Merge pull request 'Call dispatcher' (#26) from feat/call-dispatcher into main

Reviewed-on: #26

README.md

@@ -18,6 +18,7 @@ This framework is being developed as part of a Bachelor's Thesis by Louis Herede
   - [Highlighting](#highlighting)
   - [Dumping the AST](#dumping-the-ast)
   - [Dumping the Registry](#dumping-the-registry)
+  - [Generating Stubs](#generating-stubs)
   - [Showing Type Judgements](#showing-type-judgements)
   - [Validating Definitions](#validating-definitions)
 - [Tests](#tests)
@@ -116,6 +117,14 @@ midas dump-registry -t types.midas
 
 This command processes the given Midas definitions and dumps the contents of the types registry.
 
+### Generating Stubs
+
+```shell
+midas stubs types.midas -o stubs.pyi
+```
+
+This command generate Python stubs from a Midas definition file
+
 ### Showing Type Judgements
 
 ```shell

assets/icon.svg

@@ -0,0 +1,117 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<!-- Created with Inkscape (http://www.inkscape.org/) -->
+
+<svg
+   width="128"
+   height="128"
+   viewBox="0 0 128 128"
+   version="1.1"
+   id="svg1"
+   inkscape:export-filename="logo.png"
+   inkscape:export-xdpi="96"
+   inkscape:export-ydpi="96"
+   inkscape:version="1.4.4 (1:1.4.4+202605061436+dcaf3e7d9e)"
+   sodipodi:docname="logo.svg"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns:xlink="http://www.w3.org/1999/xlink"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg">
+  <sodipodi:namedview
+     id="namedview1"
+     pagecolor="#ffffff"
+     bordercolor="#000000"
+     borderopacity="0.25"
+     inkscape:showpageshadow="2"
+     inkscape:pageopacity="0.0"
+     inkscape:pagecheckerboard="0"
+     inkscape:deskcolor="#d1d1d1"
+     inkscape:document-units="mm"
+     showgrid="true"
+     inkscape:zoom="1.9332778"
+     inkscape:cx="-8.2760999"
+     inkscape:cy="112.2446"
+     inkscape:window-width="2584"
+     inkscape:window-height="1028"
+     inkscape:window-x="0"
+     inkscape:window-y="24"
+     inkscape:window-maximized="1"
+     inkscape:current-layer="layer1">
+    <inkscape:grid
+       id="grid1"
+       units="px"
+       originx="0"
+       originy="0"
+       spacingx="4"
+       spacingy="4"
+       empcolor="#0099e5"
+       empopacity="0.30196078"
+       color="#0099e5"
+       opacity="0.14901961"
+       empspacing="4"
+       enabled="true"
+       visible="true" />
+  </sodipodi:namedview>
+  <defs
+     id="defs1">
+    <linearGradient
+       inkscape:collect="always"
+       xlink:href="#linearGradient4689"
+       id="linearGradient1478"
+       gradientUnits="userSpaceOnUse"
+       gradientTransform="matrix(0.562541,0,0,0.567972,-9.399749,-5.305317)"
+       x1="26.648937"
+       y1="20.603781"
+       x2="135.66525"
+       y2="114.39767" />
+    <linearGradient
+       id="linearGradient4689">
+      <stop
+         style="stop-color:#e1be1e;stop-opacity:1;"
+         offset="0"
+         id="stop4691" />
+      <stop
+         style="stop-color:#ffeb82;stop-opacity:1;"
+         offset="1"
+         id="stop4693" />
+    </linearGradient>
+    <linearGradient
+       inkscape:collect="always"
+       xlink:href="#linearGradient4671"
+       id="linearGradient1475"
+       gradientUnits="userSpaceOnUse"
+       gradientTransform="matrix(0.562541,0,0,0.567972,-9.399749,-5.305317)"
+       x1="150.96111"
+       y1="192.35176"
+       x2="112.03144"
+       y2="137.27299" />
+    <linearGradient
+       id="linearGradient4671">
+      <stop
+         style="stop-color:#ffdc21;stop-opacity:1;"
+         offset="0"
+         id="stop4673" />
+      <stop
+         style="stop-color:#ffeb82;stop-opacity:1;"
+         offset="1"
+         id="stop4675" />
+    </linearGradient>
+  </defs>
+  <g
+     inkscape:label="Calque 1"
+     inkscape:groupmode="layer"
+     id="layer1">
+    <g
+       id="g1"
+       transform="translate(2.911719,3.414527)">
+      <path
+         style="fill:url(#linearGradient1478);fill-opacity:1"
+         d="m 60.510156,6.3979729 c -4.583653,0.021298 -8.960939,0.4122177 -12.8125,1.09375 C 36.35144,9.4962267 34.291407,13.691825 34.291406,21.429223 v 10.21875 h 26.8125 v 3.40625 h -26.8125 -10.0625 c -7.792459,0 -14.6157592,4.683717 -16.7500002,13.59375 -2.46182,10.212966 -2.5710151,16.586023 0,27.25 1.9059283,7.937852 6.4575432,13.593748 14.2500002,13.59375 h 9.21875 v -12.25 c 0,-8.849902 7.657144,-16.656248 16.75,-16.65625 h 26.78125 c 7.454951,0 13.406253,-6.138164 13.40625,-13.625 v -25.53125 c 0,-7.266339 -6.12998,-12.7247775 -13.40625,-13.9375001 -4.605987,-0.7667253 -9.385097,-1.1150483 -13.96875,-1.09375 z m -14.5,8.2187501 c 2.769547,0 5.03125,2.298646 5.03125,5.125 -2e-6,2.816336 -2.261703,5.09375 -5.03125,5.09375 -2.779476,-1e-6 -5.03125,-2.277415 -5.03125,-5.09375 -1e-6,-2.826353 2.251774,-5.125 5.03125,-5.125 z"
+         id="path1948" />
+      <path
+         style="fill:url(#linearGradient1475);fill-opacity:1"
+         d="m 91.228906,35.054223 v 11.90625 c 0,9.230755 -7.825895,16.999999 -16.75,17 h -26.78125 c -7.335833,0 -13.406249,6.278483 -13.40625,13.625 v 25.531247 c 0,7.26634 6.318588,11.54032 13.40625,13.625 8.487331,2.49561 16.626237,2.94663 26.78125,0 6.750155,-1.95439 13.406253,-5.88761 13.40625,-13.625 V 92.897973 h -26.78125 v -3.40625 h 26.78125 13.406254 c 7.79246,0 10.69625,-5.435408 13.40624,-13.59375 2.79933,-8.398886 2.68022,-16.475776 0,-27.25 -1.92578,-7.757441 -5.60387,-13.59375 -13.40624,-13.59375 z m -15.0625,64.65625 c 2.779478,3e-6 5.03125,2.277417 5.03125,5.093747 -2e-6,2.82635 -2.251775,5.125 -5.03125,5.125 -2.76955,0 -5.03125,-2.29865 -5.03125,-5.125 2e-6,-2.81633 2.261697,-5.093747 5.03125,-5.093747 z"
+         id="path1950" />
+    </g>
+  </g>
+</svg>

docs/manual.typ

@@ -0,0 +1,697 @@
+//#import "@preview/codly:1.3.0": codly, codly-init
+// Fix unaligned highlights in v0.15.0 ()
+// See https://github.com/Dherse/codly/pull/132
+#import "@local/codly:1.3.1": codly, codly-init
+
+#import "@preview/codly-languages:0.1.10": codly-languages
+#import "template.typ": TODO, project
+#import "@preview/gentle-clues:1.3.1" as gc
+
+#let midas-version = toml("../pyproject.toml").project.version
+#let head-ref = read("../.git/HEAD").split(":").at(1).trim()
+#let commit-hash = read("../.git/" + head-ref).slice(0, 8)
+
+#show: project.with(
+  title: [Midas User Manual],
+  author: "Louis Heredero",
+  version: midas-version,
+  hash: commit-hash,
+  icon-path: path("../assets/icon.svg"),
+)
+
+#show: codly-init
+#codly(
+  languages: codly-languages
+    + (
+      midas: (
+        name: "Midas",
+        color: rgb("#eedd47"),
+        icon: box(
+          image(
+            "../assets/icon.svg",
+            height: 130%,
+            fit: "contain",
+          ),
+        ),
+      ),
+    ),
+)
+
+= Introduction
+
+Python is a very popular programming language, especially in data sciences.
+However, it has been designed for simplicity, distancing itself from typed languages such as Java or C to embrace dynamic typing.
+What this means is that in Python, type checks are deferred to runtime when operations are concretely executed.
+
+For developers, it might seem like a great way of simplifying the language and making it very flexible, but it does come with a cost.
+Indeed, type errors are very easy to make in Python. While passing an integer where a string is expected might not be an issue in some cases, these are the sort of thing that can cause crashes or incorrect results without a clear diagnostic to help the user fix it.
+
+Fortunately, developers using IDEs or properly configured text editors can benefit from external type checkers such as MyPy which will perform static type analysis of their Python code. Some can also be configured to be very strict, forcing the user to make the whole code typeable statically, thus avoiding any runtime type errors.
+
+This is not the end of the problem though. Some parts of a program, especially in data related fields, may not be available at "compile-time". For example, a dataset can be loaded from an external file, or data can be fetched from an API, with no guarantees of having the expected format when analyzing the code statically.
+
+In turn, that can cause a range of loud and silent errors at runtime. A malformed number will probably crash the program when trying to convert it, but a NaN in a series of value might just produce wrong results without any exception. Combine this with often long-running data-processing pipelines and this is how developers can waste hours of precious computation time.
+
+Midas is a type system which can be used on top of Python to provide better type checking capabilities and gradual typing.
+It aims at providing optional but strict type annotations and casting operations which can produce runtime assertions. It also allows the user to define dependent types with value constraints that are translated into runtime checks.
+
+= Installation
+
+Midas comes as a very light Python package that you can install on your system in a few simple steps.
+
+== Requirements
+
+Here below are the requirements for installing Midas. All Python dependencies will be installed by `uv` in the installation process described in @install-steps.
+
+- Python 3.11+
+- `uv`
+
+== Steps <install-steps>
+
+1. Clone the repository
+  ```bash
+  git clone https://git.kb28.ch/HEL/midas.git
+  ```
+2. Navigate inside the directory
+  ```bash
+  cd midas
+  ```
+3. Install Midas as a tool in your local user space
+  ```bash
+  uv tool install .
+  ```
+
+And that's it ! You can now use Midas commands anywhere, like this:
+```bash
+midas --help
+```
+
+= Quick Start
+
+This chapter will give you the keys to quickly start using Midas in your project.
+
+== Defining custom types
+
+To begin with, you might want to define some custom types for your project, to avoid handling anonymous float values everywhere. To do so, create a `*.midas` file in your project, and write some definitions for your types. See @midas-ref for more information on syntax and features.
+
+@qs-midas shows a simple example of what it might look like.
+
+#codly(header: [types.midas])
+
+#figure(
+  ```midas
+  type Meter = float
+  extend Meter {
+    def __add__: fn(Meter, /) -> Meter
+    def __sub__: fn(Meter, /) -> Meter
+  }
+
+  type Coordinate = object
+  extend Coordinate {
+    prop x: Meter
+    prop y: Meter
+  }
+  ```,
+  caption: [Example Midas type definitions],
+) <qs-midas>
+
+You can check for any syntax error using the following command:
+```bash
+midas validate types.midas
+```
+
+When you are happy with your definitions, you can generate Python stubs to use in your source code. This allows other type checkers like MyPy to recognize your custom types and avoid reporting them as undefined. It can also help catch some type errors in your IDE.
+```bash
+midas stubs types.midas -o stubs.pyi
+```
+
+This command will generate a file as shown in @qs-stubs, providing stub classes to represent the type lattice including methods and properties.
+
+#codly(header: [stubs.pyi])
+
+#figure(
+  ```pyi
+  from __future__ import annotations
+
+  class Meter(float):
+      def __add__(self, _0: Meter, /) -> Meter: ...
+      def __sub__(self, _0: Meter, /) -> Meter: ...
+
+  class Coordinate(object):
+      x: Meter
+      y: Meter
+  ```,
+  caption: [Generated stubs from example definitions of @qs-midas],
+) <qs-stubs>
+
+== Using Midas in Python
+
+You can now write your Python program as you would normally. You can import your custom types from the generated stubs file and use them in type annotations.
+
+You can also import the `cast` and `unsafe_cast` functions from `midas.typing` to explicitly cast a value to a specific type (see @cast for more information).
+
+An example Python script is shown in @qs-python, demonstrating how you can use custom types in type annotations. Notice the comments describing errors that will be caught by the type checker in @qs-type-checking.
+
+#codly(header: [script.py])
+
+#figure(
+  ```python
+  from lib import load_coordinate
+  from midas.typing import cast
+  from stubs import Coordinate, Meter
+
+  p1 = cast(Coordinate, load_coordinate(0))
+  p2 = cast(Coordinate, load_coordinate(1))
+
+  diff_x = p2.x - p1.x
+  diff_y = p2.y - p1.y
+
+  dist = diff_x + diff_y
+
+  p2.x += cast(Meter, 1)
+  p2.y = True  # invalid, wrong type
+  p2.z = 3  # invalid, no property 'z' on Coordinate
+  p2.x.a = 3  # invalid, no properties on Meter
+  ```,
+  caption: [Example Python script],
+) <qs-python>
+
+== Type checking <qs-type-checking>
+
+Now that you have defined some types and written a script, you can run the type checker with the following command. You can also skip this step and directly run the compilation command in @qs-compilation.
+
+```bash
+midas check -t types.midas script.py
+```
+
+== Compiling <qs-compilation>
+
+The final step is to compile your code. This step will produce a runnable Python script, including runtime assertions generated by `cast` expressions.
+
+```bash
+midas compile -t types.midas script.py
+python3 build/midas/script.py
+```
+
+= Midas Language Reference <midas-ref>
+
+In this chapter, you will find a complete reference for the Midas definition language.
+
+A `*.midas` file contains a number of statements, which can be:
+- *`type`* statements (see @type-stmt): to define a new type
+- *`extend`* statements (see @extend-stmt): to define member of a type
+- *`predicate`* statements (see @predicate-stmt): to define named predicates that can be used in constraint types
+
+== Type Statement <type-stmt>
+
+A *`type`* statement lets you define a new type. It requires a unique name and base type.
+
+The simplest form of a *`type`* statement is:
+#figure(
+  ```midas
+  type MyType = float
+  ```,
+  caption: [Simple `type` statement declaring a new type "`MyType`" as a subtype of `float`],
+) <midas-simple-alias>
+
+This statement defines a new type called `MyType` which is a subtype of `float`. `MyType` is a `float` but a `float` is not necessarily `MyType`.
+
+=== Builtin / base types
+
+A number of base types are provided out of the box, which can be used to derive other types.
+
+They correspond to Python's builtin types:
+```py object```,
+```py str```,
+```py float```,
+```py int```,
+```py bool```,
+```py list```,
+```py dict```,
+```py None```.
+
+Some differences are to be noted however.
+1. ```py bool``` is not a subtype of ```py int```
+2. ```py list``` are homogeneous, i.e. all items must be of the same type
+3. ```py dict``` keys and values are homogeneous, i.e. all keys must be of the same type and all values must be of the same type (can be different from keys).
+
+=== Function types
+
+A function type is written in a similar notation to Python function definitions:
+#figure(
+  ```midas
+  type Repeater = fn(text: str, count: int) -> str
+  ```,
+  caption: [Simple function type definition],
+)
+
+Midas supports positional-only, keyword-only and mixed arguments (using the `/` and `*` separators). You may omit the name of positional-only arguments. The return type is required.
+
+Optional parameters can be indicated by adding a question mark (`?`) after their type:
+#figure(
+  ```midas
+  type Repeater = fn(text: str, count: int, *, sep: str?) -> str
+  ```,
+  caption: [Function type definition with an optional keyword-only parameter],
+)
+
+#gc.warning[
+  Sink arguments (`*args`, `**kwargs`) are not currently supported.
+]
+
+=== Constraint types
+
+A useful feature provided by Midas is the possibility to combine types with custom value constraints. For example, you might want to define a type for positive amounts of money:
+
+#figure(
+  ```midas
+  type Money = float
+  type Income = Money where _ >= 0
+  ```,
+  caption: [Simple constraint type definition],
+)
+
+Constraints can be combined with any type using the `where` keyword, followed by a constraint expression (see @constraint-expr).
+
+=== Generic types
+
+For more complex types, you might want to use type parameters. For example, to define a container, we might write:
+#figure(
+  ```midas
+  type Container[T] = object
+  ```,
+  caption: [Simple generic container type definition],
+)
+
+To better refine a generic type, you can also bound type parameters using the following syntax:
+#figure(
+  ```midas
+  type Container[T <: float] = object
+  ```,
+  caption: [Generic container type definition with a bound],
+)
+
+This can be read as "`Container` is a generic type which takes one type parameter `T` that must be a subtype of `float`".
+
+You can use a generic type, i.e. instantiate it, by using a similar syntax with concrete type as arguments:
+
+#figure(
+  ```midas
+  type MyContainer = Container[MyType]
+  ```,
+  caption: [Application of a generic type],
+)
+
+Generic types can also take multiple parameters, which are then separated by commas:
+#figure(
+  ```midas
+  type ZipCodeRegistry = dict[int, str]
+  ```,
+  caption: [Application of a multi-parameter generic type],
+)
+
+The _body_ of a generic type, i.e. the right-hand side of the definition, can contain or even be equal to any number of its parameters.#footnote[The latter is not something that is expressible in standard Python, yet it brings a semantic distinction on top of structurally equivalent values.] For example, the following is a valid type statement:
+#figure(
+  ```midas
+  type Price[T <: Currency] = T where _ > 0
+  ```,
+  caption: [Type parameters in a generic type's body],
+)
+
+#pagebreak()
+
+== Extend Statement <extend-stmt>
+
+Type statements allow you to define new types, kind of like type aliases. However, a type might have properties or methods of its own. These might override those of the parent type or be brand new members.
+
+This is where the `extend` statement comes into play. It allows defining members on a given type. Members can either be properties (`prop`) or methods (`def`). The only difference between the two is that methods must be functions and can be overloaded.
+
+Here is a simple example showing how to define a property and a method on a custom type:
+#figure(
+  ```midas
+  type MyType = float
+  extend MyType {
+    prop norm: float
+    def double: fn() -> MyType
+  }
+  ```,
+  caption: [Simple `extend` statement defining a property and a method],
+)
+
+An `extend` statement can appear anywhere after the type it extends has been defined.
+You may want to override Python's dunder methods to implement type checking for some basic operators, like `__add__` for the `+` operator.
+
+#figure(
+  ```midas
+  type Money = float
+  extend Money {
+    def __add__(Money, /) -> Money
+    def __mul__(float, /) -> Money
+  }
+  ```,
+  caption: [Simple `extend` statement overriding some dunder methods],
+)
+
+When extending generic type, you must specify the whole type, including its parameter(s):
+#figure(
+  ```midas
+  type Container[T <: float] = object
+  extend Container[T <: float] {
+    prop content: T
+    def set_content: fn(content: T) -> None
+  }
+  ```,
+  caption: [Generic `extend` statement using type parameters in the declared members],
+)
+
+#pagebreak()
+
+== Predicate Statement <predicate-stmt>
+
+A *`predicate`* statement lets you define a named constraint expression, like a function, which can then be used in other constraint expressions (either in other predicate statements or in constraint types). See @constraint-expr for more information about the syntax of constraint expressions.
+
+The left-hand side of a predicate statement is written as a function signature, without a return type. The right-hand side is a constraint expression. For example:
+#figure(
+  ```midas
+  predicate is_positive(v: float) = v >= 0
+  ```,
+  caption: [Simple `predicate` statement defining an `is_positive` predicate],
+)
+
+The left-hand side can also be curried to allow partial application. For example:
+#figure(
+  ```midas
+  predicate in_range(mn: float, mx: float)(v: float) = mn <= v & v <= mx
+  predicate is_ratio = in_range(0.0, 1.0)
+  ```,
+  caption: [Curried `predicate` statement and partial application],
+) <midas-predicate-partial>
+
+Notice that the second predicate statement doesn't take any parameters. This is simply a partial application of another predicate, kind of like an alias. You can use it in other expressions to finalize the call:
+#figure(
+  ```midas
+  type Efficiency = float where is_ratio(_)
+  ```,
+  caption: [Constraint type definition using the partially applied predicate from @midas-predicate-partial],
+)
+
+Of course you can also directly call `in_range`:
+#figure(
+  ```midas
+  type Efficiency = float where in_range(0.0, 1.0)(_)
+  ```,
+  caption: [Full call of curried predicate from @midas-predicate-partial],
+)
+
+When compiled, named predicates are translated to Python functions which are used in runtime assertions. Only predicates that are referenced are compiled.
+
+#pagebreak()
+== Constraint Expressions <constraint-expr>
+
+*Constraint expressions* are Python-like expressions which can appear in *`predicate`* statements or in constraint types.
+
+They can contain comparisons, simple computations, logical operations and must evaluate to a boolean value.
+
+Context is quite restricted inside these expressions. You can only reference some builtin functions, such as type constructors (`float(...)`, `str(...)`, etc.), parameters of predicate statements, and named predicates. In constraint type, the special variable `_` can be used to reference the value targeted by the type. For example:
+
+#figure(
+  ```midas
+  predicate not_nan(v: float) = v != float("nan")
+  type RealFloat = float where not_nan(_)
+  ```,
+  caption: [Example constraint expressions],
+) <ex-constraint-expr>
+
+In the predicate statement (@ex-constraint-expr:1), we reference the parameter `v` and the builtin `float` function.
+In the constraint type definition (@ex-constraint-expr:2), we then reference the named predicate `not_nan`, passing the value targeted by the type itself ( `_` )
+
+= Supported Python Syntax <python-ref>
+
+Midas integrates naturally in Python via type annotations. Through generated stubs, even other type checker can detect your custom types (see @cmd-stubs).
+
+It has been designed to leave the user free of typing any amount of their code but be strict about the parts that are annotated. By default, any untyped Python expression is assigned `UnknownType`.
+
+Any operation is permitted on `UnknownType` and will result in `UnknownType` values.
+
+The moment an expression can be typed, that be thanks to an annotation or a literal value, the type checker kicks in and will validate your statements.
+
+Because Python is very flexible language with many features, some expressions and statements might be more complex to properly type check, thus only a subset of the Python language is fully supported. This chapter lists all supported features of Python and how they affect type checking.
+
+Some examples are presented in the following sections in the form of code blocks. Highlights in the code blocks indicate the type assigned to each expression by the type checker. Some types may be omitted for readability. For example:
+
+#codly(
+  highlights: (
+    (
+      line: 1,
+      start: 5,
+      fill: green,
+      tag: [_int_],
+    ),
+    (
+      line: 2,
+      start: 7,
+      end: 7,
+      fill: green,
+      tag: [_int_],
+    ),
+  ),
+)
+
+```python
+v = 3
+print(v)
+```
+
+== Literals
+
+Literal Python values are type checked using builtin types. Lists and dictionaries of literals are also typed liked literals. This does not include comprehension lists/dicts (```py [. for . in .]```), nor formatted strings (```py f"..."```). @supported-literals shows the list of supported literal values and their type.
+
+#let supported-literals = table(
+  columns: 2,
+  table.header[*Example value*][*Judged Type*],
+  ```py 42```, ```py int```,
+  ```py 3.14```, ```py float```,
+  ```py True```, ```py bool```,
+  ```py "Midas"```, ```py str```,
+  ```py None```, ```py None```,
+  ```py [1, 2, 3]```, ```py list[int]```,
+  ```py {1: "One", 2: "Two"}```, ```py dict[int, str]```,
+  ```py ("1", 1, True)```, ```py tuple[str, int, bool]```,
+)
+
+#figure(
+  supported-literals,
+  caption: [Supported literal values and their judged types],
+) <supported-literals>
+
+== Assignments
+
+Variable assignments allow assigning a new value to a variable. For the type checker, this implies two things:
+1. If the variable was not already declared in the current scope, it is declared at that point with the type of the right-hand side expression
+2. If the variable was already declared, the type of the right-hand side expression is checked against the declared type of the variable. Only a subtype of the variable's type can be assigned to it
+
+Once a variable has been given a type, it cannot be changed in the same scope.
+
+The walrus operator (```py :=```) is not currently supported.
+
+A simple annotation declaration, without assigning a value, is enough to declare a variable. For example:
+#figure(
+  ```python
+  var: float
+  ```,
+  caption: [Bare Python variable annotation without assignment],
+)
+
+Because unpacking is not supported, assigning to multiple values is also not handled by the type checker.
+
+== Arithmetic
+
+- All basic binary operators are supported, through dunder methods.
+- All comparison operators except ```py in``` are supported.
+- All unary operators are supported (`+`, `-`, `~`).
+- All logical operators are supported (```py and```, ```py or```, ```py not```).
+
+== Ternary operator
+
+The ternary operator ```py . if . else .``` is supported. As for `if` statements (see @if-else), the test expression must be a boolean. Additionally, both branches must be of the same type.
+
+For example:
+#codly(
+  highlights: (
+    (
+      line: 1,
+      start: 10,
+      end: 44,
+      tag: [_str_],
+      fill: blue,
+    ),
+    (
+      line: 1,
+      start: 11,
+      end: 16,
+      tag: [_str_],
+      fill: green,
+    ),
+    (
+      line: 1,
+      start: 39,
+      end: 43,
+      tag: [_str_],
+      fill: green,
+    ),
+    (
+      line: 1,
+      start: 21,
+      end: 32,
+      tag: [_bool_],
+      fill: green,
+    ),
+  ),
+)
+#figure(
+  ```python
+  parity = ("even" if num % 2 == 0 else "odd")
+  ```,
+  caption: [Typing of ternary operator],
+)
+
+== Control flow
+
+Some control flow features are supported. For the limited code of this project, not all constructs are supported. The following are those currently handled and typ checked by Midas.
+
+=== `if` / `elif` / `else` <if-else>
+
+Conditional statements are checked relatively strictly by Midas. The test expression, i.e. what comes after the ```py if``` keyword, must be a boolean. While Python allows introducing and leaking new variables from inside an ```py if``` statement, Midas will strictly forbid leaks by restraining bindings to the scope they are defined in. For example, the following Python code will not compile with Midas:
+#figure(
+  ```python
+  age = 22
+  if age >= 18:
+      msg = "You're an adult"
+  else:
+      msg = "You're still a child"
+  print(msg)  # -> unknown variable 'msg'
+  ```,
+  caption: [`if`/`else` statement cannot leak variables],
+)
+
+=== `for` loops
+
+Simple forms of `for` loops can be used, that is using a single variable and iterating over an object implementing the `__getitem__` method. Like above in @if-else, leaking variables from inside the loop is ignored.
+
+The `for`-`else` statements are not supported. `while` loops are also not not supported.
+
+== Functions
+
+You can define functions as usual and the type checker will do its best to type it. Apart from argument sinks (`*args`, `**kwargs`), all forms of parameter specifications are supported (positional-only, keyword-only, mixed, optional).
+
+As for the rest of your code, type annotations are optional, but recommended. If you omit the return type hint, the type checker will try to infer it from the function body and its return statements. If you did specify a return type, all return paths must return values that are subtypes of the type hint.
+
+#codly(
+  highlights: (
+    (
+      line: 2,
+      start: 12,
+      end: 16,
+      tag: [_float_],
+      fill: green,
+    ),
+    (
+      line: 2,
+      start: 12,
+      tag: [_float_],
+      fill: blue,
+    ),
+    (
+      line: 3,
+      start: 10,
+      end: 15,
+      tag: [_(value: float) -> float_],
+      fill: green,
+    ),
+    (
+      line: 3,
+      start: 17,
+      end: 19,
+      tag: [_float_],
+      fill: green,
+    ),
+    (
+      line: 3,
+      start: 10,
+      tag: [_float_],
+      fill: blue,
+    ),
+  ),
+)
+#figure(
+  ```python
+  def double(value: float) -> float:
+      return value * 2
+  result = double(4.0)
+  ```,
+  caption: [Typing of function's body and call],
+)
+
+Anonymous functions (```py lambda```) are not yet supported
+
+== Casts <cast>
+
+#gc.info[
+  The functions discussed in this section are provided by the `midas.typing` submodule. You can import them in your script like so:
+  #figure(
+    ```python
+    from midas.typing import cast, unsafe_cast
+    ```,
+    caption: [Importing cast functions],
+  )
+]
+
+Sometimes, you may want to use a value whose type is not known to the type checker in a place where it expects a particular type. In that case, if you do know that the runtime type will correspond to what is expected, you can use a `cast` expression.
+
+Similar to the `cast` function from the `typing` package of Python's Standard Library, it allows telling the type checker that a value has a given type. While `typing`'s function doesn't have any runtime side-effect, Midas' will generate runtime assertions, ensuring that your statement is true when running the code. What cannot be checked statically is checked at runtime.
+
+In the following example, a runtime check would be generated to ensure that the value is indeed a `float` and that it satisfies the type's constraint (i.e. `>= 0`):
+
+#codly(
+  highlights: (
+    (
+      line: 1,
+      start: 35,
+      end: 47,
+      tag: [_UnknownType_],
+      fill: red,
+    ),
+    (
+      line: 2,
+      start: 7,
+      end: 17,
+      tag: [_PositiveFloat_],
+      fill: green,
+    ),
+  ),
+)
+#figure(
+  ```python
+  typed_value = cast(PositiveFloat, unknown_value)
+  print(typed_value)
+  ```,
+  caption: [Typing of `cast` expression],
+)
+
+There may be some cases where the cost of checking a value at runtime is simply not worth the safety, for example when dealing with a big dataset. If do wish so, you can use `unsafe_cast` which will only tell the type checker the type of the value, without generating a runtime assertion. This maps to the default behavior of `typing`'s own `cast` function.
+
+If the value passed to `cast` or `unsafe_cast` is a literal (e.g. an integer, a string, a list of literals, etc.), the assertion is evaluated _at compile-time_ and no runtime assertion is generated.
+
+= Commands <commands>
+
+#TODO
+
+== Type Checking (`check`) <cmd-check>
+== Compiling (`compile`) <cmd-compile>
+== Formatting (`format`) <cmd-format>
+== Highlighting (`highlight`) <cmd-highlight>
+== Dumping the AST (`parse`) <cmd-parse>
+== Dumping the Registry (`dump-registry`) <cmd-registry>
+== Generating Stubs (`stubs`) <cmd-stubs>
+== Showing Type Judgements (`types`) <cmd-types>
+== Validating Definitions (`validate`) <cmd-validate>

docs/midas.sublime-syntax

@@ -0,0 +1,180 @@
+%YAML 1.2
+---
+name: Midas
+file_extensions:
+  - midas
+scope: source.midas
+
+variables:
+  identifier: "[a-zA-Z_][a-zA-Z0-9_]*"
+
+contexts:
+  prototype:
+    - include: comments
+
+  main:
+    - include: keywords
+    - include: types
+
+  comments:
+    - match: "//"
+      scope: punctuation.definition.comment.midas
+      push:
+        - meta_scope: comment.line.midas
+        - match: $
+          pop: true
+    - match: /\*
+      scope: punctuation.definition.comment.midas
+      push:
+        - meta_scope: comment.block.midas
+        - match: \*/
+          pop: true
+
+  string:
+    - meta_include_prototype: false
+    - meta_scope: string.quoted.double.c
+    - match: '"'
+      pop: true
+
+  keywords:
+    - match: \btype\b
+      scope: keyword.declaration.midas
+      push: type-stmt
+    - match: \bextend\b
+      scope: keyword.declaration.midas
+      push: extend-stmt
+    - match: \bpredicate\b
+      scope: keyword.declaration.midas
+      push: predicate-stmt
+
+  type-stmt:
+    - match: "{{identifier}}"
+      scope: entity.name.type
+    - match: \[
+      push: type-params
+    - match: "="
+      scope: keyword.operator.equal.midas
+      push: type-expr
+    - match: $
+      pop: true
+
+  type-expr:
+    - match: \b(fn)\s*(\()
+      captures:
+        1: keyword.other.midas
+        2: punctuation.section.group.begin
+      push: fn-params
+    - match: \b(where)\b
+      scope: keyword.other.midas
+      set: constraint
+    - match: "{{identifier}}"
+      scope: entity.name.type
+    - match: $
+      pop: 2
+
+  fn-params:
+    - match: "({{identifier}})(:)"
+      captures:
+        1: variable.parameter.midas
+        2: punctuation.separator.annotation.midas
+      push:
+        - include: type-expr
+        - match: \?
+          scope: keyword.operator.qmark.midas
+        - match: "(?=,)"
+          scope: punctuation.separator.midas
+          pop: true
+        - match: '(?=\))'
+          pop: true
+    - include: type-expr
+    - match: '\)'
+      set:
+        - match: "->"
+          scope: keyword.operator.arrow.midas
+          set: type-expr
+
+  constraint:
+    - match: $
+      pop: 2
+
+    - match: \d+(\.\d+)?
+      scope: constant.numeric.midas
+
+    - match: \b(true|false|none)\b
+      scope: constant.language.midas
+
+    - match: '"'
+      push: string
+
+    - match: (<=|>=|<|>|==|!=|&)
+      scope: keyword.operator
+
+    - match: _
+      scope: variable.language.midas
+
+    - match: '{{identifier}}(?=\s*\()'
+      scope: variable.function.midas
+
+    - match: "{{identifier}}"
+      scope: variable.other.readwrite.midas
+
+  type-params:
+    - match: "<:"
+      scope: keyword.operator.subtype.midas
+    - match: "[a-zA-Z][a-zA-Z_0-9]*"
+      scope: entity.name.type
+    - match: "]"
+      pop: true
+
+  extend-stmt:
+    - match: "{{identifier}}"
+      scope: entity.name.type
+    - match: \[
+      push: type-params
+    - match: \{
+      scope: punctuation.section.block.begin
+      set: extend-body
+
+  extend-body:
+    - include: member-stmt
+    - match: \}
+      scope: punctuation.section.block.end
+      pop: true
+
+  member-stmt:
+    - match: \b(prop|def)\b
+      scope: keyword.other.midas
+      push:
+        - match: "{{identifier}}"
+          scope: variable.other.member
+        - match: ":"
+          push: type-expr
+        - match: $
+          pop: true
+
+  predicate-stmt:
+    - match: "{{identifier}}"
+      scope: entity.name.function.midas
+    - match: '\('
+      push: predicate-params
+    - match: "="
+      scope: keyword.operator.equal.midas
+      set: constraint
+    - match: $
+      pop: true
+
+  predicate-params:
+    - match: "({{identifier}})(:)"
+      captures:
+        1: variable.parameter.midas
+        2: punctuation.separator.annotation.midas
+      push:
+        - include: type-expr
+        - match: "(?=,)"
+          scope: punctuation.separator.midas
+          pop: true
+        - match: '(?=\))'
+          pop: true
+
+    - match: '\)'
+      pop: true

docs/template.typ

@@ -0,0 +1,143 @@
+#import "@preview/modpattern:0.2.0": modpattern
+
+#let TODO = block(
+  width: 6em,
+  height: 3em,
+  stroke: red,
+  fill: modpattern(
+    size: (10pt, 10pt),
+    line(
+      start: (0%, 0%),
+      end: (100%, 100%),
+      stroke: gray.transparentize(60%) + 2pt,
+    ),
+  ),
+  align(
+    center + horizon,
+    text(fill: red, size: 1.5em)[*TODO*],
+  ),
+)
+
+#let _render-header(version, hash) = {
+  let last-heading = query(heading.where(level: 1).before(here())).last(default: none)
+  let next-heading = query(heading.where(level: 1).after(here())).first(default: none)
+
+  let current-heading = if next-heading != none and next-heading.location().page() == here().page() {
+    next-heading
+  } else if last-heading != none {
+    last-heading
+  } else { none }
+
+  let chapter = if current-heading != none {
+    let body = current-heading.body
+    if current-heading.numbering != none {
+      let num = counter(heading).display(current-heading.numbering, at: current-heading.location())
+      body = [#num #body]
+    }
+    body
+  } else []
+
+  grid(
+    columns: (1fr, auto, 1fr),
+    align: (left, center, right),
+    document.title, [v#version - #hash], chapter,
+  )
+}
+
+#let _unshift-prefix(prefix, content) = context {
+  pad(left: -measure(prefix).width, prefix + content)
+}
+
+#let project(
+  title: none,
+  author: none,
+  version: "0.0.1",
+  hash: "abcdefgh",
+  icon-path: none,
+  doc,
+) = {
+  assert(title != none, message: "Please provide a title")
+
+  set document(
+    title: title,
+    author: author,
+  )
+  set text(
+    font: "Source Sans 3",
+  )
+
+  set raw(syntaxes: path("midas.sublime-syntax"))
+
+  let front-page() = {
+    align(center)[
+      #{
+        set text(size: 1.5em)
+        std.title()
+      }
+
+      v#version - #hash
+
+      #if icon-path != none {
+        v(1cm)
+        image(icon-path)
+      }
+    ]
+    pagebreak()
+  }
+
+  let outlines() = {
+    outline()
+    pagebreak()
+    outline(
+      title: [List of Listings],
+      target: figure.where(kind: raw),
+    )
+
+    outline(
+      title: [List of Tables],
+      target: figure.where(kind: table),
+    )
+  }
+
+  let main() = {
+    // Adapted from https://github.com/hei-templates/hei-synd-thesis/blob/7d2b941197babae0bf3afd4e5914754e09a64001/lib/template-thesis.typ#L242-L261
+    show heading.where(level: 1): it => {
+      pagebreak()
+
+      set text(size: 1.5em)
+      set block(above: 1.2em, below: 1.2em)
+      if it.numbering != none {
+        let num = numbering(it.numbering, ..counter(heading).at(it.location()))
+        let prefix = num + h(1em)
+        _unshift-prefix(prefix, it.body)
+      } else {
+        it
+      }
+    }
+
+    show heading.where(level: 2): it => {
+      if it.numbering != none {
+        let num = numbering(it.numbering, ..counter(heading).at(it.location()))
+        _unshift-prefix(num + h(0.8em), it.body)
+      } else {
+        it
+      }
+    }
+
+    set page(
+      header: context _render-header(version, hash),
+      footer: context if page.numbering != none {
+        align(center, counter(page).display(page.numbering, both: true))
+      },
+      numbering: "1 / 1",
+    )
+
+    show heading: set heading(numbering: "I.1.")
+    counter(page).update(1)
+    doc
+  }
+
+  front-page()
+  outlines()
+  main()
+}

examples/02_demonstration/demo.midas

@@ -0,0 +1,15 @@
+predicate in_range(min: float, max: float)(v: float) = min <= v & v <= max
+predicate is_ratio = in_range(0, 1)
+
+type Currency = float
+type Price[T <: Currency] = T where _ >= 0
+
+extend Price[T <: Currency] {
+    def __add__: fn(Price[T], /) -> Price[T]
+}
+
+type EUR = Currency
+type USD = Currency
+type CHF = Currency
+
+type Discount = float where is_ratio(_)

examples/02_demonstration/demo.py

@@ -0,0 +1,35 @@
+from typing import TypeVar
+
+from demo_stubs import CHF, EUR, USD, Currency, Discount, Price
+
+from midas.typing import cast, unsafe_cast
+
+T = TypeVar("T", bound=Currency)
+
+
+def apply_discount(amount: Price[T], discount: Discount) -> Price[T]:
+    return cast(Price[T], (1.0 - discount) * amount)
+
+
+a1 = cast(Price[EUR], 3.2)
+a2 = cast(Price[USD], 10.4)
+r1 = cast(Discount, 0.2)
+
+print(apply_discount(a1, r1))
+print(apply_discount(a2, r1))
+
+a3 = a1 + a1
+a4 = a1 + a2  # cannot add euros and dollars
+a3 = a2  # cannot change variable type
+
+dyn_price = float(input("Price (CHF): "))
+dyn_discount = float(input("Discount (0.0-1.0): "))
+discounted = apply_discount(
+    cast(Price[CHF], dyn_price),
+    cast(Discount, dyn_discount),
+)
+
+print(f"Discounted: CHF {discounted}")
+
+large_data = [i * 10 for i in range(100)]
+prices = unsafe_cast(list[Price[EUR]], large_data)

examples/02_demonstration/demo_stubs.pyi

@@ -0,0 +1,14 @@
+from __future__ import annotations
+from typing import Generic, TypeVar
+
+class Currency(float): ...
+
+_T0 = TypeVar("_T0", bound=Currency, covariant=True)
+
+class Price(Currency, Generic[_T0]):
+    def __add__(self, _0: Price[_T0], /) -> Price[_T0]: ...
+
+class EUR(Currency): ...
+class USD(Currency): ...
+class CHF(Currency): ...
+class Discount(float): ...

gen/midas.py

@@ -44,6 +44,11 @@ class TypeStmt:
     type: Type
 
 
+class AliasStmt:
+    name: Token
+    type: Type
+
+
 class MemberStmt:
     name: Token
     type: Type
@@ -152,4 +157,14 @@ class FunctionType:
         required: bool
 
 
+class FrameType:
+    columns: list[Column]
+
+    @dataclass(frozen=True, kw_only=True)
+    class Column:
+        location: Optional[Location] = None
+        name: Token
+        type: Type
+
+
 ###<

gen/python.py

@@ -15,7 +15,7 @@ from midas.ast.location import Location
 ###> MidasType | Type annotations | node
 class BaseType:
     base: str
-    param: Optional[MidasType]
+    args: tuple[MidasType, ...]
 
 
 class ConstraintType:
@@ -145,6 +145,7 @@ class LogicalExpr:
 class CastExpr:
     type: MidasType
     expr: Expr
+    unsafe: bool
 
 
 class TernaryExpr:
@@ -173,6 +174,10 @@ class SliceExpr:
     step: Optional[Expr]
 
 
+class TupleExpr:
+    items: tuple[Expr, ...]
+
+
 class RawExpr:
     expr: ast.expr
 

midas/ast/midas.py

@@ -51,6 +51,9 @@ class Stmt(ABC):
         @abstractmethod
         def visit_type_stmt(self, stmt: TypeStmt) -> T: ...
 
+        @abstractmethod
+        def visit_alias_stmt(self, stmt: AliasStmt) -> T: ...
+
         @abstractmethod
         def visit_member_stmt(self, stmt: MemberStmt) -> T: ...
 
@@ -71,6 +74,15 @@ class TypeStmt(Stmt):
         return visitor.visit_type_stmt(self)
 
 
+@dataclass(frozen=True)
+class AliasStmt(Stmt):
+    name: Token
+    type: Type
+
+    def accept(self, visitor: Stmt.Visitor[T]) -> T:
+        return visitor.visit_alias_stmt(self)
+
+
 @dataclass(frozen=True)
 class MemberStmt(Stmt):
     name: Token
@@ -253,6 +265,9 @@ class Type(ABC):
         @abstractmethod
         def visit_function_type(self, type: FunctionType) -> T: ...
 
+        @abstractmethod
+        def visit_frame_type(self, type: FrameType) -> T: ...
+
 
 @dataclass(frozen=True)
 class NamedType(Type):
@@ -311,3 +326,17 @@ class FunctionType(Type):
 
     def accept(self, visitor: Type.Visitor[T]) -> T:
         return visitor.visit_function_type(self)
+
+
+@dataclass(frozen=True)
+class FrameType(Type):
+    columns: list[Column]
+
+    @dataclass(frozen=True, kw_only=True)
+    class Column:
+        location: Optional[Location] = None
+        name: Token
+        type: Type
+
+    def accept(self, visitor: Type.Visitor[T]) -> T:
+        return visitor.visit_frame_type(self)

midas/ast/printer.py

@@ -105,6 +105,14 @@ class MidasAstPrinter(
             with self._child_level(single=True):
                 stmt.type.accept(self)
 
+    def visit_alias_stmt(self, stmt: m.AliasStmt) -> None:
+        self._write_line("AliasStmt")
+        with self._child_level():
+            self._write_line(f'name: "{stmt.name.lexeme}"')
+            self._write_line("type", last=True)
+            with self._child_level(single=True):
+                stmt.type.accept(self)
+
     def _print_type_param(self, param: m.TypeParam) -> None:
         self._write_line("Param")
         with self._child_level():
@@ -350,6 +358,25 @@ class MidasAstPrinter(
                 arg.type.accept(self)
             self._write_line(f"required: {arg.required}", last=True)
 
+    def visit_frame_type(self, type: m.FrameType) -> None:
+        self._write_line("FrameType")
+        with self._child_level(single=True):
+            self._write_line("columns")
+            with self._child_level():
+                for i, column in enumerate(type.columns):
+                    self._idx = i
+                    if i == len(type.columns) - 1:
+                        self._mark_last()
+                    self._print_frame_column(column)
+
+    def _print_frame_column(self, column: m.FrameType.Column) -> None:
+        self._write_line("Column")
+        with self._child_level():
+            self._write_line(f'name: "{column.name.lexeme}"')
+            self._write_line("type")
+            with self._child_level(single=True):
+                column.type.accept(self)
+
 
 class MidasPrinter(m.Expr.Visitor[str], m.Stmt.Visitor[str], m.Type.Visitor[str]):
     def __init__(self, indent: int = 4):
@@ -371,6 +398,9 @@ class MidasPrinter(m.Expr.Visitor[str], m.Stmt.Visitor[str], m.Type.Visitor[str]
         res: str = f"type {stmt.name.lexeme}{template} = {stmt.type.accept(self)}"
         return self.indented(res)
 
+    def visit_alias_stmt(self, stmt: m.AliasStmt) -> str:
+        return self.indented(f"alias {stmt.name.lexeme} = {stmt.type.accept(self)}")
+
     def _print_type_param(self, param: m.TypeParam) -> str:
         res: str = param.name.lexeme
         if param.bound is not None:
@@ -502,6 +532,23 @@ class MidasPrinter(m.Expr.Visitor[str], m.Stmt.Visitor[str], m.Type.Visitor[str]
             res += "?"
         return res
 
+    def visit_frame_type(self, type: m.FrameType) -> str:
+        res: str = self.indented("Frame[")
+        if len(type.columns) != 0:
+            res += "\n"
+            self.level += 1
+            columns: list[str] = []
+            for column in type.columns:
+                columns.append(self.indented(self._print_frame_column(column)))
+            res += ",\n".join(columns)
+            self.level -= 1
+            res += "\n"
+        res += "]"
+        return res
+
+    def _print_frame_column(self, column: m.FrameType.Column) -> str:
+        return f"{column.name.lexeme}: {column.type.accept(self)}"
+
 
 class PythonAstPrinter(
     AstPrinter,
@@ -513,7 +560,13 @@ class PythonAstPrinter(
         self._write_line("BaseType")
         with self._child_level():
             self._write_line(f"base: {node.base}")
-            self._write_optional_child("param", node.param, last=True)
+            self._write_line("args:", last=True)
+            with self._child_level():
+                for i, arg in enumerate(node.args):
+                    self._idx = i
+                    if i == len(node.args) - 1:
+                        self._mark_last()
+                    arg.accept(self)
 
     def visit_constraint_type(self, node: p.ConstraintType) -> None:
         self._write_line("ConstraintType")
@@ -757,9 +810,10 @@ class PythonAstPrinter(
             self._write_line("type")
             with self._child_level(single=True):
                 expr.type.accept(self)
-            self._write_line("expr", last=True)
+            self._write_line("expr")
             with self._child_level(single=True):
                 expr.expr.accept(self)
+            self._write_line(f"unsafe: {expr.unsafe}", last=True)
 
     def visit_ternary_expr(self, expr: p.TernaryExpr) -> None:
         self._write_line("TernaryExpr")
@@ -825,6 +879,17 @@ class PythonAstPrinter(
             self._write_optional_child("upper", expr.upper)
             self._write_optional_child("step", expr.step, last=True)
 
+    def visit_tuple_expr(self, expr: p.TupleExpr) -> None:
+        self._write_line("TupleExpr")
+        with self._child_level():
+            self._write_line("items", last=True)
+            with self._child_level():
+                for i, item in enumerate(expr.items):
+                    self._idx = i
+                    if i == len(expr.items) - 1:
+                        self._mark_last()
+                    item.accept(self)
+
     def visit_raw_expr(self, expr: p.RawExpr) -> None:
         self._write_line("RawExpr")
         with self._child_level(single=True):

midas/ast/python.py

@@ -44,7 +44,7 @@ class MidasType(ABC):
 @dataclass(frozen=True)
 class BaseType(MidasType):
     base: str
-    param: Optional[MidasType]
+    args: tuple[MidasType, ...]
 
     def accept(self, visitor: MidasType.Visitor[T]) -> T:
         return visitor.visit_base_type(self)
@@ -268,6 +268,9 @@ class Expr(ABC):
         @abstractmethod
         def visit_slice_expr(self, expr: SliceExpr) -> T: ...
 
+        @abstractmethod
+        def visit_tuple_expr(self, expr: TupleExpr) -> T: ...
+
         @abstractmethod
         def visit_raw_expr(self, expr: RawExpr) -> T: ...
 
@@ -350,6 +353,7 @@ class LogicalExpr(Expr):
 class CastExpr(Expr):
     type: MidasType
     expr: Expr
+    unsafe: bool
 
     def accept(self, visitor: Expr.Visitor[T]) -> T:
         return visitor.visit_cast_expr(self)
@@ -401,6 +405,14 @@ class SliceExpr(Expr):
         return visitor.visit_slice_expr(self)
 
 
+@dataclass(frozen=True)
+class TupleExpr(Expr):
+    items: tuple[Expr, ...]
+
+    def accept(self, visitor: Expr.Visitor[T]) -> T:
+        return visitor.visit_tuple_expr(self)
+
+
 @dataclass(frozen=True)
 class RawExpr(Expr):
     expr: ast.expr

midas/checker/builtins.midas

@@ -179,3 +179,99 @@ extend dict[K, V] {
     // def __ior__: fn(value: Iterable[tuple[K, V]], /) -> dict[K, V]
 
 }
+
+extend str {
+    def capitalize: fn() -> str
+    def casefold: fn() -> str
+    def center: fn(width: int, fillchar: str?, /) -> str
+    def count: fn(sub: str, start: None?, end: None?, /) -> int
+    def count: fn(sub: str, start: int, end: None?, /) -> int
+    def count: fn(sub: str, start: None, end: int, /) -> int
+    def count: fn(sub: str, start: int, end: int, /) -> int
+    def encode: fn(encoding: str?, errors: str?) -> bytes
+    def endswith: fn(suffix: str, start: None?, end: None?, /) -> bool
+    def endswith: fn(suffix: str, start: int, end: None?, /) -> bool
+    def endswith: fn(suffix: str, start: None, end: int, /) -> bool
+    def endswith: fn(suffix: str, start: int, end: int, /) -> bool
+    def expandtabs: fn(tabsize: int?) -> str
+    def find: fn(sub: str, start: None?, end: None?, /) -> int
+    def find: fn(sub: str, start: int, end: None?, /) -> int
+    def find: fn(sub: str, start: None, end: int, /) -> int
+    def find: fn(sub: str, start: int, end: int, /) -> int
+    // def format: fn(*args: object, **kwargs: object) -> str
+    // def format_map: fn(mapping: _FormatMapMapping, /) -> str
+    def index: fn(sub: str, start: None?, end: None?, /) -> int
+    def index: fn(sub: str, start: int, end: None?, /) -> int
+    def index: fn(sub: str, start: None, end: int, /) -> int
+    def index: fn(sub: str, start: int, end: int, /) -> int
+    def isalnum: fn() -> bool
+    def isalpha: fn() -> bool
+    def isascii: fn() -> bool
+    def isdecimal: fn() -> bool
+    def isdigit: fn() -> bool
+    def isidentifier: fn() -> bool
+    def islower: fn() -> bool
+    def isnumeric: fn() -> bool
+    def isprintable: fn() -> bool
+    def isspace: fn() -> bool
+    def istitle: fn() -> bool
+    def isupper: fn() -> bool
+    def join: fn(iterable: list[str], /) -> str  // TODO: use Iterable
+    def ljust: fn(width: int, fillchar: str?, /) -> str
+    def lower: fn() -> str
+    def lstrip: fn(chars: None?, /) -> str
+    def lstrip: fn(chars: str, /) -> str
+    def partition: fn(sep: str, /) -> tuple[str, str, str]
+    
+    def replace: fn(old: str, new: str, count: int?, /) -> str
+
+    def removeprefix: fn(prefix: str, /) -> str
+    def removesuffix: fn(suffix: str, /) -> str
+    def rfind: fn(sub: str, start: None?, end: None?, /) -> int
+    def rfind: fn(sub: str, start: int, end: None?, /) -> int
+    def rfind: fn(sub: str, start: None, end: int, /) -> int
+    def rfind: fn(sub: str, start: int, end: int, /) -> int
+    def rindex: fn(sub: str, start: None?, end: None?, /) -> int
+    def rindex: fn(sub: str, start: int, end: None?, /) -> int
+    def rindex: fn(sub: str, start: None, end: int, /) -> int
+    def rindex: fn(sub: str, start: int, end: int, /) -> int
+    def rjust: fn(width: int, fillchar: str?, /) -> str
+    def rpartition: fn(sep: str, /) -> tuple[str, str, str]
+    def rsplit: fn(sep: None?, maxsplit: int?) -> list[str]
+    def rsplit: fn(sep: str, maxsplit: int?) -> list[str]
+    def rstrip: fn(chars: None?, /) -> str
+    def rstrip: fn(chars: str, /) -> str
+    def split: fn(sep: None?, maxsplit: int?) -> list[str]
+    def split: fn(sep: str, maxsplit: int?) -> list[str]
+    def splitlines: fn(keepends: bool?) -> list[str]
+    def startswith: fn(prefix: str, start: None?, end: None?, /) -> bool
+    def startswith: fn(prefix: str, start: int, end: None?, /) -> bool
+    def startswith: fn(prefix: str, start: None, end: int, /) -> bool
+    def startswith: fn(prefix: str, start: int, end: int, /) -> bool
+    def strip: fn(chars: None?, /) -> str
+    def strip: fn(chars: str, /) -> str
+    def swapcase: fn() -> str
+    def title: fn() -> str
+    // def translate: fn(table: _TranslateTable, /) -> str
+    def upper: fn() -> str
+    def zfill: fn(width: int, /) -> str
+    def __add__: fn(value: str, /) -> str
+    // Incompatible with Sequence.__contains__
+    def __contains__: fn(key: str, /) -> bool
+    def __eq__: fn(value: object, /) -> bool
+    def __ge__: fn(value: str, /) -> bool
+    def __getitem__: fn(key: slice, /) -> str
+    def __getitem__: fn(key: int, /) -> str
+    def __gt__: fn(value: str, /) -> bool
+    def __hash__: fn() -> int
+    // def __iter__: fn() -> Iterator[str]
+    def __le__: fn(value: str, /) -> bool
+    def __len__: fn() -> int
+    def __lt__: fn(value: str, /) -> bool
+    def __mod__: fn(value: Any, /) -> str
+    def __mul__: fn(value: int, /) -> str
+    def __ne__: fn(value: object, /) -> bool
+    def __rmul__: fn(value: int, /) -> str
+    def __getnewargs__: fn() -> tuple[str]
+    def __format__: fn(format_spec: str, /) -> str
+}

midas/checker/builtins.py

@@ -15,7 +15,7 @@ if TYPE_CHECKING:
 
 
 BUILTIN_SUBTYPES: dict[str, set[str]] = {
-    "object": {"float", "list", "dict"},
+    "object": {"float", "list", "dict", "str", "bytes", "tuple"},
     "float": {"int"},
     "int": {"bool"},
 }
@@ -26,12 +26,15 @@ def define_builtins(reg: TypesRegistry):
     any = reg.define_type("Any", TopType())
     unit = reg.define_type("None", UnitType())
     object = reg.define_type("object", BaseType(name="object"))
+    bytes = reg.define_type("bytes", BaseType(name="bytes"))
     bool = reg.define_type("bool", BaseType(name="bool"))
     int = reg.define_type("int", BaseType(name="int"))
     float = reg.define_type("float", BaseType(name="float"))
     str = reg.define_type("str", BaseType(name="str"))
     slice = reg.define_type("slice", BaseType(name="slice"))
 
+    tuple = reg.define_type("tuple", BaseType(name="tuple"))
+
     list = reg.define_type(
         "list",
         GenericType(

midas/checker/diagnostic.py

@@ -9,6 +9,7 @@ class DiagnosticType(StrEnum):
     ERROR = "Error"
     WARNING = "Warning"
     INFO = "Info"
+    DEBUG = "Debug"
 
 
 @dataclass(frozen=True)

midas/checker/dispatcher.py

@@ -0,0 +1,484 @@
+import logging
+from dataclasses import dataclass
+from enum import StrEnum
+from typing import Generic, Optional, Protocol, TypeVar, Union
+
+from midas.ast.location import Location
+from midas.checker.registry import TypesRegistry
+from midas.checker.reporter import FileReporter
+from midas.checker.types import (
+    AppliedType,
+    DerivedType,
+    Function,
+    GenericType,
+    OverloadedFunction,
+    Type,
+    UnknownType,
+)
+from midas.checker.unifier import Unifier
+
+
+class HasLocation(Protocol):
+    @property
+    def location(self) -> Location: ...
+
+
+E = TypeVar("E", bound=HasLocation)
+
+TypedExpr = tuple[E, Type]
+
+
+@dataclass(frozen=True, kw_only=True)
+class MappedArgument(Generic[E]):
+    expr: E
+    type: Type
+    argument: Function.Argument
+
+
+@dataclass(frozen=True, kw_only=True)
+class OverloadCandidate:
+    function: Function
+    mapped: list[MappedArgument]
+
+
+class CallError(StrEnum):
+    INVALID_ARGS = "Invalid arguments"
+    NO_MATCHING_OVERLOAD = "No matching overload"
+    IMPOSSIBLE_UNIFICATION = "Parameters unification failed"
+    NOT_CALLABLE = "Not callable"
+
+
+@dataclass(frozen=True, kw_only=True)
+class CallResult:
+    error: Optional[CallError] = None
+    result: Type = UnknownType()
+    message: Optional[str] = None
+
+    @property
+    def is_valid(self) -> bool:
+        return self.error is None
+
+    @property
+    def error_message(self) -> str:
+        if self.message is not None:
+            return self.message
+        if self.error is not None:
+            return str(self.error)
+        return ""
+
+
+class CallDispatcher(Generic[E]):
+    def __init__(self, types: TypesRegistry, reporter: FileReporter) -> None:
+        self.types: TypesRegistry = types
+        self.reporter: FileReporter = reporter
+        self.logger: logging.Logger = logging.getLogger("CallDispatcher")
+
+    def set_reporter(self, reporter: FileReporter):
+        self.reporter = reporter
+
+    def get_result(
+        self,
+        location: Location,
+        callee: Type,
+        positional: list[TypedExpr[E]],
+        keywords: dict[str, TypedExpr[E]],
+        report_errors: bool = True,
+    ) -> CallResult:
+        """Get the result type of a function call
+
+        If the function has overloads, the function will try to resolve the
+        appropriate signature.
+        Argument types are matched to the defined parameters.
+        The function doesn't take the raw expression as a parameter to accommodate
+        for desugared calls such as for operators.
+
+        Args:
+            location (Location): the call location
+            callee (Type): the called function
+            positional (list[TypedExpr]): the list positional arguments
+            keywords (dict[str, TypedExpr]): the map of keyword arguments
+            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
+
+        Returns:
+            Type: the return type of the call, or `None` if either
+            the call is invalid or no overload matched the arguments uniquely
+        """
+        match callee:
+            case Function() as function:
+                valid: bool
+                mapped: list[MappedArgument[E]]
+                valid, mapped = self.map_call_arguments(
+                    function, location, positional, keywords
+                )
+                valid = valid and self._are_arguments_valid(mapped, report_errors)
+                if not valid:
+                    return CallResult(error=CallError.INVALID_ARGS)
+                return CallResult(result=function.returns)
+
+            case OverloadedFunction(overloads=overloads):
+                res = self._match_overload(
+                    overloads, location, positional, keywords, report_errors
+                )
+                if res[0] is None:
+                    return CallResult(
+                        error=CallError.NO_MATCHING_OVERLOAD,
+                        message=res[1],
+                    )
+                return CallResult(result=res[0].returns)
+
+            case AppliedType(body=body):
+                return self.get_result(
+                    location, body, positional, keywords, report_errors
+                )
+
+            case UnknownType():
+                return CallResult(result=UnknownType())
+
+            case DerivedType(type=base):
+                return self.get_result(
+                    location, base, positional, keywords, report_errors
+                )
+
+            case GenericType():
+                unifier: Unifier = Unifier(self.types)
+                pos: list[Type] = [a[1] for a in positional]
+                kw: dict[str, Type] = {k: v[1] for k, v in keywords.items()}
+                unified: Optional[Type] = unifier.unify_call(callee, pos, kw)
+                if unified is None:
+                    pos_str: str = ", ".join(str(t) for t in pos)
+                    kw_str: str = ", ".join(f"{k}: {v}" for k, v in kw.items())
+                    message: str = (
+                        f"Could not unify {callee}={callee.body} with pos=[{pos_str}] and kw={{{kw_str}}}"
+                    )
+                    if report_errors:
+                        self.reporter.error(location, message)
+                    return CallResult(
+                        error=CallError.IMPOSSIBLE_UNIFICATION,
+                        message=message,
+                    )
+                return self.get_result(
+                    location,
+                    unified,
+                    positional,
+                    keywords,
+                    report_errors,
+                )
+
+            case _:
+                message: str = f"{callee} ({callee.__class__.__name__}) is not callable"
+                if report_errors:
+                    self.reporter.error(location, message)
+                return CallResult(
+                    error=CallError.NOT_CALLABLE,
+                    message=message,
+                )
+
+    def _unwrap_function(
+        self,
+        callee: Type,
+        positional: list[TypedExpr[E]],
+        keywords: dict[str, TypedExpr[E]],
+    ) -> Union[tuple[Function, None], tuple[None, CallError]]:
+        match callee:
+            case DerivedType(type=base):
+                return self._unwrap_function(base, positional, keywords)
+
+            case GenericType():
+                unifier: Unifier = Unifier(self.types)
+                unified: Optional[Type] = unifier.unify_call(
+                    callee,
+                    [a[1] for a in positional],
+                    {k: v[1] for k, v in keywords.items()},
+                )
+                if unified is None:
+                    return None, CallError.IMPOSSIBLE_UNIFICATION
+                return self._unwrap_function(unified, positional, keywords)
+
+            case Function():
+                return callee, None
+
+            case AppliedType(body=body):
+                return self._unwrap_function(body, positional, keywords)
+
+            case _:
+                return None, CallError.NOT_CALLABLE
+
+    def _are_arguments_valid(
+        self,
+        arguments: list[MappedArgument[E]],
+        report_errors: bool = True,
+    ) -> bool:
+        """Check whether the passed argument types correspond to their matched parameter definitions
+
+        Args:
+            arguments (list[MappedArgument]): the list of argument/parameter pairs
+            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
+
+        Returns:
+            bool: True if all arguments fit the matching parameter definitions, False otherwise
+        """
+        valid: bool = True
+        for arg in arguments:
+            if not self.types.is_subtype(arg.type, arg.argument.type):
+                if report_errors:
+                    self.reporter.error(
+                        arg.expr.location,
+                        f"Wrong type for argument '{arg.argument.name}', expected {arg.argument.type}, got {arg.type}",
+                    )
+                valid = False
+        return valid
+
+    def _match_overload(
+        self,
+        overloads: list[Type],
+        location: Location,
+        positional: list[TypedExpr[E]],
+        keywords: dict[str, TypedExpr[E]],
+        report_errors: bool = True,
+    ) -> Union[tuple[Function, None], tuple[None, str]]:
+        """Try and resolve the appropriate overload for the given arguments
+
+        Args:
+            overloads (list[Type]): the list of possible overloads
+            location (Location): the call location
+            positional (list[TypedExpr]): the list of positional arguments
+            keywords (dict[str, TypedExpr]): the map of keywords arguments
+            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
+
+        Returns:
+            Optional[Function]: the resolved function signature if it can be
+            determined unambiguously, or `None`.
+        """
+        candidates: list[OverloadCandidate] = []
+        errors: list[CallError] = []
+        for overload in overloads:
+            function, unwrap_error = self._unwrap_function(
+                overload, positional, keywords
+            )
+            if function is None:
+                errors.append(unwrap_error)  # type: ignore
+                continue
+
+            valid, mapped = self.map_call_arguments(
+                function=function,
+                location=location,
+                positional=positional,
+                keywords=keywords,
+                report_errors=False,
+            )
+            if valid and self._are_arguments_valid(mapped, report_errors=False):
+                candidates.append(
+                    OverloadCandidate(
+                        function=function,
+                        mapped=mapped,
+                    )
+                )
+
+        pos_types: str = ", ".join(str(type) for _, type in positional)
+        kw_types: str = ", ".join(
+            f"{name}: {type}" for name, (_, type) in keywords.items()
+        )
+        for_args: str = f"for arguments pos=[{pos_types}] and kw={{{kw_types}}}"
+
+        n_candidates: int = len(candidates)
+
+        # Exactly 1 match -> return it
+        if n_candidates == 1:
+            return candidates[0].function, None
+
+        # No match -> invalid call
+        if n_candidates == 0:
+            overloads_str: str = ", ".join(map(str, overloads))
+            errors_str: str = ", ".join(errors)
+            message: str = (
+                f"No matching overload in [{overloads_str}] {for_args} (errors: {errors_str})"
+            )
+            if report_errors:
+                self.reporter.error(location, message)
+            return None, message
+
+        # Multiple matches -> see if one <: all others (more specific)
+        for i1, c1 in enumerate(candidates):
+            mapped1: list[MappedArgument[E]] = c1.mapped
+            best_match: bool = True
+            for i2, c2 in enumerate(candidates):
+                if i1 == i2:
+                    continue
+                mapped2: list[MappedArgument[E]] = c2.mapped
+                if not self._are_mapped_subtypes(mapped1, mapped2):
+                    best_match = False
+                    break
+                self.logger.debug(f"{c1.function} is a full overload of {c2.function}")
+            if best_match:
+                return c1.function, None
+
+        candidates_str: str = ", ".join(
+            str(candidate.function) for candidate in candidates
+        )
+        message: str = f"Multiple matching overloads {for_args}: {candidates_str}"
+        if report_errors:
+            self.reporter.error(location, message)
+        return None, message
+
+    def map_call_arguments(
+        self,
+        function: Function,
+        location: Location,
+        positional: list[TypedExpr[E]],
+        keywords: dict[str, TypedExpr[E]],
+        report_errors: bool = True,
+    ) -> tuple[bool, list[MappedArgument]]:
+        """Map call arguments to a function's parameters as defined in its signature
+
+        This method maps positional-only, keyword-only and mixed parameter definitions
+        with the arguments passed at the call site
+
+        Any mismatched, missing or unexpected argument is reported as a diagnostic,
+        unless `report_errors` is set to `False`
+
+        Args:
+            function (Function): the function definition
+            location (Location): the call location
+            positional (list[TypedExpr]): the list of positional arguments
+            keywords (dict[str, TypedExpr]): the map of keyword arguments
+            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
+
+        Returns:
+            tuple[bool, list[MappedArgument]]: a boolean reporting whether
+            the call is valid and the list of mapped arguments
+        """
+        set_args: set[str] = set()
+
+        required_positional: list[str] = [
+            arg.name for arg in function.pos_args + function.args if arg.required
+        ]
+        required_keyword: list[str] = [
+            arg.name for arg in function.kw_args if arg.required
+        ]
+
+        mapped: list[MappedArgument[E]] = []
+
+        pos_params: list[Function.Argument] = list(function.pos_args)
+        mixed_params: list[Function.Argument] = list(function.args)
+        kw_params: dict[str, Function.Argument] = {
+            arg.name: arg for arg in function.kw_args
+        }
+
+        valid_call: bool = True
+
+        # TODO: handle *args and **kwargs sinks
+        for arg in positional:
+            param: Function.Argument
+            if len(pos_params) != 0:
+                param = pos_params.pop(0)
+            elif len(mixed_params) != 0:
+                param = mixed_params.pop(0)
+            else:
+                if report_errors:
+                    self.reporter.error(
+                        arg[0].location, "Too many positional arguments"
+                    )
+                valid_call = False
+                break
+            name: str = param.name
+            if name in required_positional:
+                required_positional.remove(name)
+            if name in required_keyword:
+                required_keyword.remove(name)
+            set_args.add(name)
+            mapped.append(
+                MappedArgument(
+                    expr=arg[0],
+                    type=arg[1],
+                    argument=param,
+                )
+            )
+
+        kw_params.update({arg.name: arg for arg in mixed_params})
+        for name, arg in keywords.items():
+            param: Function.Argument
+            if name not in kw_params:
+                if report_errors:
+                    if name in set_args:
+                        self.reporter.error(
+                            arg[0].location, f"Multiple values for argument '{name}'"
+                        )
+                    else:
+                        self.reporter.error(
+                            arg[0].location, f"Unknown keyword argument '{name}'"
+                        )
+                valid_call = False
+                continue
+            param = kw_params.pop(name)
+            if name in required_positional:
+                required_positional.remove(name)
+            if name in required_keyword:
+                required_keyword.remove(name)
+            set_args.add(name)
+            mapped.append(
+                MappedArgument(
+                    expr=arg[0],
+                    type=arg[1],
+                    argument=param,
+                )
+            )
+
+        def join_args(args: list[str]) -> str:
+            args = list(map(lambda a: f"'{a}'", args))
+            if len(args) == 0:
+                return ""
+            if len(args) == 1:
+                return args[0]
+            return ", ".join(args[:-1]) + " and " + args[-1]
+
+        if len(required_positional) != 0:
+            plural: str = "" if len(required_positional) == 1 else "s"
+            args: str = join_args(required_positional)
+            if report_errors:
+                self.reporter.error(
+                    location,
+                    f"Missing required positional argument{plural}: {args}",
+                )
+            valid_call = False
+
+        if len(required_keyword) != 0:
+            plural: str = "" if len(required_keyword) == 1 else "s"
+            args: str = join_args(required_keyword)
+            if report_errors:
+                self.reporter.error(
+                    location,
+                    f"Missing required keyword argument{plural}: {args}",
+                )
+            valid_call = False
+
+        return valid_call, mapped
+
+    def _are_mapped_subtypes(
+        self, mapped1: list[MappedArgument[E]], mapped2: list[MappedArgument[E]]
+    ) -> bool:
+        """Check whether the given argument mappings are subtype/supertype of one another
+
+        This function checks whether the argument mappings `mapped1` are subtypes
+        of `mapped2`. If any of the parameter type in `mapped1` is not a subtype
+        of the corresponding parameter in `mapped2`, `False` is returned.
+
+        This is used to check whether a given overload is
+        a more specific function/ a subtype of another.
+
+        Args:
+            mapped1 (list[MappedArgument]): the first argument mappings (subtype)
+            mapped2 (list[MappedArgument]): the second argument mappings (supertype)
+
+        Returns:
+            bool: `True` if `mapped1` is a subtype of `mapped2`, `False` otherwise
+        """
+        by_expr: dict[E, Type] = {}
+        for arg in mapped1:
+            by_expr[arg.expr] = arg.argument.type
+
+        for arg in mapped2:
+            type2: Type = arg.argument.type
+            type1: Type = by_expr[arg.expr]
+            if not self.types.is_subtype(type1, type2):
+                return False
+        return True

midas/checker/evaluator.py

@@ -0,0 +1,172 @@
+from dataclasses import dataclass
+from typing import Any, Callable, Optional
+
+import midas.ast.midas as m
+from midas.checker.preamble import Preamble
+from midas.checker.registry import TypesRegistry
+from midas.checker.reporter import FileReporter
+from midas.checker.types import Function, Predicate
+from midas.lexer.token import TokenType
+
+
+@dataclass(frozen=True, kw_only=True)
+class PartialPredicate(Predicate):
+    scope: dict[str, Any]
+
+
+class Evaluator(m.Expr.Visitor[Any]):
+    def __init__(self, types: TypesRegistry, reporter: Optional[FileReporter] = None):
+        self.types: TypesRegistry = types
+        self.reporter: Optional[FileReporter] = reporter
+        self.preamble: Preamble = Preamble(self.types)
+        self.scopes: list[dict[str, Any]] = [{}]
+
+    def evaluate(self, expr: m.Expr) -> Any:
+        value: Any = expr.accept(self)
+        if self.reporter is not None:
+            self.reporter.debug(expr.location, f"Value: {value}")
+        return value
+
+    def get_value(self, name: str) -> Any:
+        scope: dict[str, Any] = self.scopes[-1]
+        return scope[name]
+
+    def set_value(self, name: str, value: Any, force_declare: bool = False):
+        if not force_declare:
+            for scope in reversed(self.scopes):
+                if name in scope:
+                    scope[name] = value
+                    return
+        self.scopes[-1][name] = value
+
+    def visit_logical_expr(self, expr: m.LogicalExpr) -> Any:
+        def left():
+            return self.evaluate(expr.left)
+
+        def right():
+            return self.evaluate(expr.right)
+
+        match expr.operator.type:
+            case TokenType.AND:
+                return left() and right()
+            case _:
+                raise NotImplementedError
+
+    def visit_binary_expr(self, expr: m.BinaryExpr) -> Any:
+        left: Any = self.evaluate(expr.left)
+        right: Any = self.evaluate(expr.right)
+        match expr.operator.type:
+            case TokenType.MINUS:
+                return left - right
+            case TokenType.STAR:
+                return left * right
+            case TokenType.SLASH:
+                return left / right
+            case TokenType.GREATER:
+                return left > right
+            case TokenType.GREATER_EQUAL:
+                return left >= right
+            case TokenType.LESS:
+                return left < right
+            case TokenType.LESS_EQUAL:
+                return left <= right
+            case TokenType.EQUAL_EQUAL:
+                return left == right
+            case TokenType.BANG_EQUAL:
+                return left != right
+            case _:
+                raise NotImplementedError
+
+    def visit_unary_expr(self, expr: m.UnaryExpr) -> Any:
+        right: Any = self.evaluate(expr.right)
+        match expr.operator.type:
+            case TokenType.MINUS:
+                return -right
+            case _:
+                raise NotImplementedError
+
+    def visit_call_expr(self, expr: m.CallExpr) -> Any:
+        callee: Any = self.evaluate(expr.callee)
+        args: list[Any] = [self.evaluate(arg) for arg in expr.arguments]
+        kwargs: dict[str, Any] = {
+            name: self.evaluate(arg) for name, arg in expr.keywords.items()
+        }
+
+        match callee:
+            case Predicate():
+                return self._evaluate_predicate(callee, args, kwargs)
+            case _ if callable(callee):
+                return callee(*args, **kwargs)
+            case _:
+                return NotImplementedError
+
+    def visit_get_expr(self, expr: m.GetExpr) -> Any:
+        obj: Any = self.evaluate(expr.expr)
+        return getattr(obj, expr.name.lexeme)
+
+    def visit_variable_expr(self, expr: m.VariableExpr) -> Any:
+        name: str = expr.name.lexeme
+        for scope in reversed(self.scopes):
+            if name in scope:
+                return scope[name]
+
+        predicate: Optional[Predicate] = self.types.lookup_predicate(name)
+        if predicate is not None:
+            if predicate.alias:
+                return self.evaluate(predicate.body)
+            return predicate
+
+        glob: Optional[Callable] = self.preamble.get_py_func(name)
+        if glob is not None:
+            return glob
+        raise NameError(f"Unknown variable '{name}'")
+
+    def visit_grouping_expr(self, expr: m.GroupingExpr) -> Any:
+        return self.evaluate(expr.expr)
+
+    def visit_literal_expr(self, expr: m.LiteralExpr) -> Any:
+        return expr.value
+
+    def visit_wildcard_expr(self, expr: m.WildcardExpr) -> Any:
+        return self.get_value("_")
+
+    def _evaluate_predicate(
+        self, predicate: Predicate, args: list[Any], kwargs: dict[str, Any]
+    ) -> Any:
+        res: Any = None
+        if isinstance(predicate, PartialPredicate):
+            self.scopes.append(predicate.scope)
+        else:
+            self.scopes.append({})
+        match predicate.type:
+            case Function(returns=Function() as inner):
+                self._map_args(predicate.type, args, kwargs)
+                res = PartialPredicate(
+                    type=inner,
+                    body=predicate.body,
+                    alias=False,
+                    scope=self.scopes[-1],
+                )
+
+            case Function():
+                self._map_args(predicate.type, args, kwargs)
+                res = self.evaluate(predicate.body)
+
+            case _:
+                raise NotImplementedError
+        self.scopes.pop()
+        return res
+
+    def _map_args(self, function: Function, args: list[Any], kwargs: dict[str, Any]):
+        positional: list[Function.Argument] = function.pos_args + function.args
+        keywords: dict[str, Function.Argument] = {
+            arg.name: arg for arg in function.args + function.kw_args
+        }
+
+        for i, arg in enumerate(args):
+            param: Function.Argument = positional[i]
+            self.set_value(param.name, arg)
+
+        for name, arg in kwargs.items():
+            param: Function.Argument = keywords[name]
+            self.set_value(param.name, arg)

midas/checker/frame_methods.py

@@ -0,0 +1,201 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, Any, Callable, Optional
+
+import midas.ast.python as p
+from midas.ast.location import Location
+from midas.checker.dispatcher import CallDispatcher, CallResult
+from midas.checker.registry import TypesRegistry
+from midas.checker.reporter import FileReporter
+from midas.checker.types import (
+    ColumnType,
+    DataFrameType,
+    Function,
+    OverloadedFunction,
+    TopType,
+    Type,
+    UnknownType,
+    unfold_type,
+)
+
+if TYPE_CHECKING:
+    from midas.checker.python import PythonTyper, TypedExpr
+
+
+@staticmethod
+def frame_method(*names: str):
+    def wrapper(func):
+        names_: tuple[str, ...] = names
+        if len(names_) == 0:
+            names_ = (func.__name__,)
+        setattr(func, "__method_names__", names_)
+        return func
+
+    return wrapper
+
+
+@dataclass(frozen=True, kw_only=True)
+class Call:
+    location: Location
+    frame: DataFrameType
+    positional: list[TypedExpr]
+    keywords: dict[str, TypedExpr]
+
+
+class _MethodRegistryMeta(type):
+    _methods: dict[str, Callable[..., Type]] = {}
+
+    def __new__(
+        cls,
+        name: str,
+        bases: tuple[type, ...],
+        namespace: dict[str, Any],
+    ):
+        new_class = super().__new__(cls, name, bases, namespace)
+        new_class._methods = {}
+        for attr in namespace.values():
+            if callable(attr) and hasattr(attr, "__method_names__"):
+                for name in attr.__method_names__:  # type: ignore
+                    new_class._methods[name] = attr  # type: ignore
+        return new_class
+
+
+class MethodRegistry(metaclass=_MethodRegistryMeta):
+    def __init__(self, typer: PythonTyper) -> None:
+        self.typer: PythonTyper = typer
+
+    @property
+    def reporter(self) -> FileReporter:
+        return self.typer.reporter
+
+    @property
+    def types(self) -> TypesRegistry:
+        return self.typer.types
+
+    @property
+    def dispatcher(self) -> CallDispatcher[p.Expr]:
+        return self.typer.dispatcher
+
+    def call(
+        self,
+        method: str,
+        call: Call,
+    ) -> Type:
+        func: Optional[Callable[..., Type]] = self._methods.get(method)
+        if func is None:
+            self.reporter.warning(call.location, f"Unknown method {method}")
+            return UnknownType()
+        return func(self, call)
+
+    @frame_method("add", "__add__")
+    def add(
+        self,
+        call: Call,
+    ) -> Type:
+        # TODO: support add with scalar, sequence, Series, dict
+        # TODO: check operation exists on inner column types
+
+        new_columns: list[DataFrameType.Column] = []
+
+        by_name: dict[str, DataFrameType.Column] = {}
+        frame2: Optional[DataFrameType] = None
+        if len(call.positional) != 0:
+            other: Type = call.positional[0][1]
+            unfolded_other: Type = unfold_type(other)
+            if isinstance(unfolded_other, DataFrameType):
+                frame2 = unfolded_other
+                by_name = {
+                    col.name: col for col in frame2.columns if col.name is not None
+                }
+
+        in_frame1: set[str] = set()
+        for column in call.frame.columns:
+            if column.name is not None:
+                in_frame1.add(column.name)
+
+            col_type1: Type = column.type
+            col_type: Type = ColumnType(type=UnknownType())
+            if column.name in by_name:
+                column2 = by_name[column.name]
+                col_type2: Type = column2.type
+                if self.types.are_equivalent(col_type2, col_type1):
+                    col_type = col_type1
+
+            new_column = DataFrameType.Column(
+                index=column.index,
+                name=column.name,
+                type=col_type,
+            )
+            new_columns.append(new_column)
+
+        if frame2 is not None:
+            for column in frame2.columns:
+                if column.name in in_frame1:
+                    continue
+                new_columns.append(
+                    DataFrameType.Column(
+                        index=len(new_columns),
+                        name=column.name,
+                        type=ColumnType(type=UnknownType()),
+                    )
+                )
+
+        signature = Function(
+            args=[
+                Function.Argument(
+                    pos=0,
+                    name="other",
+                    type=DataFrameType(columns=[]),
+                    required=True,
+                ),
+            ],
+            returns=DataFrameType(columns=new_columns),
+        )
+
+        result: CallResult = self.dispatcher.get_result(
+            location=call.location,
+            callee=signature,
+            positional=call.positional,
+            keywords=call.keywords,
+        )
+        return result.result
+
+    @frame_method()
+    def mean(self, call: Call) -> Type:
+        with_axis = Function(
+            kw_args=[
+                Function.Argument(
+                    pos=0,
+                    name="axis",
+                    type=self.types.get_type("int"),
+                    required=False,
+                )
+            ],
+            returns=ColumnType(type=TopType()),
+        )
+        without_axis = Function(
+            kw_args=[
+                Function.Argument(
+                    pos=0,
+                    name="axis",
+                    type=self.types.get_type("None"),
+                    required=True,
+                )
+            ],
+            returns=TopType(),
+        )
+        overload = OverloadedFunction(
+            overloads=[
+                with_axis,
+                without_axis,
+            ]
+        )
+
+        result: CallResult = self.dispatcher.get_result(
+            location=call.location,
+            callee=overload,
+            positional=call.positional,
+            keywords=call.keywords,
+        )
+        return result.result

midas/checker/frames.py

@@ -0,0 +1,154 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Optional, TypeGuard, cast
+
+import midas.ast.python as p
+from midas.ast.location import Location
+from midas.checker.frame_methods import Call, MethodRegistry
+from midas.checker.reporter import FileReporter
+from midas.checker.types import ColumnType, DataFrameType, TupleType, Type, UnknownType
+
+if TYPE_CHECKING:
+    from midas.checker.python import PythonTyper, TypedExpr
+
+
+def is_list_of_literals(exprs: list[p.Expr]) -> TypeGuard[list[p.LiteralExpr]]:
+    return all(isinstance(expr, p.LiteralExpr) for expr in exprs)
+
+
+class FrameManager:
+    def __init__(self, typer: PythonTyper) -> None:
+        self.typer: PythonTyper = typer
+        self.method_resolver: MethodRegistry = MethodRegistry(self.typer)
+
+    def assign(
+        self,
+        reporter: FileReporter,
+        location: Location,
+        frame: DataFrameType,
+        index: p.Expr,
+        value_type: Type,
+    ) -> Type:
+        match index:
+            case p.LiteralExpr(value=str() as name):
+                return self.assign_column(reporter, location, frame, name, value_type)
+
+            case p.ListExpr(items=indices) if is_list_of_literals(indices) and all(
+                isinstance(idx, str) for idx in indices
+            ):
+                raise NotImplementedError
+
+            case _:
+                reporter.error(location, f"Invalid index type {index} on {frame}")
+                return UnknownType()
+
+    def assign_column(
+        self,
+        reporter: FileReporter,
+        location: Location,
+        frame: DataFrameType,
+        name: str,
+        type: Type,
+    ) -> Type:
+        if not isinstance(type, ColumnType):
+            reporter.error(
+                location,
+                f"Cannot assign {type} to dataframe column. Must be a ColumnType",
+            )
+            return frame
+        return self._set_column(frame, name, type)
+
+    def get(
+        self,
+        reporter: FileReporter,
+        location: Location,
+        frame: DataFrameType,
+        index: p.Expr,
+    ) -> Type:
+        match index:
+            case p.LiteralExpr(value=str() as name):
+                column: Optional[ColumnType] = FrameManager._get_column(frame, name)
+                if column is None:
+                    reporter.error(location, f"Unknown column '{name}' on {frame}")
+                    return UnknownType()
+                return column
+
+            case p.ListExpr(items=indices) if is_list_of_literals(indices) and all(
+                isinstance(index.value, str) for index in indices
+            ):
+                names: list[str] = [cast(str, index.value) for index in indices]
+                columns: list[ColumnType] = []
+                for name in names:
+                    column: Optional[ColumnType] = FrameManager._get_column(frame, name)
+                    if column is None:
+                        reporter.error(location, f"Unknown column '{name}' on {frame}")
+                        return UnknownType()
+                    columns.append(column)
+                return TupleType(items=tuple(columns))
+
+            case _:
+                reporter.error(location, f"Invalid index type {index} on {frame}")
+                return UnknownType()
+
+    @classmethod
+    def _set_column(
+        cls, frame: DataFrameType, name: str, column: ColumnType
+    ) -> DataFrameType:
+        new_columns: list[DataFrameType.Column] = []
+        index: int = len(frame.columns)
+        replace: bool = False
+        for i, col in enumerate(frame.columns):
+            if col.name == name:
+                index = i
+                replace = True
+                # TODO: check column type here to prevent changing it
+            new_columns.append(col)
+
+        new_col: DataFrameType.Column = DataFrameType.Column(
+            index=index,
+            name=name,
+            type=column,
+        )
+        if replace:
+            new_columns[index] = new_col
+        else:
+            new_columns.append(new_col)
+
+        return DataFrameType(columns=new_columns)
+
+    @classmethod
+    def _set_columns(
+        cls, frame: DataFrameType, names: list[str], columns: list[ColumnType]
+    ) -> DataFrameType:
+        for name, col in zip(names, columns):
+            frame = cls._set_column(frame, name, col)
+        return frame
+
+    @classmethod
+    def _get_column(cls, frame: DataFrameType, name: str) -> Optional[ColumnType]:
+        for col in frame.columns:
+            if col.name == name:
+                return col.type
+        return None
+
+    @classmethod
+    def _get_columns(
+        cls, frame: DataFrameType, names: list[str]
+    ) -> list[Optional[ColumnType]]:
+        return [cls._get_column(frame, name) for name in names]
+
+    def call(
+        self,
+        method: str,
+        location: Location,
+        frame: DataFrameType,
+        positional: list[TypedExpr],
+        keywords: dict[str, TypedExpr],
+    ) -> Type:
+        call: Call = Call(
+            location=location,
+            frame=frame,
+            positional=positional,
+            keywords=keywords,
+        )
+        return self.method_resolver.call(method, call)

midas/checker/midas.py

@@ -6,25 +6,25 @@ from typing import Optional
 import midas.ast.midas as m
 from midas.ast.location import Location
 from midas.checker.builtins import define_builtins
+from midas.checker.dispatcher import CallDispatcher, CallResult
 from midas.checker.environment import Environment
 from midas.checker.operators import MIDAS_BINARY_METHODS, MIDAS_UNARY_METHODS
 from midas.checker.preamble import Preamble
 from midas.checker.registry import TypesRegistry
 from midas.checker.reporter import FileReporter, Reporter
 from midas.checker.types import (
-    AliasType,
+    ColumnType,
-    AppliedType,
     ComplexType,
     ConstraintType,
+    DataFrameType,
+    DerivedType,
     ExtensionType,
     Function,
     GenericType,
-    OverloadedFunction,
     Predicate,
     Type,
     TypeVar,
     UnknownType,
-    unfold_type,
 )
 from midas.checker.variance import VarianceInferrer
 from midas.lexer.midas import MidasLexer
@@ -39,9 +39,6 @@ class TypedParamSpec:
     kw: list[Function.Argument]
 
 
-TypedExpr = tuple[m.Expr, Type]
-
-
 class ReturnException(Exception):
     pass
 
@@ -65,8 +62,11 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
     def __init__(self, types: TypesRegistry, reporter: Reporter) -> None:
         self.logger: logging.Logger = logging.getLogger("MidasTyper")
         self.reporter: FileReporter = reporter.for_file(None)
-
         self.types: TypesRegistry = types
+        self.dispatcher: CallDispatcher[m.Expr] = CallDispatcher[m.Expr](
+            self.types, self.reporter
+        )
+
         self._local_variables: dict[str, TypeVar] = {}
 
         self._predicate_params: dict[str, Type] = {}
@@ -81,8 +81,14 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
 
         self._preamble: Environment = Preamble(self.types)
 
+    def set_reporter(self, reporter: FileReporter):
+        self.reporter = reporter
+        self.dispatcher.set_reporter(reporter)
+
     def process(self, source: str, path: Optional[str]):
-        self.reporter = self.reporter.for_file(path)
+        reporter: FileReporter = self.reporter.for_file(path)
+        self.set_reporter(reporter)
+
         lexer: MidasLexer = MidasLexer(source)
         tokens: list[Token] = lexer.process()
         parser: MidasParser = MidasParser(tokens)
@@ -152,11 +158,18 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
         if len(params) != 0:
             type = GenericType(name=name, params=params, body=type)
         else:
-            type = AliasType(name=name, type=type)
+            type = DerivedType(name=name, type=type)
         self.types.define_type(name, type)
         self._local_variables.clear()
         self._current_name = None
 
+    def visit_alias_stmt(self, stmt: m.AliasStmt) -> None:
+        name: str = stmt.name.lexeme
+        self._current_name = name
+        type: Type = stmt.type.accept(self)
+        self.types.define_type(name, type)
+        self._current_name = None
+
     def visit_member_stmt(self, stmt: m.MemberStmt) -> None: ...
 
     def visit_extend_stmt(self, stmt: m.ExtendStmt) -> None:
@@ -173,7 +186,7 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
                 base_name,
                 member.name.lexeme,
                 member_type,
-                member.kind == m.MemberKind.METHOD,
+                member.kind,
             )
 
     def visit_predicate_stmt(self, stmt: m.PredicateStmt) -> None:
@@ -250,13 +263,13 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
             )
             return UnknownType()
 
-        result: Optional[Type] = self._get_call_result(
+        result: CallResult = self.dispatcher.get_result(
-            location,
+            location=location,
-            operation,
+            callee=operation,
-            [(right_expr, right)],
+            positional=[(right_expr, right)],
-            {},
+            keywords={},
         )
-        return result or UnknownType()
+        return result.result
 
     def visit_unary_expr(self, expr: m.UnaryExpr) -> Type:
         method: Optional[str] = MIDAS_UNARY_METHODS.get(expr.operator.type)
@@ -276,31 +289,29 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
             )
             return UnknownType()
 
-        result: Optional[Type] = self._get_call_result(
+        result: CallResult = self.dispatcher.get_result(
-            expr.location,
+            location=expr.location,
-            operation,
+            callee=operation,
-            [],
+            positional=[],
-            {},
+            keywords={},
         )
-        return result or UnknownType()
+        return result.result
 
     def visit_call_expr(self, expr: m.CallExpr) -> Type:
         callee: Type = expr.callee.accept(self)
-        positional: list[TypedExpr] = [
+        positional: list[tuple[m.Expr, Type]] = [
             (arg, self.type_of(arg)) for arg in expr.arguments
         ]
-        keywords: dict[str, TypedExpr] = {
+        keywords: dict[str, tuple[m.Expr, Type]] = {
             name: (arg, self.type_of(arg)) for name, arg in expr.keywords.items()
         }
-        return (
+        result: CallResult = self.dispatcher.get_result(
-            self._get_call_result(
+            location=expr.location,
-                expr.location,
+            callee=callee,
-                callee,
+            positional=positional,
-                positional,
+            keywords=keywords,
-                keywords,
-            )
-            or UnknownType()
         )
+        return result.result
 
     def visit_get_expr(self, expr: m.GetExpr) -> Type:
         object: Type = expr.expr.accept(self)
@@ -401,6 +412,18 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
             kw=[process_arg(arg, i + n_pos + n_mixed) for i, arg in enumerate(spec.kw)],
         )
 
+    def visit_frame_type(self, type: m.FrameType) -> Type:
+        def process_column(i: int, col: m.FrameType.Column) -> DataFrameType.Column:
+            return DataFrameType.Column(
+                index=i,
+                name=col.name.lexeme,
+                type=ColumnType(type=col.type.accept(self)),
+            )
+
+        return DataFrameType(
+            columns=[process_column(i, col) for i, col in enumerate(type.columns)]
+        )
+
     def _resolve_type_params(self, params: list[m.TypeParam]):
         vars: list[TypeVar] = []
         for param in params:
@@ -412,343 +435,3 @@ class MidasTyper(m.Stmt.Visitor[None], m.Expr.Visitor[Type], m.Type.Visitor[Type
             self._local_variables[name] = var
             vars.append(var)
         return vars
-
-    def _get_call_result(
-        self,
-        location: Location,
-        callee: Type,
-        positional: list[TypedExpr],
-        keywords: dict[str, TypedExpr],
-        report_errors: bool = True,
-    ) -> Optional[Type]:
-        """Get the result type of a function call
-
-        If the function has overloads, the function will try to resolve the
-        appropriate signature.
-        Argument types are matched to the defined parameters.
-        The function doesn't take the raw expression as a parameter to accommodate
-        for desugared calls such as for operators.
-
-        Args:
-            location (Location): the call location
-            callee (Type): the called function
-            positional (list[TypedExpr]): the list positional arguments
-            keywords (dict[str, TypedExpr]): the map of keyword arguments
-            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
-
-        Returns:
-            Type: the return type of the call, or `None` if either
-            the call is invalid or no overload matched the arguments uniquely
-        """
-        match callee:
-            case Function() as function:
-                valid: bool
-                mapped: list[MappedArgument]
-                valid, mapped = self.map_call_arguments(
-                    function, location, positional, keywords
-                )
-                valid = valid and self._are_arguments_valid(mapped, report_errors)
-                if not valid:
-                    return None
-                return function.returns
-
-            case OverloadedFunction(overloads=overloads):
-                function = self._match_overload(
-                    overloads, location, positional, keywords, report_errors
-                )
-                if function is None:
-                    return None
-                return function.returns
-
-            case AppliedType(body=body):
-                return self._get_call_result(
-                    location, body, positional, keywords, report_errors
-                )
-
-            case UnknownType():
-                return UnknownType()
-
-            case _:
-                if report_errors:
-                    self.reporter.error(location, f"{callee} is not callable")
-                return None
-
-    def _are_arguments_valid(
-        self,
-        arguments: list[MappedArgument],
-        report_errors: bool = True,
-    ) -> bool:
-        """Check whether the passed argument types correspond to their matched parameter definitions
-
-        Args:
-            arguments (list[MappedArgument]): the list of argument/parameter pairs
-            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
-
-        Returns:
-            bool: True if all arguments fit the matching parameter definitions, False otherwise
-        """
-        valid: bool = True
-        for arg in arguments:
-            if not self.types.is_subtype(arg.type, arg.argument.type):
-                if report_errors:
-                    self.reporter.error(
-                        arg.expr.location,
-                        f"Wrong type for argument '{arg.argument.name}', expected {arg.argument.type}, got {arg.type}",
-                    )
-                valid = False
-        return valid
-
-    def _match_overload(
-        self,
-        overloads: list[Type],
-        location: Location,
-        positional: list[TypedExpr],
-        keywords: dict[str, TypedExpr],
-        report_errors: bool = True,
-    ) -> Optional[Function]:
-        """Try and resolve the appropriate overload for the given arguments
-
-        Args:
-            overloads (list[Type]): the list of possible overloads
-            location (Location): the call location
-            positional (list[TypedExpr]): the list of positional arguments
-            keywords (dict[str, TypedExpr]): the map of keywords arguments
-            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
-
-        Returns:
-            Optional[Function]: the resolved function signature if it can be
-            determined unambiguously, or `None`.
-        """
-        candidates: list[OverloadCandidate] = []
-        for overload in overloads:
-            function: Type = unfold_type(overload)
-            if not isinstance(function, Function):
-                if report_errors:
-                    self.logger.error(
-                        f"Overload is not a function: {overload} is {function}"
-                    )
-                continue
-            valid, mapped = self.map_call_arguments(
-                function=function,
-                location=location,
-                positional=positional,
-                keywords=keywords,
-                report_errors=False,
-            )
-            if valid and self._are_arguments_valid(mapped, report_errors=False):
-                candidates.append(
-                    OverloadCandidate(
-                        function=function,
-                        mapped=mapped,
-                    )
-                )
-
-        pos_types: str = ", ".join(str(type) for _, type in positional)
-        kw_types: str = ", ".join(
-            f"{name}: {type}" for name, (_, type) in keywords.items()
-        )
-        for_args: str = f"for arguments pos=[{pos_types}] and kw={{{kw_types}}}"
-
-        n_candidates: int = len(candidates)
-
-        # Exactly 1 match -> return it
-        if n_candidates == 1:
-            return candidates[0].function
-
-        # No match -> invalid call
-        if n_candidates == 0:
-            overloads_str: str = ", ".join(map(str, overloads))
-            if report_errors:
-                self.reporter.error(
-                    location,
-                    f"No matching overload in [{overloads_str}] {for_args}",
-                )
-            return None
-
-        # Multiple matches -> see if one <: all others (more specific)
-        for i1, c1 in enumerate(candidates):
-            mapped1: list[MappedArgument] = c1.mapped
-            best_match: bool = True
-            for i2, c2 in enumerate(candidates):
-                if i1 == i2:
-                    continue
-                mapped2: list[MappedArgument] = c2.mapped
-                if not self._are_mapped_subtypes(mapped1, mapped2):
-                    best_match = False
-                    break
-                self.logger.debug(f"{c1.function} is a full overload of {c2.function}")
-            if best_match:
-                return c1.function
-
-        candidates_str: str = ", ".join(
-            str(candidate.function) for candidate in candidates
-        )
-        if report_errors:
-            self.reporter.error(
-                location,
-                f"Multiple matching overloads {for_args}: {candidates_str}",
-            )
-        return None
-
-    def map_call_arguments(
-        self,
-        function: Function,
-        location: Location,
-        positional: list[TypedExpr],
-        keywords: dict[str, TypedExpr],
-        report_errors: bool = True,
-    ) -> tuple[bool, list[MappedArgument]]:
-        """Map call arguments to a function's parameters as defined in its signature
-
-        This method maps positional-only, keyword-only and mixed parameter definitions
-        with the arguments passed at the call site
-
-        Any mismatched, missing or unexpected argument is reported as a diagnostic,
-        unless `report_errors` is set to `False`
-
-        Args:
-            function (Function): the function definition
-            location (Location): the call location
-            positional (list[TypedExpr]): the list of positional arguments
-            keywords (dict[str, TypedExpr]): the map of keyword arguments
-            report_errors (bool, optional): whether type errors should be reported as diagnostics. Defaults to True.
-
-        Returns:
-            tuple[bool, list[MappedArgument]]: a boolean reporting whether
-            the call is valid and the list of mapped arguments
-        """
-        set_args: set[str] = set()
-
-        required_positional: list[str] = [
-            arg.name for arg in function.pos_args + function.args if arg.required
-        ]
-        required_keyword: list[str] = [
-            arg.name for arg in function.kw_args if arg.required
-        ]
-
-        mapped: list[MappedArgument] = []
-
-        pos_params: list[Function.Argument] = list(function.pos_args)
-        mixed_params: list[Function.Argument] = list(function.args)
-        kw_params: dict[str, Function.Argument] = {
-            arg.name: arg for arg in function.kw_args
-        }
-
-        valid_call: bool = True
-
-        # TODO: handle *args and **kwargs sinks
-        for arg in positional:
-            param: Function.Argument
-            if len(pos_params) != 0:
-                param = pos_params.pop(0)
-            elif len(mixed_params) != 0:
-                param = mixed_params.pop(0)
-            else:
-                if report_errors:
-                    self.reporter.error(
-                        arg[0].location, "Too many positional arguments"
-                    )
-                valid_call = False
-                break
-            name: str = param.name
-            if name in required_positional:
-                required_positional.remove(name)
-            if name in required_keyword:
-                required_keyword.remove(name)
-            set_args.add(name)
-            mapped.append(
-                MappedArgument(
-                    expr=arg[0],
-                    type=arg[1],
-                    argument=param,
-                )
-            )
-
-        kw_params.update({arg.name: arg for arg in mixed_params})
-        for name, arg in keywords.items():
-            param: Function.Argument
-            if name not in kw_params:
-                if report_errors:
-                    if name in set_args:
-                        self.reporter.error(
-                            arg[0].location, f"Multiple values for argument '{name}'"
-                        )
-                    else:
-                        self.reporter.error(
-                            arg[0].location, f"Unknown keyword argument '{name}'"
-                        )
-                valid_call = False
-                continue
-            param = kw_params.pop(name)
-            if name in required_positional:
-                required_positional.remove(name)
-            if name in required_keyword:
-                required_keyword.remove(name)
-            set_args.add(name)
-            mapped.append(
-                MappedArgument(
-                    expr=arg[0],
-                    type=arg[1],
-                    argument=param,
-                )
-            )
-
-        def join_args(args: list[str]) -> str:
-            args = list(map(lambda a: f"'{a}'", args))
-            if len(args) == 0:
-                return ""
-            if len(args) == 1:
-                return args[0]
-            return ", ".join(args[:-1]) + " and " + args[-1]
-
-        if len(required_positional) != 0:
-            plural: str = "" if len(required_positional) == 1 else "s"
-            args: str = join_args(required_positional)
-            if report_errors:
-                self.reporter.error(
-                    location,
-                    f"Missing required positional argument{plural}: {args}",
-                )
-            valid_call = False
-
-        if len(required_keyword) != 0:
-            plural: str = "" if len(required_keyword) == 1 else "s"
-            args: str = join_args(required_keyword)
-            if report_errors:
-                self.reporter.error(
-                    location,
-                    f"Missing required keyword argument{plural}: {args}",
-                )
-            valid_call = False
-
-        return valid_call, mapped
-
-    def _are_mapped_subtypes(
-        self, mapped1: list[MappedArgument], mapped2: list[MappedArgument]
-    ) -> bool:
-        """Check whether the given argument mappings are subtype/supertype of one another
-
-        This function checks whether the argument mappings `mapped1` are subtypes
-        of `mapped2`. If any of the parameter type in `mapped1` is not a subtype
-        of the corresponding parameter in `mapped2`, `False` is returned.
-
-        This is used to check whether a given overload is
-        a more specific function/ a subtype of another.
-
-        Args:
-            mapped1 (list[MappedArgument]): the first argument mappings (subtype)
-            mapped2 (list[MappedArgument]): the second argument mappings (supertype)
-
-        Returns:
-            bool: `True` if `mapped1` is a subtype of `mapped2`, `False` otherwise
-        """
-        by_expr: dict[m.Expr, Type] = {}
-        for arg in mapped1:
-            by_expr[arg.expr] = arg.argument.type
-
-        for arg in mapped2:
-            type2: Type = arg.argument.type
-            type1: Type = by_expr[arg.expr]
-            if not self.types.is_subtype(type1, type2):
-                return False
-        return True

midas/checker/operators.py

@@ -41,7 +41,7 @@ PY_UNARY_METHODS: dict[Type[ast.unaryop], str] = {
 
 
 MIDAS_BINARY_METHODS: dict[TokenType, str] = {
-    # TokenType.PLUS: "__add__",
+    TokenType.PLUS: "__add__",
     TokenType.MINUS: "__sub__",
     TokenType.STAR: "__mul__",
     TokenType.SLASH: "__truediv__",

midas/checker/preamble.py

@@ -1,8 +1,17 @@
 from dataclasses import dataclass
+from typing import Any, Callable, Optional
 
 from midas.checker.environment import Environment
 from midas.checker.registry import TypesRegistry
-from midas.checker.types import Function, GenericType, TopType, Type, TypeVar, UnitType
+from midas.checker.types import (
+    Function,
+    GenericType,
+    OverloadedFunction,
+    TopType,
+    Type,
+    TypeVar,
+    UnitType,
+)
 
 
 @dataclass(frozen=True)
@@ -16,23 +25,26 @@ class Preamble(Environment):
     def __init__(self, types: TypesRegistry) -> None:
         super().__init__()
         self._types: TypesRegistry = types
+        self._python_funcs: dict[str, Callable[..., Any]] = {}
 
-        self._def_type_constructor("object")
+        self._def_type_constructor("object", object)
-        self._def_type_constructor("float")
+        self._def_type_constructor("float", float)
-        self._def_type_constructor("int")
+        self._def_type_constructor("int", int)
-        self._def_type_constructor("bool")
+        self._def_type_constructor("bool", bool)
-        self._def_type_constructor("str")
+        self._def_type_constructor("str", str)
         self._def_function(
             name="list",
             pos=[Param("object", TopType())],
             returns=self._list_of(TopType()),
+            py_function=list,
         )
 
         # TODO: use sink
         self._def_function(
             name="print",
-            pos=[Param("object", TopType())],
+            pos=[Param("object", TopType(), required=False)],
             returns=UnitType(),
+            py_function=print,
         )
 
         map_in = TypeVar(name="T", bound=None)
@@ -52,17 +64,62 @@ class Preamble(Environment):
                 ),
             ],
             returns=self._list_of(map_out),  # TODO: replace with Iterable[U]
+            type_vars=[map_in, map_out],
+            py_function=map,
+        )
+        self._def_function(
+            name="input",
+            pos=[Param("prompt", TopType(), required=False)],
+            returns=self._types.get_type("str"),
+        )
+        self._def_function(
+            name="len",
+            pos=[Param("object", TopType())],
+            returns=self._types.get_type("int"),
+        )
+
+        T = TypeVar(name="T", bound=None)
+        self._def_overloads(
+            name="max",
+            py_function=max,
+            signatures=[
+                (
+                    [Param("arg1", T), Param("arg2", T)],
+                    [],
+                    [],
+                    T,
+                    [T],
+                ),
+                ([Param("iterable", self._list_of(T))], [], [], T, [T]),
+            ],
+        )
+        self._def_overloads(
+            name="min",
+            py_function=min,
+            signatures=[
+                (
+                    [Param("arg1", T), Param("arg2", T)],
+                    [],
+                    [],
+                    T,
+                    [T],
+                ),
+                ([Param("iterable", self._list_of(T))], [], [], T, [T]),
+            ],
         )
 
     def _list_of(self, item_type: Type) -> Type:
         return self._types.apply_generic(self._types.get_type("list"), [item_type])
 
-    def _def_type_constructor(self, name: str):
+    def _def_type_constructor(
+        self, name: str, py_function: Optional[Callable[..., Any]] = None
+    ):
         # TODO: more specific arg types
         self._def_function(
             name=name,
             pos=[Param("object", TopType(), required=False)],
             returns=self._types.get_type(name),
+            py_function=py_function,
         )
 
     def _make_function(
@@ -109,6 +166,7 @@ class Preamble(Environment):
         kw: list[Param] = [],
         returns: Type = UnitType(),
         type_vars: list[TypeVar] = [],
+        py_function: Optional[Callable[..., Any]] = None,
     ):
         function: Type = self._make_function(
             name=name,
@@ -119,3 +177,34 @@ class Preamble(Environment):
             type_vars=type_vars,
         )
         self.define(name, function)
+        if py_function is not None:
+            self._python_funcs[name] = py_function
+
+    def _def_overloads(
+        self,
+        *,
+        name: str,
+        signatures: list[
+            tuple[list[Param], list[Param], list[Param], Type, list[TypeVar]]
+        ],
+        py_function: Optional[Callable[..., Any]] = None,
+    ):
+        overloads: list[Type] = []
+        for pos, mixed, kw, returns, type_vars in signatures:
+            overloads.append(
+                self._make_function(
+                    name=name,
+                    pos=pos,
+                    mixed=mixed,
+                    kw=kw,
+                    returns=returns,
+                    type_vars=type_vars,
+                )
+            )
+        function: Type = OverloadedFunction(overloads=overloads)
+        self.define(name, function)
+        if py_function is not None:
+            self._python_funcs[name] = py_function
+
+    def get_py_func(self, name: str) -> Optional[Callable[..., Any]]:
+        return self._python_funcs.get(name)

midas/checker/registry.py

@@ -1,19 +1,24 @@
 import logging
+from dataclasses import dataclass
 from typing import Optional
 
+from midas.ast.midas import MemberKind
 from midas.checker.builtins import BUILTIN_SUBTYPES
 from midas.checker.types import (
-    AliasType,
     AppliedType,
     BaseType,
+    ColumnType,
     ComplexType,
     ConstraintType,
+    DataFrameType,
+    DerivedType,
     ExtensionType,
     Function,
     GenericType,
     OverloadedFunction,
     Predicate,
     TopType,
+    TupleType,
     Type,
     TypeVar,
     UnknownType,
@@ -22,11 +27,17 @@ from midas.checker.types import (
 )
 
 
+@dataclass
+class Member:
+    kind: MemberKind
+    type: Type
+
+
 class TypesRegistry:
     def __init__(self) -> None:
         self.logger: logging.Logger = logging.getLogger("TypesRegistry")
         self._types: dict[str, Type] = {}
-        self._members: dict[str, dict[str, Type]] = {}
+        self._members: dict[str, dict[str, Member]] = {}
         self._predicates: dict[str, Predicate] = {}
 
     def get_type(self, name: str) -> Type:
@@ -64,26 +75,38 @@ class TypesRegistry:
         return type
 
     def define_member(
-        self, type_name: str, member_name: str, member_type: Type, is_method: bool
+        self,
+        type_name: str,
+        member_name: str,
+        member_type: Type,
+        kind: MemberKind,
     ):
-        members: dict[str, Type] = self._members.setdefault(type_name, {})
+        members: dict[str, Member] = self._members.setdefault(type_name, {})
         if member_name in members:
-            if not is_method:
+            current: Member = members[member_name]
+            if current.kind != kind:
                 self.logger.error(
-                    f"Member '{member_name}' already defined for type {type_name}"
+                    f"Member '{member_name}' is already defined as a {current.kind},"
+                    + f" cannot define a {kind} with the same name"
                 )
                 return
-            current: Type = members[member_name]
+            if kind != MemberKind.METHOD:
+                self.logger.error(
+                    f"Member '{member_name}' already defined for type {type_name},"
+                    + " only methods can be overloaded"
+                )
+                return
+
             combined: Type
-            match current:
+            match current.type:
                 case OverloadedFunction(overloads=overloads):
                     combined = OverloadedFunction(overloads=overloads + [member_type])
                 case _:
-                    combined = OverloadedFunction(overloads=[current, member_type])
+                    combined = OverloadedFunction(overloads=[current.type, member_type])
-            members[member_name] = combined
+            members[member_name] = Member(kind=current.kind, type=combined)
 
         else:
-            members[member_name] = member_type
+            members[member_name] = Member(kind=kind, type=member_type)
 
     def define_predicate(self, name: str, predicate: Predicate):
         if name in self._predicates:
@@ -110,7 +133,20 @@ class TypesRegistry:
             case (_, TopType()):
                 return True
 
-            case (AliasType(type=base1), _):
+            case (_, UnknownType()):
+                return True
+
+            case (TypeVar(bound=bound), _):
+                if bound is None:
+                    return False
+                return self.is_subtype(bound, type2)
+
+            case (_, TypeVar(bound=bound)):
+                if bound is None:
+                    return True
+                return self.is_subtype(type1, bound)
+
+            case (DerivedType(type=base1), _):
                 return self.is_subtype(base1, type2)
 
             case (BaseType(name=name1), BaseType(name=name2)):
@@ -124,14 +160,27 @@ class TypesRegistry:
                         return False
                 return True
 
+            case (DataFrameType(columns=columns1), DataFrameType(columns=columns2)):
+                # TODO: check order?
+                by_name1: dict[str, DataFrameType.Column] = {
+                    col.name: col for col in columns1 if col.name is not None
+                }
+                for col2 in columns2:
+                    if col2.name not in by_name1:
+                        return False
+                    if not self.is_subtype(by_name1[col2.name].type, col2.type):
+                        return False
+                return True
+
+            case (ColumnType(type=inner1), ColumnType(type=inner2)):
+                # TODO: invariant, replace ColumnType with simple GenericType
+                if not self.are_equivalent(inner1, inner2):
+                    return False
+                return True
+
             case (Function(), Function()):
                 return self.is_func_subtype(type1, type2)
 
-            case (TypeVar(bound=bound), _):
-                if bound is None:
-                    return False
-                return self.is_subtype(bound, type2)
-
             case (ConstraintType(type=base1), _):
                 return self.is_subtype(base1, type2)
 
@@ -153,8 +202,15 @@ class TypesRegistry:
                             return False
                 return True
 
+            # TODO: verify legitimacy
+            case (AppliedType(body=body), _):
+                return self.is_subtype(body, type2)
+
         return False
 
+    def are_equivalent(self, type1: Type, type2: Type) -> bool:
+        return self.is_subtype(type1, type2) and self.is_subtype(type2, type1)
+
     # TODO: verify the logic in here
     def is_func_subtype(self, func1: Function, func2: Function) -> bool:
         """Check whether a function is a subtype of another
@@ -262,8 +318,8 @@ class TypesRegistry:
 
     def apply_generic(self, type: Type, args: list[Type]) -> Type:
         match type:
-            case AliasType(name=name, type=base):
+            case DerivedType(name=name, type=base):
-                return AliasType(name=name, type=self.apply_generic(base, args))
+                return DerivedType(name=name, type=self.apply_generic(base, args))
 
             case GenericType(name=name, params=type_vars, body=body):
                 n_args: int = len(args)
@@ -291,6 +347,9 @@ class TypesRegistry:
                     body=substitute_typevars(body, substitutions),
                 )
 
+            case BaseType(name="tuple"):
+                return TupleType(items=tuple(args))
+
             case _:
                 raise ValueError(f"{type} is not a generic type")
 
@@ -327,13 +386,13 @@ class TypesRegistry:
             case BaseType(name=name):
                 if name in self._members:
                     if member_name in self._members[name]:
-                        return self._members[name][member_name]
+                        return self._members[name][member_name].type
                 return None
 
-            case AliasType(name=name, type=base):
+            case DerivedType(name=name, type=base):
                 if name in self._members:
                     if member_name in self._members[name]:
-                        return self._members[name][member_name]
+                        return self._members[name][member_name].type
                 return self.lookup_member(base, member_name)
 
             case AppliedType(name=name, body=body, args=args):
@@ -347,7 +406,7 @@ class TypesRegistry:
                 }
                 if name in self._members:
                     if member_name in self._members[name]:
-                        member_type: Type = self._members[name][member_name]
+                        member_type: Type = self._members[name][member_name].type
                         return substitute_typevars(member_type, substitutions)
 
                 member_type2: Optional[Type] = self.lookup_member(body, member_name)
@@ -369,6 +428,12 @@ class TypesRegistry:
                 )
                 return self.lookup_member(base, member_name)
 
+            case ConstraintType(type=base):
+                return self.lookup_member(base, member_name)
+
+            case TypeVar(bound=bound) if bound is not None:
+                return self.lookup_member(bound, member_name)
+
             case UnknownType():
                 return UnknownType()
 

midas/checker/reporter.py

@@ -61,3 +61,10 @@ class FileReporter:
             location=location,
             message=message,
         )
+
+    def debug(self, location: Location, message: str):
+        self.report(
+            type=DiagnosticType.DEBUG,
+            location=location,
+            message=message,
+        )

midas/checker/resolver.py

@@ -128,6 +128,10 @@ class Resolver(p.Stmt.Visitor[None], p.Expr.Visitor[None]):
 
             case p.GetExpr():
                 target.accept(self)
+
+            case p.SubscriptExpr():
+                target.accept(self)
+
             case _:
                 raise Exception(f"Unsupported assignment to {target}")
 
@@ -232,5 +236,9 @@ class Resolver(p.Stmt.Visitor[None], p.Expr.Visitor[None]):
         if expr.step is not None:
             self.resolve(expr.step)
 
+    def visit_tuple_expr(self, expr: p.TupleExpr) -> None:
+        for item in expr.items:
+            self.resolve(item)
+
     def visit_raw_expr(self, expr: p.RawExpr) -> None:
         pass

midas/checker/types.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 from dataclasses import dataclass, field
 from enum import StrEnum
-from typing import Optional, assert_never
+from typing import Optional, assert_never, cast
 
 import midas.ast.midas as m
 from midas.ast.printer import MidasPrinter
@@ -23,7 +23,7 @@ class BaseType:
 
 
 @dataclass(frozen=True, kw_only=True)
-class AliasType:
+class DerivedType:
     name: str
     type: Type
 
@@ -156,6 +156,37 @@ class ConstraintType:
         return f"{self.type} where {printer.print(self.constraint)}"
 
 
+@dataclass(frozen=True, kw_only=True)
+class TupleType:
+    items: tuple[Type, ...]
+
+    def __str__(self) -> str:
+        return f"({', '.join(map(str, self.items))})"
+
+
+@dataclass(frozen=True, kw_only=True)
+class ColumnType:
+    type: Type
+
+    def __str__(self) -> str:
+        return f"Column[{self.type}]"
+
+
+@dataclass(frozen=True, kw_only=True)
+class DataFrameType:
+    columns: list[Column]
+
+    def __str__(self) -> str:
+        schema: list[str] = [f"{col.name}: {col.type}" for col in self.columns]
+        return f"Frame[{', '.join(schema)}]"
+
+    @dataclass(frozen=True, kw_only=True)
+    class Column:
+        index: int
+        name: Optional[str]
+        type: ColumnType
+
+
 def substitute_typevars(type: Type, substitutions: dict[str, Type]) -> Type:
     def sub_argument(arg: Function.Argument):
         return Function.Argument(
@@ -165,15 +196,27 @@ def substitute_typevars(type: Type, substitutions: dict[str, Type]) -> Type:
             required=arg.required,
         )
 
+    def sub_column(col: DataFrameType.Column):
+        return DataFrameType.Column(
+            index=col.index,
+            name=col.name,
+            type=cast(ColumnType, substitute_typevars(col.type, substitutions)),
+        )
+
     match type:
+        case TopType():
+            return type
+
         case BaseType(name=name) if name in substitutions:
             return substitutions[name]
 
         case BaseType():
             return type
 
-        case AliasType(name=name, type=type2):
+        case DerivedType(name=name, type=type2):
-            return AliasType(name=name, type=substitute_typevars(type2, substitutions))
+            return DerivedType(
+                name=name, type=substitute_typevars(type2, substitutions)
+            )
 
         case Function(
             pos_args=pos_args,
@@ -232,10 +275,41 @@ def substitute_typevars(type: Type, substitutions: dict[str, Type]) -> Type:
                 return substitutions[name]
             raise ValueError(f"Missing TypeVar substitution for {name}")
 
+        case GenericType(name=name, params=params, body=body):
+            params2: list[TypeVar] = []
+            for param in params:
+                param2: Type = substitute_typevars(param, substitutions)
+                if not isinstance(param2, TypeVar):
+                    raise ValueError(
+                        f"Invalid type parameter substitution, expected TypeVar, got {param2}"
+                    )
+                params2.append(param2)
+            return GenericType(
+                name=name,
+                params=params2,
+                body=substitute_typevars(body, substitutions),
+            )
+
+        case TupleType(items=items):
+            return TupleType(
+                items=tuple(substitute_typevars(item, substitutions) for item in items),
+            )
+
+        case ColumnType(type=items_type):
+            return ColumnType(
+                type=substitute_typevars(items_type, substitutions),
+            )
+
+        case DataFrameType(columns=columns):
+            return DataFrameType(
+                columns=list(map(sub_column, columns)),
+            )
+
         case UnknownType() | UnitType():
             return type
 
         case TopType() | GenericType():
+
             raise NotImplementedError(f"Unsupported type {type}")
 
         # Ensure exhaustiveness
@@ -245,7 +319,7 @@ def substitute_typevars(type: Type, substitutions: dict[str, Type]) -> Type:
 
 def unfold_type(type: Type) -> Type:
     match type:
-        case AliasType(type=ref_type):
+        case DerivedType(type=ref_type):
             return unfold_type(ref_type)
         case _:
             return type
@@ -268,7 +342,7 @@ def to_annotation(type: Type) -> str:
         case BaseType(name=name):
             return name
 
-        case AliasType(name=name):
+        case DerivedType(name=name):
             return name
 
         case UnknownType():
@@ -299,6 +373,15 @@ def to_annotation(type: Type) -> str:
         case ConstraintType():
             return str(type)
 
+        case TupleType(items=items):
+            return f"Tuple[{', '.join(map(to_annotation, items))}]"
+
+        case ColumnType():
+            return "pd.Series"
+
+        case DataFrameType():
+            return "pd.DataFrame"
+
         case _:
             assert_never(type)
 
@@ -313,7 +396,7 @@ class Predicate:
 Type = (
     TopType
     | BaseType
-    | AliasType
+    | DerivedType
     | UnknownType
     | UnitType
     | Function
@@ -324,4 +407,7 @@ Type = (
     | GenericType
     | AppliedType
     | ConstraintType
+    | TupleType
+    | ColumnType
+    | DataFrameType
 )

midas/checker/unifier.py

@@ -0,0 +1,169 @@
+import logging
+from typing import Optional
+
+from midas.checker.registry import TypesRegistry
+from midas.checker.types import (
+    AppliedType,
+    Function,
+    GenericType,
+    TopType,
+    Type,
+    TypeVar,
+)
+
+
+class UnificationError(Exception): ...
+
+
+class Unifier:
+    def __init__(self, types: TypesRegistry) -> None:
+        self.types: TypesRegistry = types
+        self.logger: logging.Logger = logging.getLogger("Unifier")
+
+    def unify_call(
+        self,
+        type: GenericType,
+        positional: list[Type],
+        keywords: dict[str, Type],
+    ) -> Optional[Type]:
+        concrete_func: Function = Function(
+            pos_args=[
+                Function.Argument(
+                    pos=i,
+                    name=str(i),
+                    type=arg,
+                    required=True,
+                )
+                for i, arg in enumerate(positional)
+            ],
+            args=[],
+            kw_args=[
+                Function.Argument(
+                    pos=len(positional) + i,
+                    name=name,
+                    type=arg,
+                    required=True,
+                )
+                for i, (name, arg) in enumerate(keywords.items())
+            ],
+            returns=TopType(),  # TODO: use expected type
+        )
+        return self.unify_generic(type, concrete_func, match_return=False)
+
+    def unify_generic(
+        self,
+        template: GenericType,
+        concrete: Type,
+        match_return: bool = True,
+    ) -> Optional[Type]:
+        substitutions: dict[str, Type]
+        try:
+            substitutions = self.match(template.body, concrete, match_return)
+        except UnificationError:
+            return None
+
+        args: list[Type] = []
+        for param in template.params:
+            if param.name not in substitutions:
+                return None
+            args.append(substitutions[param.name])
+
+        applied: Type = self.types.apply_generic(template, args)
+        return applied
+
+    def match(
+        self,
+        template: Type,
+        concrete: Type,
+        match_return: bool = True,
+    ) -> dict[str, Type]:
+        # TODO: if concrete is Generic, record bound TypeVar. Then when merging
+        # substitutions, check that the constraint is respected
+        match (template, concrete):
+            case (TypeVar(name=name), _):
+                return {name: concrete}
+
+            case (
+                AppliedType(name=template_name, args=template_args),
+                AppliedType(name=concrete_name, args=concrete_args),
+            ) if template_name == concrete_name and len(template_args) == len(
+                concrete_args
+            ):
+                substitutions: dict[str, Type] = {}
+                for template_arg, concrete_arg in zip(template_args, concrete_args):
+                    new_substistutions: dict[str, Type] = self.match(
+                        template_arg, concrete_arg
+                    )
+                    substitutions = self.merge(substitutions, new_substistutions)
+
+                return substitutions
+
+            case (Function(), Function()):
+                mapped: list[tuple[Function.Argument, Function.Argument]] = (
+                    self.map_params(template, concrete)
+                )
+                substitutions: dict[str, Type] = {}
+                for template_arg, concrete_arg in mapped:
+                    arg_subs: dict[str, Type] = self.match(
+                        template_arg.type, concrete_arg.type
+                    )
+                    substitutions = self.merge(substitutions, arg_subs)
+
+                if match_return:
+                    return_subs: dict[str, Type] = self.match(
+                        template.returns, concrete.returns
+                    )
+                    substitutions = self.merge(substitutions, return_subs)
+
+                return substitutions
+
+            case _:
+                self.logger.debug(f"Can't match {concrete!r} with {template!r}")
+                return {}
+
+    def merge(self, subs1: dict[str, Type], subs2: dict[str, Type]) -> dict[str, Type]:
+        merged: dict[str, Type] = subs1.copy()
+
+        for k, v in subs2.items():
+            if k in merged and merged[k] != v:
+                self.logger.debug(
+                    f"Substitution already defined for {k} with type {merged[k]}, got {v}"
+                )
+                raise UnificationError
+            merged[k] = v
+        return merged
+
+    def map_params(
+        self, func1: Function, func2: Function
+    ) -> list[tuple[Function.Argument, Function.Argument]]:
+        pos1: list[Function.Argument] = func1.pos_args
+        mixed1: list[Function.Argument] = func1.args
+        kw1: list[Function.Argument] = func1.kw_args
+
+        pos2: list[Function.Argument] = func2.pos_args
+        mixed2: list[Function.Argument] = func2.args
+        kw2: list[Function.Argument] = func2.kw_args
+
+        mapped: list[tuple[Function.Argument, Function.Argument]] = []
+
+        by_pos2: dict[int, Function.Argument] = {arg.pos: arg for arg in pos2 + mixed2}
+        by_name2: dict[str, Function.Argument] = {arg.name: arg for arg in mixed2 + kw2}
+
+        for arg1 in pos1:
+            if (arg2 := by_pos2.get(arg1.pos)) is not None:
+                mapped.append((arg1, arg2))
+
+        for arg1 in mixed1:
+            # Match both positionally and by name, conflicts are caught
+            # when merging substitutions
+            if (arg2 := by_pos2.get(arg1.pos)) is not None:
+                mapped.append((arg1, arg2))
+
+            if (arg2 := by_name2.get(arg1.name)) is not None:
+                mapped.append((arg1, arg2))
+
+        for arg1 in kw1:
+            if (arg2 := by_name2.get(arg1.name)) is not None:
+                mapped.append((arg1, arg2))
+
+        return mapped

midas/checker/variance.py

@@ -1,6 +1,6 @@
 from typing import Literal, Optional, cast
 
-from midas.checker.registry import TypesRegistry
+from midas.checker.registry import Member, TypesRegistry
 from midas.checker.types import (
     AppliedType,
     ConstraintType,
@@ -54,9 +54,9 @@ class VarianceInferrer:
         self.tracker = Tracker(type.params)
 
         self.walk(type.body, 1, type.name)
-        members: dict[str, Type] = self.types._members.get(type.name, {})
+        members: dict[str, Member] = self.types._members.get(type.name, {})
         for name, member in members.items():
-            self.walk(member, 1, type.name, [f"member:'{name}'"])
+            self.walk(member.type, 1, type.name, [f"member:'{name}'"])
 
         return GenericType(
             name=type.name,

midas/cli/commands/__init__.py

@@ -4,5 +4,6 @@ from .format import format as format
 from .highlight import highlight as highlight
 from .parse import parse as parse
 from .registry import dump_registry as dump_registry
+from .stubs import stubs as stubs
 from .types import types as types
 from .validate import validate as validate

midas/cli/commands/compile.py

@@ -5,7 +5,7 @@
 
 import sys
 from pathlib import Path
-from typing import TextIO
+from typing import Optional, TextIO
 
 import click
 
@@ -19,24 +19,33 @@ from midas.utils import TypedAST
 @click.command(help="Compile source")
 @click.argument("file", type=click.File("r"))
 @click.option("-t", "--types", type=click.File("r"), multiple=True)
+@click.option("-s", "--stubs", type=str, multiple=True)
+@click.option("--ignore-errors", is_flag=True)
 def compile(
     file: TextIO,
     types: tuple[TextIO],
+    stubs: tuple[str],
+    ignore_errors: bool,
 ):
     source: str = file.read()
     source_path: Path = Path(file.name).resolve()
 
     checker = TypeChecker()
-    for types_file in types:
+    type_files: list[tuple[Path, Optional[str]]] = []
-        checker.import_midas(Path(types_file.name).resolve())
+    for i, types_file in enumerate(types):
+        in_path: Path = Path(types_file.name).resolve()
+        checker.import_midas(in_path)
+        type_files.append((in_path, stubs[i] if i < len(stubs) else None))
 
     typed_ast: TypedAST = checker.type_check_source(source, str(source_path))
     diagnostics: list[Diagnostic] = checker.diagnostics.copy()
     printer = DiagnosticPrinter()
     printer.print_all(diagnostics)
 
-    if any(map(lambda d: d.type == DiagnosticType.ERROR, diagnostics)):
+    if not ignore_errors and any(
+        map(lambda d: d.type == DiagnosticType.ERROR, diagnostics)
+    ):
         sys.exit(1)
 
     generator = Generator(workdir=source_path.parent, types=checker.types)
-    generator.generate(typed_ast, source_path)
+    generator.generate(typed_ast, source_path, type_files=type_files)

midas/cli/commands/registry.py

@@ -10,14 +10,15 @@ import click
 
 from midas.ast.printer import MidasPrinter
 from midas.checker.checker import TypeChecker
-from midas.checker.types import AliasType, AppliedType, BaseType, GenericType, Type
+from midas.checker.registry import Member
+from midas.checker.types import AppliedType, BaseType, DerivedType, GenericType, Type
 
 
 def base_type(type: Type) -> Type:
     match type:
         case BaseType():
             return type
-        case AliasType(type=base):
+        case DerivedType(type=base):
             return base
         case AppliedType(body=body):
             return body
@@ -38,7 +39,7 @@ def dump_registry(
 
     print("##### Types #####")
     for name, type in checker.types._types.items():
-        members: dict[str, Type] = checker.types._members.get(name, {})
+        members: dict[str, Member] = checker.types._members.get(name, {})
         params: str = ""
         if isinstance(type, GenericType):
             params = ", ".join(map(str, type.params))
@@ -46,8 +47,9 @@ def dump_registry(
         print(f"{name}{params} = {base_type(type)}")
         if len(members) != 0:
             print(" " * 4 + "Members:")
-            for member_name, member_type in members.items():
+            for member_name, member in members.items():
-                print(" " * 8 + f"{member_name}: {member_type}")
+                kind: str = member.kind.name
+                print(" " * 8 + f"({kind:8}) {member_name}: {member.type}")
 
     print("##### Predicates #####")
     printer = MidasPrinter()

midas/cli/commands/stubs.py

@@ -0,0 +1,66 @@
+import ast
+import time
+from pathlib import Path
+from typing import Optional, TextIO
+
+import black
+import click
+from watchdog.events import DirModifiedEvent, FileModifiedEvent, FileSystemEventHandler
+from watchdog.observers import Observer
+
+from midas.checker.checker import TypeChecker
+from midas.generator.stubs import StubsGenerator
+
+
+def generate_stubs(in_path: Path, out_path: Path):
+    checker = TypeChecker()
+    checker.import_midas(in_path)
+
+    generator = StubsGenerator(checker.types)
+    module: ast.Module = generator.generate_stubs()
+    module = ast.fix_missing_locations(module)
+
+    output: str = ast.unparse(module)
+    output = black.format_str(output, mode=black.Mode(is_pyi=True))
+
+    out_path.write_text(output)
+
+
+class Handler(FileSystemEventHandler):
+    def __init__(self, in_path: Path, out_path: Path) -> None:
+        super().__init__()
+        self.in_path: Path = in_path
+        self.out_path: Path = out_path
+
+    def on_modified(self, event: DirModifiedEvent | FileModifiedEvent) -> None:
+        generate_stubs(self.in_path, self.out_path)
+
+
+@click.command(help="Generate stubs from Midas definitions")
+@click.argument("file", type=click.File("r"))
+@click.option("-o", "--output", type=click.File("w"))
+@click.option("-w", "--watch", is_flag=True)
+def stubs(
+    file: TextIO,
+    output: Optional[TextIO],
+    watch: bool,
+):
+    source_path: Path = Path(file.name).resolve()
+    out_path: Path = source_path.with_suffix(".pyi")
+    if output is not None:
+        out_path = Path(output.name).resolve()
+    generate_stubs(source_path, out_path)
+
+    if watch:
+        print(f"Watching {source_path}...")
+        print("Press CTRL+C to stop")
+        handler = Handler(source_path, out_path)
+        observer = Observer()
+        observer.schedule(handler, str(source_path))
+        observer.start()
+        try:
+            while True:
+                time.sleep(1)
+        except KeyboardInterrupt:
+            observer.stop()
+        observer.join()

midas/cli/commands/types.py

@@ -41,6 +41,7 @@ def types(
                 message=f"Type: {type}",
             )
         )
+    diagnostics.extend(checker.diagnostics)
     printer = DiagnosticPrinter()
     printer.print_all(diagnostics)
 

midas/cli/highlighter.py

@@ -134,9 +134,9 @@ class PythonHighlighter(
 
     def visit_base_type(self, node: p.BaseType) -> None:
         self.wrap(node, "base-type")
-        if node.param is not None:
+        for arg in node.args:
-            self.wrap(node.param, "param")
+            self.wrap(arg, "arg")
-            node.param.accept(self)
+            arg.accept(self)
 
     def visit_constraint_type(self, node: p.ConstraintType) -> None:
         self.wrap(node, "constraint-type")
@@ -228,6 +228,13 @@ class PythonHighlighter(
         for item in expr.items:
             item.accept(self)
 
+    def visit_dict_expr(self, expr: p.DictExpr) -> None:
+        for key in expr.keys:
+            if key is not None:
+                key.accept(self)
+        for value in expr.values:
+            value.accept(self)
+
     def visit_subscript_expr(self, expr: p.SubscriptExpr) -> None:
         expr.object.accept(self)
         expr.index.accept(self)
@@ -240,6 +247,14 @@ class PythonHighlighter(
         if expr.step is not None:
             expr.step.accept(self)
 
+    def visit_tuple_expr(self, expr: p.TupleExpr) -> None:
+        for item in expr.items:
+            item.accept(self)
+
+    def visit_raw_expr(self, expr: p.RawExpr) -> None: ...
+
+    def visit_raw_stmt(self, stmt: p.RawStmt) -> None: ...
+
 
 class MidasHighlighter(
     Highlighter, m.Stmt.Visitor[None], m.Expr.Visitor[None], m.Type.Visitor[None]
@@ -266,8 +281,9 @@ class MidasHighlighter(
     def visit_predicate_stmt(self, stmt: m.PredicateStmt) -> None:
         self.wrap(stmt, "predicate")
         self.wrap(LocatableToken(stmt.name), "predicate-name")
-        stmt.type.accept(self)
+        for spec in stmt.params:
-        stmt.condition.accept(self)
+            self._visit_param_spec(spec)
+        stmt.body.accept(self)
 
     def visit_logical_expr(self, expr: m.LogicalExpr) -> None:
         self.wrap(expr, "logical-expr")
@@ -283,6 +299,14 @@ class MidasHighlighter(
         self.wrap(expr, "unary-expr")
         expr.right.accept(self)
 
+    def visit_call_expr(self, expr: m.CallExpr) -> None:
+        self.wrap(expr, "call-expr")
+        expr.callee.accept(self)
+        for arg in expr.arguments:
+            arg.accept(self)
+        for arg in expr.keywords.values():
+            arg.accept(self)
+
     def visit_get_expr(self, expr: m.GetExpr) -> None:
         self.wrap(expr, "get-expr")
         expr.expr.accept(self)
@@ -318,8 +342,7 @@ class MidasHighlighter(
 
     def visit_function_type(self, type: m.FunctionType) -> None:
         self.wrap(type, "function")
-        for arg in type.pos_args + type.args + type.kw_args:
+        self._visit_param_spec(type.params)
-            arg.type.accept(self)
         type.returns.accept(self)
 
     def visit_extension_type(self, type: m.ExtensionType) -> None:
@@ -327,6 +350,18 @@ class MidasHighlighter(
         type.base.accept(self)
         type.extension.accept(self)
 
+    def _visit_param_spec(self, spec: m.ParamSpec) -> None:
+        for param in spec.pos + spec.mixed + spec.kw:
+            param.type.accept(self)
+
+    def visit_frame_type(self, type: m.FrameType) -> None:
+        self.wrap(type, "frame")
+        for column in type.columns:
+            self._visit_frame_column(column)
+
+    def _visit_frame_column(self, column: m.FrameType.Column) -> None:
+        self.wrap(column, "column")
+
 
 class DiagnosticsHighlighter(Highlighter):
     EXTRA_CSS_PATH: Optional[Path] = Path(__file__).parent / "hl_diagnostic.css"

midas/cli/hl_python.css

@@ -3,7 +3,7 @@ span {
         --col: 108, 233, 108;
     }
 
-    &.param {
+    &.arg {
         --col: 103, 192, 224;
     }
 

midas/cli/main.py

@@ -18,6 +18,7 @@ midas.add_command(commands.highlight)
 midas.add_command(commands.parse)
 midas.add_command(commands.dump_registry)
 midas.add_command(commands.types)
+midas.add_command(commands.stubs)
 midas.add_command(commands.validate)
 
 

midas/cli/utils.py

@@ -1,3 +1,4 @@
+from collections import defaultdict
 from pathlib import Path
 from typing import Optional
 
@@ -7,6 +8,13 @@ from midas.cli.ansi import Ansi
 
 
 class DiagnosticPrinter:
+    COLORS: dict[DiagnosticType, int] = {
+        DiagnosticType.ERROR: Ansi.RED,
+        DiagnosticType.WARNING: Ansi.YELLOW,
+        DiagnosticType.INFO: Ansi.CYAN,
+        DiagnosticType.DEBUG: Ansi.MAGENTA,
+    }
+
     def __init__(self) -> None:
         self.files: dict[Optional[str], list[str]] = {}
 
@@ -22,10 +30,25 @@ class DiagnosticPrinter:
         return self.files[filename]
 
     def print_all(self, diagnostics: list[Diagnostic], indent: int = 4):
+        by_type: dict[DiagnosticType, int] = defaultdict(int)
         for diagnostic in diagnostics:
             filename: Optional[str] = diagnostic.file_path
             lines = self.get_lines(filename)
             self.print(lines, diagnostic, indent=indent)
+            by_type[diagnostic.type] += 1
+
+        if len(diagnostics) == 0:
+            return
+
+        counts: list[str] = []
+        for type in DiagnosticType:
+            if type not in by_type:
+                continue
+            count: int = by_type[type]
+            color: int = self.COLORS.get(type, Ansi.WHITE)
+            counts.append(f"{Ansi.FG(color)}{type.value}s{Ansi.RESET}: {count}")
+
+        print("    ".join(counts))
 
     def print(self, lines: list[str], diagnostic: Diagnostic, indent: int = 4):
         """Pretty-print a diagnostic, showing some context if possible
@@ -45,7 +68,7 @@ class DiagnosticPrinter:
 
         loc: Location = diagnostic.location
         if loc.lineno != loc.end_lineno:
-            print(diagnostic)
+            self.print_multiline(lines, diagnostic, indent)
             return
 
         start_offset: int = loc.col_offset
@@ -55,11 +78,7 @@ class DiagnosticPrinter:
         before: str = line[:start_offset]
         after: str = line[end_offset:]
 
-        color: int = {
+        color: int = self.COLORS.get(diagnostic.type, Ansi.WHITE)
-            DiagnosticType.ERROR: Ansi.RED,
-            DiagnosticType.WARNING: Ansi.YELLOW,
-            DiagnosticType.INFO: Ansi.CYAN,
-        }.get(diagnostic.type, Ansi.WHITE)
 
         subject: str = Ansi.FG(color) + line[start_offset:end_offset] + Ansi.RESET
         cursor: str = (
@@ -76,3 +95,27 @@ class DiagnosticPrinter:
         print(indent_str + before + subject + after)
         print(indent_str + cursor)
         print()
+
+    def print_multiline(
+        self, all_lines: list[str], diagnostic: Diagnostic, indent: int = 4
+    ):
+        loc: Location = diagnostic.location
+        lines: list[str] = all_lines[loc.lineno - 1 : loc.end_lineno]
+
+        start_offset: int = loc.col_offset
+        end_offset: int = loc.end_col_offset or (start_offset + 1)
+
+        indent_str: str = " " * indent
+        color: int = self.COLORS.get(diagnostic.type, Ansi.WHITE)
+        res: str = indent_str + lines[0][:start_offset]
+        res += Ansi.FG(color) + lines[0][start_offset:]
+        for line in lines[1:-1]:
+            res += "\n" + indent_str + line
+        res += "\n" + indent_str + lines[-1][:end_offset]
+        res += Ansi.RESET + lines[-1][end_offset:]
+
+        print(diagnostic.location_str + ":")
+        print(res)
+        print()
+        print(Ansi.FG(color) + diagnostic.message + Ansi.RESET)
+        print()

midas/generator/generator.py

@@ -1,4 +1,5 @@
 import ast
+import logging
 import shutil
 from dataclasses import dataclass, field
 from pathlib import Path
@@ -8,42 +9,52 @@ import midas.ast.midas as m
 import midas.ast.python as p
 from midas.ast.location import Location
 from midas.ast.printer import MidasPrinter
+from midas.checker.checker import TypeChecker
 from midas.checker.registry import TypesRegistry
 from midas.checker.types import (
-    AliasType,
     AppliedType,
     BaseType,
+    ColumnType,
     ComplexType,
     ConstraintType,
+    DataFrameType,
+    DerivedType,
     ExtensionType,
     Function,
     GenericType,
     OverloadedFunction,
     TopType,
+    TupleType,
     Type,
     TypeVar,
     UnitType,
     UnknownType,
 )
 from midas.generator.constraints import ConstraintGenerator
+from midas.generator.stubs import StubsGenerator
 from midas.utils import TypedAST
 
 
 @dataclass
 class Scope:
-    pre_assertions: list[ast.stmt] = field(default_factory=list)
+    pre_assertions: list[ast.stmt] = field(default_factory=list[ast.stmt])
-    aliases: list[str] = field(default_factory=list)
+    aliases: list[str] = field(default_factory=list[str])
 
 
 class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
+    IS_DATAFRAME_FUNC = "__midas_is_dataframe__"
+    IS_COLUMN_FUNC = "__midas_is_column__"
+
     def __init__(self, workdir: Path, types: TypesRegistry) -> None:
         self.workdir: Path = workdir.resolve()
         self.build_dir: Path = self.workdir / "build" / "midas"
         self.rel_src_path: Path = Path()
+        self.logger: logging.Logger = logging.getLogger("Generator")
 
         self._typed_ast: TypedAST = TypedAST(
             stmts=[],
             judgements=[],
+            evaluated_casts=[],
         )
         self._alias_count: int = 0
         self._predicate_count: int = 0
@@ -52,20 +63,37 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
         self._constraint_generator: ConstraintGenerator = ConstraintGenerator(types)
         self._constraints: list[tuple[m.Expr, ast.expr]] = []
 
-    def generate_ast(self, typed_ast: TypedAST, src_path: Path) -> ast.AST:
+        self.define_is_dataframe: bool = False
-        self.rel_src_path = src_path.resolve().relative_to(self.workdir)
+        self.define_is_column: bool = False
+
+    def set_src_path(self, path: Path):
+        self.rel_src_path = path.resolve().relative_to(self.workdir)
+
+    def generate_ast(self, typed_ast: TypedAST) -> ast.AST:
         self._typed_ast = typed_ast
         body: list[ast.stmt] = self._visit_body(typed_ast.stmts)
         predicates: list[ast.stmt] = self._constraint_generator.get_definitions()
-        module = ast.Module(body=predicates + body, type_ignores=[])
+
+        body = predicates + body
+
+        if self.define_is_dataframe:
+            body = [self._is_dataframe_definition()] + body
+
+        if self.define_is_column:
+            body = [self._is_column_definition()] + body
+
+        module = ast.Module(body=body, type_ignores=[])
         module = ast.fix_missing_locations(module)
         return module
 
     def generate(
-        self, typed_ast: TypedAST, src_path: Path, out_path: Optional[Path] = None
+        self,
+        typed_ast: TypedAST,
+        src_path: Path,
+        out_path: Optional[Path] = None,
+        type_files: Optional[list[tuple[Path, Optional[str]]]] = None,
     ) -> Path:
-        module: ast.AST = self.generate_ast(typed_ast, src_path)
+        self.set_src_path(src_path)
-        compiled: str = ast.unparse(module)
         if out_path is None:
             if self.build_dir.exists():
                 shutil.rmtree(self.build_dir)
@@ -77,10 +105,30 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
                 raise ValueError(
                     f"Directory traversal, {self.rel_src_path} points outside of parent directory"
                 )
-            out_path.parent.mkdir(parents=True, exist_ok=True)
+        out_dir: Path = out_path.parent
+        out_dir.parent.mkdir(parents=True, exist_ok=True)
+
+        if type_files is not None:
+            for in_path, out_name in type_files:
+                if out_name is None:
+                    out_name = in_path.stem
+                self.generate_stubs(in_path, out_dir / f"{out_name}.py")
+
+        module: ast.AST = self.generate_ast(typed_ast)
+        compiled: str = ast.unparse(module)
+
         out_path.write_text(compiled)
         return out_path
 
+    def generate_stubs(self, in_path: Path, out_path: Path):
+        checker = TypeChecker()
+        checker.import_midas(in_path)
+        generator = StubsGenerator(checker.types)
+        module: ast.Module = generator.generate_stubs()
+        module = ast.fix_missing_locations(module)
+        output: str = ast.unparse(module)
+        out_path.write_text(output)
+
     def visit_binary_expr(self, expr: p.BinaryExpr) -> ast.expr:
         return ast.BinOp(
             left=expr.left.accept(self),
@@ -131,10 +179,16 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
 
     def visit_cast_expr(self, expr: p.CastExpr) -> ast.expr:
         expr2: ast.expr = expr.expr.accept(self)
+
+        if expr in self._typed_ast.evaluated_casts or expr.unsafe:
+            return expr2
+
         alias: ast.expr = self._make_alias(expr2)
 
         type: Type = self._get_expr_type(expr)
-        self._make_cast_asserts(expr.location, alias, type)
+        asserts: list[ast.stmt] = self._make_cast_asserts(expr.location, alias, type)
+        for assert_ in asserts:
+            self._add_assert(assert_)
 
         return alias
 
@@ -169,6 +223,11 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
             step=expr.step.accept(self) if expr.step is not None else None,
         )
 
+    def visit_tuple_expr(self, expr: p.TupleExpr) -> ast.expr:
+        return ast.Tuple(
+            elts=[item.accept(self) for item in expr.items],
+        )
+
     def visit_raw_expr(self, expr: p.RawExpr) -> ast.expr:
         return expr.expr
 
@@ -269,61 +328,156 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
         )
         return alias
 
-    def _add_assert(self, expr: ast.expr, message: str | ast.expr):
+    def _build_assert(self, expr: ast.expr, message: str | ast.expr) -> ast.stmt:
         if isinstance(message, str):
             message = ast.Constant(value=message)
-        self._scopes[-1].pre_assertions.append(
+        return ast.Assert(
-            ast.Assert(
+            test=expr,
-                test=expr,
+            msg=message,
-                msg=message,
-            )
         )
 
+    def _add_assert(self, assertion: ast.stmt):
+        self._scopes[-1].pre_assertions.append(assertion)
+
     def _get_expr_type(self, query: p.Expr) -> Type:
         for expr, type in self._typed_ast.judgements:
             if expr == query:
                 return type
         raise RuntimeError(f"Cannot get type judgement for {query}")
 
-    def _make_cast_asserts(self, src_location: Location, expr: ast.expr, type: Type):
+    def _make_cast_asserts(
+        self, src_location: Location, expr: ast.expr, type: Type
+    ) -> list[ast.stmt]:
         match type:
             case UnknownType():
-                pass
+                return []
 
             case BaseType(name=name):
-                self._add_assert(
+                return [
-                    ast.Call(
+                    self._build_assert(
-                        func=ast.Name(id="isinstance"),
+                        ast.Call(
-                        args=[expr, ast.Name(id=name)],
+                            func=ast.Name(id="isinstance"),
-                        keywords=[],
+                            args=[expr, ast.Name(id=name)],
-                    ),
+                            keywords=[],
-                    self._make_cast_assert_message(src_location, expr, type),
+                        ),
-                )
+                        self._make_cast_assert_message(src_location, expr, type),
+                    )
+                ]
 
-            case AliasType(type=base):
+            case DerivedType(type=base):
-                self._make_cast_asserts(src_location, expr, base)
+                return self._make_cast_asserts(src_location, expr, base)
 
             case UnitType():
-                self._add_assert(
+                return [
-                    ast.Compare(
+                    self._build_assert(
-                        left=expr,
+                        ast.Compare(
-                        ops=[ast.Is()],
+                            left=expr,
-                        comparators=[
+                            ops=[ast.Is()],
-                            ast.Constant(value=None),
+                            comparators=[
-                        ],
+                                ast.Constant(value=None),
+                            ],
+                        ),
+                        self._make_cast_assert_message(src_location, expr, type),
                     ),
-                    self._make_cast_assert_message(src_location, expr, type),
+                ]
-                )
 
             case AppliedType(body=body):
-                self._make_cast_asserts(src_location, expr, body)
+                return self._make_cast_asserts(src_location, expr, body)
 
             case ConstraintType(type=base, constraint=constraint):
-                self._make_cast_asserts(src_location, expr, base)
+                asserts: list[ast.stmt] = self._make_cast_asserts(
-                self._make_constraint_assert(src_location, expr, constraint)
+                    src_location, expr, base
+                )
+                asserts.append(
+                    self._make_constraint_assert(src_location, expr, constraint)
+                )
+                return asserts
 
-            case TypeVar():
+            case TypeVar(bound=bound):
-                raise RuntimeError("Unexpected TypeVar")
+                # TODO: check with type from arguments / use call-site context
+                if bound is None:
+                    return []
+                return self._make_cast_asserts(src_location, expr, bound)
+
+            case TupleType(items=items):
+                asserts: list[ast.stmt] = [
+                    self._build_assert(
+                        ast.Call(
+                            func=ast.Name(id="isinstance"),
+                            args=[expr, ast.Name(id="tuple")],
+                            keywords=[],
+                        ),
+                        self._make_cast_assert_message(src_location, expr, type),
+                    ),
+                ]
+                assert isinstance(expr, ast.Tuple)
+                for item, item_type in zip(expr.elts, items):
+                    asserts.extend(
+                        self._make_cast_asserts(src_location, item, item_type)
+                    )
+                return asserts
+
+            case DataFrameType(columns=columns):
+                self.define_is_dataframe = True
+                asserts: list[ast.stmt] = [
+                    self._build_assert(
+                        ast.Call(
+                            func=ast.Name(id=self.IS_DATAFRAME_FUNC),
+                            args=[expr],
+                            keywords=[],
+                        ),
+                        self._make_cast_assert_message(
+                            src_location, expr, type, ": Not a dataframe"
+                        ),
+                    ),
+                ]
+                for column in columns:
+                    asserts.append(
+                        self._build_assert(
+                            ast.Compare(
+                                left=ast.Constant(value=column.name),
+                                ops=[ast.In()],
+                                comparators=[expr],
+                            ),
+                            self._make_cast_assert_message(
+                                src_location,
+                                expr,
+                                type,
+                                f": Missing column {column.name}",
+                            ),
+                        )
+                    )
+                    asserts.extend(
+                        self._make_cast_asserts(
+                            src_location,
+                            ast.Subscript(
+                                value=expr, slice=ast.Constant(value=column.name)
+                            ),
+                            column.type,
+                        )
+                    )
+                return asserts
+
+            case ColumnType():
+                self.define_is_column = True
+                asserts: list[ast.stmt] = [
+                    self._build_assert(
+                        ast.Call(
+                            func=ast.Name(id=self.IS_COLUMN_FUNC),
+                            args=[expr],
+                            keywords=[],
+                        ),
+                        self._make_cast_assert_message(
+                            src_location, expr, type, ": Not a column"
+                        ),
+                    ),
+                ]
+                inner_assert: Optional[ast.stmt] = self._make_column_inner_assert(
+                    src_location, expr, type
+                )
+                if inner_assert is not None:
+                    asserts.append(inner_assert)
+                return asserts
 
             case (
                 TopType()
@@ -333,14 +487,19 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
                 | ExtensionType()
                 | GenericType()
             ):
-                raise NotImplementedError(f"Can't make assertion for type {type}")
+                self.logger.warning(f"Can't make assertion for type {type}")
+                return []
 
             # Ensure exhaustiveness
             case _:
                 assert_never(type)
 
     def _make_cast_assert_message(
-        self, location: Location, expr: ast.expr, type: Type
+        self,
+        location: Location,
+        expr: ast.expr,
+        type: Type,
+        extra: Optional[str] = None,
     ) -> ast.expr:
         loc_str: str = f"{self.rel_src_path}:L{location.lineno}:{location.col_offset+1}"
         # f"file.py:L1:1: CastError: Cannot cast {type(expr).__name__} to Type"
@@ -358,15 +517,15 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
                     ),
                     conversion=-1,
                 ),
-                ast.Constant(f" to {type}"),
+                ast.Constant(f" to {type}{extra or ''}"),
             ]
         )
 
     def _make_constraint_assert(
         self, src_location: Location, expr: ast.expr, constraint: m.Expr
-    ):
+    ) -> ast.stmt:
         test_func: ast.expr = self._get_constraint(constraint)
-        self._add_assert(
+        return self._build_assert(
             ast.Call(
                 func=test_func,
                 args=[expr],
@@ -394,3 +553,90 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
         constraint: ast.expr = self._constraint_generator.generate(expr)
         self._constraints.append((expr, constraint))
         return constraint
+
+    def _is_dataframe_definition(self) -> ast.stmt:
+        """
+        def IS_DATAFRAME_FUNC(obj) -> bool:
+            import pandas as pd
+            return isinstance(obj, pd.DataFrame)
+        """
+
+        return ast.FunctionDef(
+            name=self.IS_DATAFRAME_FUNC,
+            args=ast.arguments(
+                posonlyargs=[ast.arg(arg="obj")],
+                args=[],
+                kwonlyargs=[],
+                defaults=[],
+                kw_defaults=[],
+            ),
+            body=[
+                ast.Import(names=[ast.alias(name="pandas", asname="pd")]),
+                ast.Return(
+                    value=ast.Call(
+                        func=ast.Name(id="isinstance"),
+                        args=[
+                            ast.Name(id="obj"),
+                            ast.Attribute(
+                                value=ast.Name(id="pd"),
+                                attr="DataFrame",
+                            ),
+                        ],
+                        keywords=[],
+                    )
+                ),
+            ],
+            decorator_list=[],
+            returns=ast.Name(id="bool"),
+        )
+
+    def _is_column_definition(self) -> ast.stmt:
+        """
+        def IS_COLUMN_FUNC(obj) -> bool:
+            import pandas as pd
+            return isinstance(obj, pd.Series)
+        """
+
+        return ast.FunctionDef(
+            name=self.IS_COLUMN_FUNC,
+            args=ast.arguments(
+                posonlyargs=[ast.arg(arg="obj")],
+                args=[],
+                kwonlyargs=[],
+                defaults=[],
+                kw_defaults=[],
+            ),
+            body=[
+                ast.Import(names=[ast.alias(name="pandas", asname="pd")]),
+                ast.Return(
+                    value=ast.Call(
+                        func=ast.Name(id="isinstance"),
+                        args=[
+                            ast.Name(id="obj"),
+                            ast.Attribute(
+                                value=ast.Name(id="pd"),
+                                attr="Series",
+                            ),
+                        ],
+                        keywords=[],
+                    )
+                ),
+            ],
+            decorator_list=[],
+            returns=ast.Name(id="bool"),
+        )
+
+    def _make_column_inner_assert(
+        self, src_location: Location, column: ast.expr, type: ColumnType
+    ) -> Optional[ast.stmt]:
+        # TODO: improve message, maybe chain contexts
+        col: ast.expr = ast.Name(id="col")
+        body: list[ast.stmt] = self._make_cast_asserts(src_location, col, type.type)
+        if len(body) == 0:
+            return None
+        return ast.For(
+            target=col,
+            iter=column,
+            body=body,
+            orelse=[],
+        )

midas/generator/stubs.py

@@ -0,0 +1,427 @@
+import ast
+from typing import Optional, assert_never
+
+import midas.ast.midas as m
+from midas.checker.registry import Member, TypesRegistry
+from midas.checker.types import (
+    AppliedType,
+    BaseType,
+    ColumnType,
+    ComplexType,
+    ConstraintType,
+    DataFrameType,
+    DerivedType,
+    ExtensionType,
+    Function,
+    GenericType,
+    OverloadedFunction,
+    TopType,
+    TupleType,
+    Type,
+    TypeVar,
+    UnitType,
+    UnknownType,
+    Variance,
+    substitute_typevars,
+)
+
+Empty = ast.Constant(value=...)
+
+
+class StubsGenerator:
+    def __init__(self, types: TypesRegistry) -> None:
+        self.types: TypesRegistry = types
+        self.stubs: list[ast.stmt] = []
+        self.typing_imports: set[str] = set()
+        self.import_pandas: bool = False
+        self.protocol_idx: int = 0
+        self.stub_idx: int = 0
+        self.type_var_idx: int = 0
+        self.substitutions: dict[str, dict[str, Type]] = {}
+
+    def generate_stubs(self) -> ast.Module:
+        self.stubs = []
+        self.typing_imports = set()
+        self.import_pandas = False
+        for name, type in self.types._types.items():
+            # Skip builtin types, not just based on name so the user can override
+            # TODO: check if added members on builtin type
+            match type:
+                case BaseType(name=name_) if name == name_:
+                    continue
+                case GenericType(
+                    name=name1,
+                    body=BaseType(name=name2),
+                ) if (
+                    name == name1 == name2
+                ):
+                    continue
+            self.generate_stub(name, type)
+
+        imports: list[ast.stmt] = [
+            ast.ImportFrom(
+                module="__future__",
+                names=[ast.alias(name="annotations")],
+                level=0,
+            )
+        ]
+        if len(self.typing_imports) != 0:
+            imports.append(
+                ast.ImportFrom(
+                    module="typing",
+                    names=[
+                        ast.alias(name=name) for name in sorted(self.typing_imports)
+                    ],
+                    level=0,
+                )
+            )
+        if self.import_pandas:
+            imports.append(
+                ast.Import(
+                    names=[
+                        ast.alias(
+                            name="pandas",
+                            asname="pd",
+                        )
+                    ],
+                )
+            )
+        return ast.Module(body=imports + self.stubs, type_ignores=[])
+
+    def generate_stub(self, name: str, type: Type):
+        base_type: Type = type
+
+        members: dict[str, Member] = self.types._members.get(name, {})
+        if isinstance(base_type, (BaseType, TopType, UnitType)) and len(members) == 0:
+            return
+
+        bases: list[ast.expr] = []
+        substitutions: dict[str, Type] = {}
+        bases, substitutions = self.get_bases(type)
+        self.substitutions[name] = substitutions
+
+        body = self.generate_body(members, substitutions)
+        stub = ast.ClassDef(
+            name=name,
+            bases=bases,
+            body=body,
+            keywords=[],
+            decorator_list=[],
+        )
+        self.add_stub(stub)
+
+    def get_bases(self, type: Type) -> tuple[list[ast.expr], dict[str, Type]]:
+        match type:
+            case DerivedType(type=base):
+                return [self.dump_type(base)], {}
+
+            case GenericType(params=params, body=body):
+                self.add_typing_import("Generic")
+                type_vars: ast.expr
+
+                params2: list[TypeVar] = self.define_type_vars(params)
+                if len(params) == 1:
+                    type_vars = ast.Name(id=params2[0].name)
+                else:
+                    type_vars = ast.Tuple(
+                        elts=[ast.Name(id=param.name) for param in params2]
+                    )
+
+                substitutions: dict[str, TypeVar] = {
+                    param.name: param2 for param, param2 in zip(params, params2)
+                }
+
+                body_bases, body_subsitutions = self.get_bases(body)
+                return (
+                    body_bases
+                    + [
+                        ast.Subscript(
+                            value=ast.Name(id="Generic"),
+                            slice=type_vars,
+                        )
+                    ],
+                    body_subsitutions | substitutions,
+                )
+
+            case ConstraintType(type=base):
+                return self.get_bases(base)
+
+            case TypeVar(bound=bound) if bound is not None:
+                return [self.dump_type(bound)], {}
+
+            case _:
+                return [], {}
+
+    def generate_body(
+        self, members: dict[str, Member], substitutions: dict[str, Type]
+    ) -> list[ast.stmt]:
+        if len(members) == 0:
+            return [ast.Expr(value=Empty)]
+
+        body: list[ast.stmt] = []
+        for name, member in members.items():
+            type: Type = member.type
+            type = substitute_typevars(type, substitutions)
+            match member.kind:
+                case m.MemberKind.PROPERTY:
+                    body.append(
+                        ast.AnnAssign(
+                            target=ast.Name(id=name),
+                            annotation=self.dump_type(type),
+                            simple=1,
+                        )
+                    )
+                case m.MemberKind.METHOD:
+                    body.extend(self.dump_method(name, type))
+        return body
+
+    def dump_type(self, type: Type) -> ast.expr:
+        match type:
+            case DerivedType(name=name) | GenericType(name=name) if (
+                name in self.substitutions
+            ):
+                type = substitute_typevars(type, self.substitutions[name])
+
+        match type:
+            case TopType() | UnknownType():
+                self.add_typing_import("Any")
+                return ast.Name(id="Any")
+
+            case BaseType(name=name):
+                return ast.Name(id=name)
+
+            case DerivedType(name=name):
+                return ast.Name(id=name)
+
+            case UnitType():
+                return ast.Constant(value=None)
+
+            case Function():
+                name: str = self.define_protocol(type)
+                return ast.Name(id=name)
+
+            case OverloadedFunction(overloads=overloads):
+                if len(overloads) == 1:
+                    return self.dump_type(overloads[0])
+                return ast.BinOp(
+                    left=self.dump_type(OverloadedFunction(overloads=overloads[:-1])),
+                    op=ast.BitOr(),
+                    right=self.dump_type(overloads[-1]),
+                )
+
+            case ComplexType():
+                name: str = self.new_stub_name()
+                self.generate_stub(name, type)
+                return ast.Name(id=name)
+
+            case ExtensionType():
+                raise NotImplementedError
+
+            case TypeVar():
+                return ast.Name(id=type.name)
+
+            case GenericType(name=name):
+                params: ast.expr
+                if len(type.params) == 1:
+                    params = self.dump_type(type.params[0])
+                else:
+                    params = ast.Tuple(
+                        elts=[self.dump_type(param) for param in type.params]
+                    )
+                return ast.Subscript(
+                    value=ast.Name(id=type.name),
+                    slice=params,
+                )
+
+            case AppliedType():
+                args: ast.expr
+                if len(type.args) == 1:
+                    args = self.dump_type(type.args[0])
+                else:
+                    args = ast.Tuple(elts=[self.dump_type(arg) for arg in type.args])
+                return ast.Subscript(
+                    value=ast.Name(id=type.name),
+                    slice=args,
+                )
+
+            case ConstraintType():
+                return self.dump_type(type.type)
+
+            case TupleType(items=items):
+                return ast.Subscript(
+                    value=ast.Name(id="tuple"),
+                    slice=ast.Tuple(
+                        elts=[self.dump_type(item) for item in items],
+                    ),
+                )
+
+            case ColumnType(type=inner):
+                self.import_pandas = True
+                return ast.Subscript(
+                    value=ast.Attribute(
+                        value=ast.Name(id="pd"),
+                        attr="Series",
+                    ),
+                    slice=self.dump_type(inner),
+                )
+
+            case DataFrameType():
+                self.import_pandas = True
+                return ast.Attribute(
+                    value=ast.Name(id="pd"),
+                    attr="DataFrame",
+                )
+
+            case _:
+                assert_never(type)
+
+    def dump_method(
+        self, name: str, method: Type, overloaded: bool = False
+    ) -> list[ast.stmt]:
+        match method:
+            case Function():
+                if overloaded:
+                    self.add_typing_import("overload")
+                return [
+                    ast.FunctionDef(
+                        name=name,
+                        args=self.dump_args(method, with_self=True),
+                        returns=self.dump_type(method.returns),
+                        body=[ast.Expr(value=Empty)],
+                        decorator_list=[ast.Name(id="overload")] if overloaded else [],
+                    )
+                ]
+            case OverloadedFunction(overloads=overloads):
+                stmts: list[ast.stmt] = []
+                for overload in overloads:
+                    stmts.extend(self.dump_method(name, overload, True))
+                return stmts
+            case _:
+                return [
+                    ast.AnnAssign(
+                        target=ast.Name(id=name),
+                        annotation=self.dump_type(method),
+                        simple=1,
+                    )
+                ]
+
+    def dump_args(self, func: Function, with_self: bool = False) -> ast.arguments:
+        pos: list[ast.arg] = [
+            ast.arg(arg=f"_{arg.pos}", annotation=self.dump_type(arg.type))
+            for arg in func.pos_args
+        ]
+        mixed: list[ast.arg] = [
+            ast.arg(arg=arg.name, annotation=self.dump_type(arg.type))
+            for arg in func.args
+        ]
+        kw: list[ast.arg] = [
+            ast.arg(arg=arg.name, annotation=self.dump_type(arg.type))
+            for arg in func.kw_args
+        ]
+        defaults: list[ast.expr] = [
+            Empty for arg in func.pos_args + func.args if not arg.required
+        ]
+        kw_defaults: list[Optional[ast.expr]] = [
+            None if arg.required else Empty for arg in func.kw_args
+        ]
+        if with_self:
+            arg = ast.arg(arg="self", annotation=None)
+            if len(pos) != 0:
+                pos.insert(0, arg)
+            else:
+                mixed.insert(0, arg)
+        return ast.arguments(
+            posonlyargs=pos,
+            args=mixed,
+            kwonlyargs=kw,
+            defaults=defaults,
+            kw_defaults=kw_defaults,
+        )
+
+    def define_protocol(self, func: Function) -> str:
+        self.add_typing_import("Protocol")
+        name: str = self.new_protocol_name()
+        protocol = ast.ClassDef(
+            name=name,
+            bases=[ast.Name(id="Protocol")],
+            keywords=[],
+            body=[
+                ast.FunctionDef(
+                    name="__call__",
+                    args=self.dump_args(func, with_self=True),
+                    returns=self.dump_type(func.returns),
+                    body=[ast.Expr(value=Empty)],
+                    decorator_list=[],
+                ),
+            ],
+            decorator_list=[],
+        )
+        self.add_stub(protocol)
+        return name
+
+    def new_protocol_name(self) -> str:
+        name: str = f"_Protocol{self.protocol_idx}"
+        self.protocol_idx += 1
+        return name
+
+    def new_stub_name(self) -> str:
+        name: str = f"_Stub_{self.stub_idx}"
+        self.stub_idx += 1
+        return name
+
+    def new_type_var_name(self) -> str:
+        name: str = f"_T{self.type_var_idx}"
+        self.type_var_idx += 1
+        return name
+
+    def add_stub(self, stub: ast.stmt):
+        self.stubs.append(stub)
+
+    def add_typing_import(self, name: str):
+        self.typing_imports.add(name)
+
+    def define_type_vars(self, vars: list[TypeVar]) -> list[TypeVar]:
+        vars2: list[TypeVar] = []
+        for var in vars:
+            vars2.append(self.define_type_var(var))
+        return vars2
+
+    def define_type_var(self, var: TypeVar) -> TypeVar:
+        name: str = self.new_type_var_name()
+        self.add_typing_import("TypeVar")
+
+        kwargs: list[ast.keyword] = []
+        if var.bound is not None:
+            kwargs.append(
+                ast.keyword(
+                    arg="bound",
+                    value=self.dump_type(var.bound),
+                )
+            )
+        if var.variance == Variance.COVARIANT:
+            kwargs.append(
+                ast.keyword(
+                    arg="covariant",
+                    value=ast.Constant(value=True),
+                )
+            )
+        elif var.variance == Variance.CONTRAVARIANT:
+            kwargs.append(
+                ast.keyword(
+                    arg="contravariant",
+                    value=ast.Constant(value=True),
+                )
+            )
+        self.add_stub(
+            ast.Assign(
+                targets=[ast.Name(id=name)],
+                value=ast.Call(
+                    func=ast.Name(id="TypeVar"),
+                    args=[
+                        ast.Constant(value=name),
+                    ],
+                    keywords=kwargs,
+                ),
+            )
+        )
+        return TypeVar(name=name, bound=None)

midas/lexer/midas.py

@@ -46,8 +46,8 @@ class MidasLexer(Lexer):
                 self.add_token(TokenType.UNDERSCORE)
             case "-" if self.match(">"):
                 self.add_token(TokenType.ARROW)
-            # case "+":
+            case "+":
-            #     self.add_token(TokenType.PLUS)
+                self.add_token(TokenType.PLUS)
             case "-":
                 self.add_token(TokenType.MINUS)
             case "*":

midas/lexer/token.py

@@ -25,7 +25,7 @@ class TokenType(Enum):
     DOT = auto()
 
     # Operators
-    # PLUS = auto()
+    PLUS = auto()
     MINUS = auto()
     STAR = auto()
     SLASH = auto()
@@ -47,6 +47,7 @@ class TokenType(Enum):
 
     # Keywords
     TYPE = auto()
+    ALIAS = auto()
     PREDICATE = auto()
     EXTEND = auto()
     WHERE = auto()
@@ -63,6 +64,7 @@ class TokenType(Enum):
 
 KEYWORDS: dict[str, TokenType] = {
     "type": TokenType.TYPE,
+    "alias": TokenType.ALIAS,
     "predicate": TokenType.PREDICATE,
     "extend": TokenType.EXTEND,
     "where": TokenType.WHERE,

midas/parser/midas.py

@@ -2,6 +2,7 @@ from typing import Optional
 
 from midas.ast.location import Location
 from midas.ast.midas import (
+    AliasStmt,
     BinaryExpr,
     CallExpr,
     ComplexType,
@@ -9,6 +10,7 @@ from midas.ast.midas import (
     Expr,
     ExtendStmt,
     ExtensionType,
+    FrameType,
     FunctionType,
     GenericType,
     GetExpr,
@@ -79,6 +81,8 @@ class MidasParser(Parser):
         try:
             if self.match(TokenType.TYPE):
                 return self.type_declaration()
+            if self.match(TokenType.ALIAS):
+                return self.alias_declaration()
             if self.match(TokenType.EXTEND):
                 return self.extend_declaration()
             if self.match(TokenType.PREDICATE):
@@ -158,6 +162,25 @@ class MidasParser(Parser):
         self.consume(TokenType.RIGHT_BRACKET, "Missing ']' after type parameters")
         return params
 
+    def alias_declaration(self) -> AliasStmt:
+        """Parse an alias declaration
+
+        Returns:
+            AliasStmt: the parsed alias declaration statement
+        """
+        keyword: Token = self.previous()
+        name: Token = self.consume_identifier("Expected type name")
+
+        self.consume(TokenType.EQUAL, "Expected '=' before alias definition")
+
+        type: Type = self.type_expr()
+
+        return AliasStmt(
+            location=keyword.location_to(self.previous()),
+            name=name,
+            type=type,
+        )
+
     def type_expr(self) -> Type:
         """Parse a type expression
 
@@ -204,8 +227,10 @@ class MidasParser(Parser):
         return self.generic_type()
 
     def generic_type(self) -> Type:
-        type: Type = self.named_type()
+        type: NamedType = self.named_type()
         if self.check(TokenType.LEFT_BRACKET):
+            if type.name.lexeme == "Frame":
+                return self.frame_type()
             args: list[Type] = self.type_args()
             return GenericType(
                 location=Location.span(type.location, self.previous().get_location()),
@@ -224,7 +249,7 @@ class MidasParser(Parser):
         self.consume(TokenType.RIGHT_BRACKET, "Missing ']' after generic arguments")
         return args
 
-    def named_type(self) -> Type:
+    def named_type(self) -> NamedType:
         name: Token = self.consume_identifier("Expected type name")
         return NamedType(
             location=name.get_location(),
@@ -259,6 +284,32 @@ class MidasParser(Parser):
             members=members,
         )
 
+    def frame_type(self) -> FrameType:
+        keyword: Token = self.previous()
+        self.consume(TokenType.LEFT_BRACKET, "Expected '[' to start frame schema")
+
+        columns: list[FrameType.Column] = []
+        while not self.check(TokenType.RIGHT_BRACKET) and not self.is_at_end():
+            name: Token = self.advance()
+            self.consume(TokenType.COLON, "Expected ':' between column name and type")
+            type: Type = self.type_expr()
+            columns.append(
+                FrameType.Column(
+                    location=name.location_to(self.previous()),
+                    name=name,
+                    type=type,
+                )
+            )
+            if not self.match(TokenType.COMMA):
+                break
+
+        self.consume(TokenType.RIGHT_BRACKET, "Unclosed frame schema")
+
+        return FrameType(
+            location=keyword.location_to(self.previous()),
+            columns=columns,
+        )
+
     def constraint(self) -> Expr:
         """Parse a constraint
 
@@ -310,13 +361,35 @@ class MidasParser(Parser):
         Returns:
             Expr: the parsed expression
         """
-        expr: Expr = self.unary()
+        expr: Expr = self.term()
         while self.match(
             TokenType.LESS,
             TokenType.LESS_EQUAL,
             TokenType.GREATER,
             TokenType.GREATER_EQUAL,
         ):
+            operator: Token = self.previous()
+            right: Expr = self.term()
+            location: Location = Location.span(expr.location, right.location)
+            expr = BinaryExpr(
+                location=location, left=expr, operator=operator, right=right
+            )
+        return expr
+
+    def term(self) -> Expr:
+        expr: Expr = self.factor()
+        while self.match(TokenType.PLUS, TokenType.MINUS):
+            operator: Token = self.previous()
+            right: Expr = self.factor()
+            location: Location = Location.span(expr.location, right.location)
+            expr = BinaryExpr(
+                location=location, left=expr, operator=operator, right=right
+            )
+        return expr
+
+    def factor(self) -> Expr:
+        expr: Expr = self.unary()
+        while self.match(TokenType.STAR, TokenType.SLASH):
             operator: Token = self.previous()
             right: Expr = self.unary()
             location: Location = Location.span(expr.location, right.location)
@@ -348,7 +421,7 @@ class MidasParser(Parser):
         pos_args: list[Expr] = []
         kw_args: dict[str, Expr] = {}
         keywords: bool = False
-        while not self.match(TokenType.RIGHT_PAREN):
+        while not self.check(TokenType.RIGHT_PAREN):
             if self.check_identifier() and self.check_next(TokenType.EQUAL):
                 keywords = True
                 keyword: Token = self.advance()

midas/parser/python.py

@@ -30,6 +30,7 @@ from midas.ast.python import (
     Stmt,
     SubscriptExpr,
     TernaryExpr,
+    TupleExpr,
     TypeAssign,
     UnaryExpr,
     VariableExpr,
@@ -49,6 +50,7 @@ class UnsupportedSyntaxError(Exception):
 
 class PythonParser:
     CAST_FUNCTION = "cast"
+    UNSAFE_CAST_FUNCTION = "unsafe_cast"
 
     def parse_module(self, node: ast.Module) -> list[Stmt]:
         statements: list[Stmt] = []
@@ -299,26 +301,28 @@ class PythonParser:
             case ast.Subscript(value=ast.Name(id="Frame"), slice=schema):
                 return self._parse_frame_type(schema)
 
-            case ast.Subscript(value=ast.Name(id=name), slice=param):
+            case ast.Subscript(value=ast.Name(id=name), slice=arg):
+                args: tuple[MidasType, ...] = (
+                    tuple(self._parse_type(a) for a in arg.elts)
+                    if isinstance(arg, ast.Tuple)
+                    else (self._parse_type(arg),)
+                )
                 return BaseType(
                     location=loc,
                     base=name,
-                    param=self._parse_type(param),
+                    args=args,
                 )
 
             case ast.Name(id=name):
                 return BaseType(
                     location=loc,
                     base=name,
-                    param=None,
+                    args=(),
                 )
 
             case ast.BinOp(left=left_expr, op=ast.Add(), right=right_expr):
                 left = self._parse_type(left_expr)
                 match left:
-                    case None:
-                        raise InvalidSyntaxError()
-
                     # If chained constraints, separate base type and rebuild constraint
                     case ConstraintType(type=left_type, constraint=left_constraint):
                         constraint = ast.BinOp(
@@ -344,7 +348,7 @@ class PythonParser:
                 return BaseType(
                     location=loc,
                     base="None",
-                    param=None,
+                    args=(),
                 )
 
             case _:
@@ -423,6 +427,9 @@ class PythonParser:
             case ast.Call(func=ast.Name(id=self.CAST_FUNCTION)):
                 return self.parse_cast(node)
 
+            case ast.Call(func=ast.Name(id=self.UNSAFE_CAST_FUNCTION)):
+                return self.parse_cast(node)
+
             case ast.Call():
                 return self.parse_call(node)
 
@@ -473,6 +480,12 @@ class PythonParser:
                     step=self.parse_expr(step) if step is not None else None,
                 )
 
+            case ast.Tuple(elts=items):
+                return TupleExpr(
+                    location=location,
+                    items=tuple(self.parse_expr(item) for item in items),
+                )
+
             case _:
                 print(f"Unsupported expression: {ast.unparse(node)}")
                 return RawExpr(location=location, expr=node)
@@ -527,16 +540,19 @@ class PythonParser:
         return expr
 
     def parse_cast(self, node: ast.Call) -> CastExpr:
+        assert isinstance(node.func, ast.Name)
+        func: str = node.func.id
         match node:
             case ast.Call(args=[type, expr], keywords=[]):
                 return CastExpr(
                     location=Location.from_ast(node),
                     type=self._parse_type(type),
                     expr=self.parse_expr(expr),
+                    unsafe=func == self.UNSAFE_CAST_FUNCTION,
                 )
             case _:
                 raise InvalidSyntaxError(
-                    f"Invalid call to {self.CAST_FUNCTION}, expected type and expression"
+                    f"Invalid call to {func}, expected type and expression"
                 )
 
     def parse_call(self, node: ast.Call) -> CallExpr:

midas/typing.py

@@ -0,0 +1,52 @@
+from typing import Generic, TypeVar
+from typing import cast as typing_cast
+
+cast = typing_cast
+"""### Midas documentation
+Cast a value to a type.
+
+- **Compile-time**: tells the type checker that the return value has the designated type.
+- **Run-time**: generates assertions to ensure the value can be interpreted as the given type.
+
+---
+<br>
+<br>
+<br>
+
+_**Internal Python documentation**_
+"""
+
+
+unsafe_cast = typing_cast
+"""### Midas documentation
+Cast a value to a type.
+
+- **Compile-time**: tells the type checker that the return value has the designated type.
+- **Run-time**: -
+
+This operation is unsound, use at your own risk!
+
+---
+<br>
+<br>
+<br>
+
+_**Internal Python documentation**_
+"""
+
+
+T = TypeVar("T")
+
+
+class Frame(Generic[T]):
+    """A `Frame` is the abstract type implemented by `DataFrame`
+
+    A frame contains any number of named columns (see :class:`Column`)
+    """
+
+
+class Column(Generic[T]):
+    """A `Column` is the abstract type implemented by `Series`
+
+    A column contains a any number of values of the same type
+    """

midas/utils.py

@@ -62,3 +62,4 @@ class UniversalJSONDumper:
 class TypedAST:
     stmts: list[p.Stmt]
     judgements: list[tuple[p.Expr, Type]]
+    evaluated_casts: list[p.CastExpr]

pyproject.toml

@@ -8,7 +8,11 @@ authors = [
     { name = "Louis Heredero", email = "louis.heredero@students.hevs.ch" },
 ]
 classifiers = ["Programming Language :: Python :: 3"]
-dependencies = ["click>=8.4.1"]
+dependencies = [
+    "black>=26.5.1",
+    "click>=8.4.1",
+    "watchdog>=6.0.0",
+]
 
 [project.urls]
 Homepage = "https://git.kbk28.ch/HEL/midas"

tests/cases/checker/01_simple_types.py.ref.json

@@ -4,7 +4,35 @@
       "type": "Warning",
       "location": {
         "start": [
-          6,
+          8,
+          12
+        ],
+        "end": [
+          8,
+          43
+        ]
+      },
+      "message": "ConstraintType not yet supported"
+    },
+    {
+      "type": "Warning",
+      "location": {
+        "start": [
+          10,
+          10
+        ],
+        "end": [
+          10,
+          18
+        ]
+      },
+      "message": "Unknown type 'datetime'"
+    },
+    {
+      "type": "Warning",
+      "location": {
+        "start": [
+          13,
           4
         ],
         "end": [
@@ -12,7 +40,7 @@
           5
         ]
       },
-      "message": "FrameType not yet supported"
+      "message": "Unknown type '_'"
     }
   ],
   "judgments": []

tests/cases/checker/03_functions.py.ref.json

@@ -328,6 +328,19 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L6:9",
+        "to": "L6:10"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
     {
       "location": {
         "from": "L6:5",
@@ -373,19 +386,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L6:9",
-        "to": "L6:10"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
     {
       "location": {
         "from": "L6:5",
@@ -407,6 +407,32 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L7:9",
+        "to": "L7:10"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L7:12",
+        "to": "L7:15"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L7:5",
@@ -452,32 +478,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L7:9",
-        "to": "L7:10"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L7:12",
-        "to": "L7:15"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
     {
       "location": {
         "from": "L7:5",
@@ -503,6 +503,32 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L8:9",
+        "to": "L8:10"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L8:14",
+        "to": "L8:17"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L8:5",
@@ -548,32 +574,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L8:9",
-        "to": "L8:10"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L8:14",
-        "to": "L8:17"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
     {
       "location": {
         "from": "L8:5",
@@ -600,6 +600,45 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L9:9",
+        "to": "L9:10"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:12",
+        "to": "L9:15"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:17",
+        "to": "L9:23"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": "test"
+      },
+      "type": {
+        "name": "str"
+      }
+    },
     {
       "location": {
         "from": "L9:5",
@@ -645,45 +684,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L9:9",
-        "to": "L9:10"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L9:12",
-        "to": "L9:15"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
-    {
-      "location": {
-        "from": "L9:17",
-        "to": "L9:23"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": "test"
-      },
-      "type": {
-        "name": "str"
-      }
-    },
     {
       "location": {
         "from": "L9:5",
@@ -713,6 +713,45 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L10:9",
+        "to": "L10:10"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:12",
+        "to": "L10:15"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:19",
+        "to": "L10:22"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 3.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L10:5",
@@ -758,45 +797,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L10:9",
-        "to": "L10:10"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L10:12",
-        "to": "L10:15"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
-    {
-      "location": {
-        "from": "L10:19",
-        "to": "L10:22"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 3.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
     {
       "location": {
         "from": "L10:5",
@@ -827,6 +827,19 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L11:11",
+        "to": "L11:12"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
     {
       "location": {
         "from": "L11:5",
@@ -872,19 +885,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L11:11",
-        "to": "L11:12"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
     {
       "location": {
         "from": "L11:5",
@@ -906,6 +906,19 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L12:11",
+        "to": "L12:17"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": "test"
+      },
+      "type": {
+        "name": "str"
+      }
+    },
     {
       "location": {
         "from": "L12:5",
@@ -951,19 +964,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L12:11",
-        "to": "L12:17"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": "test"
-      },
-      "type": {
-        "name": "str"
-      }
-    },
     {
       "location": {
         "from": "L12:5",
@@ -985,6 +985,45 @@
       },
       "type": {}
     },
+    {
+      "location": {
+        "from": "L14:10",
+        "to": "L14:11"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L14:13",
+        "to": "L14:16"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L14:20",
+        "to": "L14:26"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": "test"
+      },
+      "type": {
+        "name": "str"
+      }
+    },
     {
       "location": {
         "from": "L14:6",
@@ -1030,45 +1069,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L14:10",
-        "to": "L14:11"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L14:13",
-        "to": "L14:16"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
-    {
-      "location": {
-        "from": "L14:20",
-        "to": "L14:26"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": "test"
-      },
-      "type": {
-        "name": "str"
-      }
-    },
     {
       "location": {
         "from": "L14:6",
@@ -1101,6 +1101,45 @@
         "name": "bool"
       }
     },
+    {
+      "location": {
+        "from": "L15:10",
+        "to": "L15:11"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L15:15",
+        "to": "L15:18"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L15:22",
+        "to": "L15:28"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": "test"
+      },
+      "type": {
+        "name": "str"
+      }
+    },
     {
       "location": {
         "from": "L15:6",
@@ -1146,45 +1185,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L15:10",
-        "to": "L15:11"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L15:15",
-        "to": "L15:18"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
-    {
-      "location": {
-        "from": "L15:22",
-        "to": "L15:28"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": "test"
-      },
-      "type": {
-        "name": "str"
-      }
-    },
     {
       "location": {
         "from": "L15:6",
@@ -1217,6 +1217,45 @@
         "name": "bool"
       }
     },
+    {
+      "location": {
+        "from": "L16:10",
+        "to": "L16:11"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L16:15",
+        "to": "L16:21"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": "test"
+      },
+      "type": {
+        "name": "str"
+      }
+    },
+    {
+      "location": {
+        "from": "L16:25",
+        "to": "L16:28"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2.0
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L16:6",
@@ -1262,45 +1301,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L16:10",
-        "to": "L16:11"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 1
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L16:15",
-        "to": "L16:21"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": "test"
-      },
-      "type": {
-        "name": "str"
-      }
-    },
-    {
-      "location": {
-        "from": "L16:25",
-        "to": "L16:28"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 2.0
-      },
-      "type": {
-        "name": "float"
-      }
-    },
     {
       "location": {
         "from": "L16:6",
@@ -1333,6 +1333,45 @@
         "name": "bool"
       }
     },
+    {
+      "location": {
+        "from": "L18:10",
+        "to": "L18:13"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": "a"
+      },
+      "type": {
+        "name": "str"
+      }
+    },
+    {
+      "location": {
+        "from": "L18:15",
+        "to": "L18:16"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 3
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L18:20",
+        "to": "L18:25"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": false
+      },
+      "type": {
+        "name": "bool"
+      }
+    },
     {
       "location": {
         "from": "L18:6",
@@ -1378,45 +1417,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L18:10",
-        "to": "L18:13"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": "a"
-      },
-      "type": {
-        "name": "str"
-      }
-    },
-    {
-      "location": {
-        "from": "L18:15",
-        "to": "L18:16"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": 3
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L18:20",
-        "to": "L18:25"
-      },
-      "expr": {
-        "_type": "LiteralExpr",
-        "value": false
-      },
-      "type": {
-        "name": "bool"
-      }
-    },
     {
       "location": {
         "from": "L18:6",

tests/cases/checker/04_custom_types.py.ref.json

@@ -1,6 +1,19 @@
 {
   "diagnostics": [],
   "judgments": [
+    {
+      "location": {
+        "from": "L4:30",
+        "to": "L4:36"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 123.45
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L4:18",
@@ -11,12 +24,13 @@
         "type": {
           "_type": "BaseType",
           "base": "Meter",
-          "param": null
+          "args": []
         },
         "expr": {
           "_type": "LiteralExpr",
           "value": 123.45
-        }
+        },
+        "unsafe": false
       },
       "type": {
         "name": "Meter",
@@ -25,6 +39,19 @@
         }
       }
     },
+    {
+      "location": {
+        "from": "L5:28",
+        "to": "L5:31"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 6.7
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L5:15",
@@ -35,12 +62,13 @@
         "type": {
           "_type": "BaseType",
           "base": "Second",
-          "param": null
+          "args": []
         },
         "expr": {
           "_type": "LiteralExpr",
           "value": 6.7
-        }
+        },
+        "unsafe": false
       },
       "type": {
         "name": "Second",

tests/cases/checker/06_subtyping.py.ref.json

@@ -100,6 +100,32 @@
         "name": "float"
       }
     },
+    {
+      "location": {
+        "from": "L11:13",
+        "to": "L11:15"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "v1"
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L11:17",
+        "to": "L11:19"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "v2"
+      },
+      "type": {
+        "name": "float"
+      }
+    },
     {
       "location": {
         "from": "L11:5",
@@ -135,32 +161,6 @@
         }
       }
     },
-    {
-      "location": {
-        "from": "L11:13",
-        "to": "L11:15"
-      },
-      "expr": {
-        "_type": "VariableExpr",
-        "name": "v1"
-      },
-      "type": {
-        "name": "int"
-      }
-    },
-    {
-      "location": {
-        "from": "L11:17",
-        "to": "L11:19"
-      },
-      "expr": {
-        "_type": "VariableExpr",
-        "name": "v2"
-      },
-      "type": {
-        "name": "float"
-      }
-    },
     {
       "location": {
         "from": "L11:5",

tests/cases/checker/07_variance.py.ref.json

@@ -72,29 +72,6 @@
     }
   ],
   "judgments": [
-    {
-      "location": {
-        "from": "L26:0",
-        "to": "L26:5"
-      },
-      "expr": {
-        "_type": "VariableExpr",
-        "name": "print"
-      },
-      "type": {
-        "pos_args": [
-          {
-            "pos": 0,
-            "name": "object",
-            "type": {},
-            "required": true
-          }
-        ],
-        "args": [],
-        "kw_args": [],
-        "returns": {}
-      }
-    },
     {
       "location": {
         "from": "L27:4",
@@ -325,6 +302,29 @@
         }
       }
     },
+    {
+      "location": {
+        "from": "L26:0",
+        "to": "L26:5"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "print"
+      },
+      "type": {
+        "pos_args": [
+          {
+            "pos": 0,
+            "name": "object",
+            "type": {},
+            "required": false
+          }
+        ],
+        "args": [],
+        "kw_args": [],
+        "returns": {}
+      }
+    },
     {
       "location": {
         "from": "L26:0",

tests/cases/checker/08_unification.py

@@ -0,0 +1,14 @@
+def double(value: float) -> float:
+    return value * 2
+
+
+def is_odd(value: int) -> bool:
+    return bool(value % 2)
+
+
+floats: list[float] = [0.2, 0.5, 0.1, 1.2]
+ints: list[int] = [1, 2, 6, -3]
+
+doubled_floats = map(double, floats)
+doubled_ints = map(double, ints)
+odd_ints = map(is_odd, ints)

tests/cases/checker/08_unification.py.ref.json

@@ -0,0 +1,874 @@
+{
+  "diagnostics": [
+    {
+      "type": "Error",
+      "location": {
+        "start": [
+          13,
+          15
+        ],
+        "end": [
+          13,
+          32
+        ]
+      },
+      "message": "Could not unify map[T, U]=(transform: (v: T, /) -> U, iterable: list[T], /) -> list[U] with pos=[(value: float) -> float, list[int]] and kw={}"
+    }
+  ],
+  "judgments": [
+    {
+      "location": {
+        "from": "L2:11",
+        "to": "L2:16"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "value"
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L2:19",
+        "to": "L2:20"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L2:11",
+        "to": "L2:20"
+      },
+      "expr": {
+        "_type": "BinaryExpr",
+        "left": {
+          "_type": "VariableExpr",
+          "name": "value"
+        },
+        "operator": "*",
+        "right": {
+          "_type": "LiteralExpr",
+          "value": 2
+        }
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L6:16",
+        "to": "L6:21"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "value"
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L6:24",
+        "to": "L6:25"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L6:16",
+        "to": "L6:25"
+      },
+      "expr": {
+        "_type": "BinaryExpr",
+        "left": {
+          "_type": "VariableExpr",
+          "name": "value"
+        },
+        "operator": "%",
+        "right": {
+          "_type": "LiteralExpr",
+          "value": 2
+        }
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L6:11",
+        "to": "L6:15"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "bool"
+      },
+      "type": {
+        "pos_args": [
+          {
+            "pos": 0,
+            "name": "object",
+            "type": {},
+            "required": false
+          }
+        ],
+        "args": [],
+        "kw_args": [],
+        "returns": {
+          "name": "bool"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L6:11",
+        "to": "L6:26"
+      },
+      "expr": {
+        "_type": "CallExpr",
+        "callee": {
+          "_type": "VariableExpr",
+          "name": "bool"
+        },
+        "arguments": [
+          {
+            "_type": "BinaryExpr",
+            "left": {
+              "_type": "VariableExpr",
+              "name": "value"
+            },
+            "operator": "%",
+            "right": {
+              "_type": "LiteralExpr",
+              "value": 2
+            }
+          }
+        ],
+        "keywords": {}
+      },
+      "type": {
+        "name": "bool"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:23",
+        "to": "L9:26"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 0.2
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:28",
+        "to": "L9:31"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 0.5
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:33",
+        "to": "L9:36"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 0.1
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:38",
+        "to": "L9:41"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1.2
+      },
+      "type": {
+        "name": "float"
+      }
+    },
+    {
+      "location": {
+        "from": "L9:22",
+        "to": "L9:42"
+      },
+      "expr": {
+        "_type": "ListExpr",
+        "items": [
+          {
+            "_type": "LiteralExpr",
+            "value": 0.2
+          },
+          {
+            "_type": "LiteralExpr",
+            "value": 0.5
+          },
+          {
+            "_type": "LiteralExpr",
+            "value": 0.1
+          },
+          {
+            "_type": "LiteralExpr",
+            "value": 1.2
+          }
+        ]
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "float"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L10:19",
+        "to": "L10:20"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 1
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:22",
+        "to": "L10:23"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 2
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:25",
+        "to": "L10:26"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 6
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:29",
+        "to": "L10:30"
+      },
+      "expr": {
+        "_type": "LiteralExpr",
+        "value": 3
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:28",
+        "to": "L10:30"
+      },
+      "expr": {
+        "_type": "UnaryExpr",
+        "operator": "-",
+        "right": {
+          "_type": "LiteralExpr",
+          "value": 3
+        }
+      },
+      "type": {
+        "name": "int"
+      }
+    },
+    {
+      "location": {
+        "from": "L10:18",
+        "to": "L10:31"
+      },
+      "expr": {
+        "_type": "ListExpr",
+        "items": [
+          {
+            "_type": "LiteralExpr",
+            "value": 1
+          },
+          {
+            "_type": "LiteralExpr",
+            "value": 2
+          },
+          {
+            "_type": "LiteralExpr",
+            "value": 6
+          },
+          {
+            "_type": "UnaryExpr",
+            "operator": "-",
+            "right": {
+              "_type": "LiteralExpr",
+              "value": 3
+            }
+          }
+        ]
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "int"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L12:21",
+        "to": "L12:27"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "double"
+      },
+      "type": {
+        "pos_args": [],
+        "args": [
+          {
+            "pos": 0,
+            "name": "value",
+            "type": {
+              "name": "float"
+            },
+            "required": true
+          }
+        ],
+        "kw_args": [],
+        "returns": {
+          "name": "float"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L12:29",
+        "to": "L12:35"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "floats"
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "float"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L12:17",
+        "to": "L12:20"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "map"
+      },
+      "type": {
+        "name": "map",
+        "params": [
+          {
+            "name": "T",
+            "bound": null,
+            "variance": "INVARIANT"
+          },
+          {
+            "name": "U",
+            "bound": null,
+            "variance": "INVARIANT"
+          }
+        ],
+        "body": {
+          "pos_args": [
+            {
+              "pos": 0,
+              "name": "transform",
+              "type": {
+                "pos_args": [
+                  {
+                    "pos": 0,
+                    "name": "v",
+                    "type": {
+                      "name": "T",
+                      "bound": null,
+                      "variance": "INVARIANT"
+                    },
+                    "required": true
+                  }
+                ],
+                "args": [],
+                "kw_args": [],
+                "returns": {
+                  "name": "U",
+                  "bound": null,
+                  "variance": "INVARIANT"
+                }
+              },
+              "required": true
+            },
+            {
+              "pos": 1,
+              "name": "iterable",
+              "type": {
+                "name": "list",
+                "args": [
+                  {
+                    "name": "T",
+                    "bound": null,
+                    "variance": "INVARIANT"
+                  }
+                ],
+                "body": {
+                  "name": "list"
+                }
+              },
+              "required": true
+            }
+          ],
+          "args": [],
+          "kw_args": [],
+          "returns": {
+            "name": "list",
+            "args": [
+              {
+                "name": "U",
+                "bound": null,
+                "variance": "INVARIANT"
+              }
+            ],
+            "body": {
+              "name": "list"
+            }
+          }
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L12:17",
+        "to": "L12:36"
+      },
+      "expr": {
+        "_type": "CallExpr",
+        "callee": {
+          "_type": "VariableExpr",
+          "name": "map"
+        },
+        "arguments": [
+          {
+            "_type": "VariableExpr",
+            "name": "double"
+          },
+          {
+            "_type": "VariableExpr",
+            "name": "floats"
+          }
+        ],
+        "keywords": {}
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "float"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L13:19",
+        "to": "L13:25"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "double"
+      },
+      "type": {
+        "pos_args": [],
+        "args": [
+          {
+            "pos": 0,
+            "name": "value",
+            "type": {
+              "name": "float"
+            },
+            "required": true
+          }
+        ],
+        "kw_args": [],
+        "returns": {
+          "name": "float"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L13:27",
+        "to": "L13:31"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "ints"
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "int"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L13:15",
+        "to": "L13:18"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "map"
+      },
+      "type": {
+        "name": "map",
+        "params": [
+          {
+            "name": "T",
+            "bound": null,
+            "variance": "INVARIANT"
+          },
+          {
+            "name": "U",
+            "bound": null,
+            "variance": "INVARIANT"
+          }
+        ],
+        "body": {
+          "pos_args": [
+            {
+              "pos": 0,
+              "name": "transform",
+              "type": {
+                "pos_args": [
+                  {
+                    "pos": 0,
+                    "name": "v",
+                    "type": {
+                      "name": "T",
+                      "bound": null,
+                      "variance": "INVARIANT"
+                    },
+                    "required": true
+                  }
+                ],
+                "args": [],
+                "kw_args": [],
+                "returns": {
+                  "name": "U",
+                  "bound": null,
+                  "variance": "INVARIANT"
+                }
+              },
+              "required": true
+            },
+            {
+              "pos": 1,
+              "name": "iterable",
+              "type": {
+                "name": "list",
+                "args": [
+                  {
+                    "name": "T",
+                    "bound": null,
+                    "variance": "INVARIANT"
+                  }
+                ],
+                "body": {
+                  "name": "list"
+                }
+              },
+              "required": true
+            }
+          ],
+          "args": [],
+          "kw_args": [],
+          "returns": {
+            "name": "list",
+            "args": [
+              {
+                "name": "U",
+                "bound": null,
+                "variance": "INVARIANT"
+              }
+            ],
+            "body": {
+              "name": "list"
+            }
+          }
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L13:15",
+        "to": "L13:32"
+      },
+      "expr": {
+        "_type": "CallExpr",
+        "callee": {
+          "_type": "VariableExpr",
+          "name": "map"
+        },
+        "arguments": [
+          {
+            "_type": "VariableExpr",
+            "name": "double"
+          },
+          {
+            "_type": "VariableExpr",
+            "name": "ints"
+          }
+        ],
+        "keywords": {}
+      },
+      "type": {}
+    },
+    {
+      "location": {
+        "from": "L14:15",
+        "to": "L14:21"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "is_odd"
+      },
+      "type": {
+        "pos_args": [],
+        "args": [
+          {
+            "pos": 0,
+            "name": "value",
+            "type": {
+              "name": "int"
+            },
+            "required": true
+          }
+        ],
+        "kw_args": [],
+        "returns": {
+          "name": "bool"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L14:23",
+        "to": "L14:27"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "ints"
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "int"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L14:11",
+        "to": "L14:14"
+      },
+      "expr": {
+        "_type": "VariableExpr",
+        "name": "map"
+      },
+      "type": {
+        "name": "map",
+        "params": [
+          {
+            "name": "T",
+            "bound": null,
+            "variance": "INVARIANT"
+          },
+          {
+            "name": "U",
+            "bound": null,
+            "variance": "INVARIANT"
+          }
+        ],
+        "body": {
+          "pos_args": [
+            {
+              "pos": 0,
+              "name": "transform",
+              "type": {
+                "pos_args": [
+                  {
+                    "pos": 0,
+                    "name": "v",
+                    "type": {
+                      "name": "T",
+                      "bound": null,
+                      "variance": "INVARIANT"
+                    },
+                    "required": true
+                  }
+                ],
+                "args": [],
+                "kw_args": [],
+                "returns": {
+                  "name": "U",
+                  "bound": null,
+                  "variance": "INVARIANT"
+                }
+              },
+              "required": true
+            },
+            {
+              "pos": 1,
+              "name": "iterable",
+              "type": {
+                "name": "list",
+                "args": [
+                  {
+                    "name": "T",
+                    "bound": null,
+                    "variance": "INVARIANT"
+                  }
+                ],
+                "body": {
+                  "name": "list"
+                }
+              },
+              "required": true
+            }
+          ],
+          "args": [],
+          "kw_args": [],
+          "returns": {
+            "name": "list",
+            "args": [
+              {
+                "name": "U",
+                "bound": null,
+                "variance": "INVARIANT"
+              }
+            ],
+            "body": {
+              "name": "list"
+            }
+          }
+        }
+      }
+    },
+    {
+      "location": {
+        "from": "L14:11",
+        "to": "L14:28"
+      },
+      "expr": {
+        "_type": "CallExpr",
+        "callee": {
+          "_type": "VariableExpr",
+          "name": "map"
+        },
+        "arguments": [
+          {
+            "_type": "VariableExpr",
+            "name": "is_odd"
+          },
+          {
+            "_type": "VariableExpr",
+            "name": "ints"
+          }
+        ],
+        "keywords": {}
+      },
+      "type": {
+        "name": "list",
+        "args": [
+          {
+            "name": "bool"
+          }
+        ],
+        "body": {
+          "name": "list"
+        }
+      }
+    }
+  ]
+}

tests/cases/generator/01_simple_types.py.ref.txt

@@ -7,68 +7,14 @@ Module(
         alias(name='Meter'),
         alias(name='Second')],
       level=0),
-    Assign(
-      targets=[
-        Name(id='__midas_a0__')],
-      value=Constant(value=123.45)),
-    Assert(
-      test=Call(
-        func=Name(id='isinstance'),
-        args=[
-          Name(id='__midas_a0__'),
-          Name(id='float')],
-        keywords=[]),
-      msg=JoinedStr(
-        values=[
-          Constant(value='01_simple_types.py:L3:19: CastError: Cannot cast '),
-          FormattedValue(
-            value=Attribute(
-              value=Call(
-                func=Name(id='type'),
-                args=[
-                  Name(id='__midas_a0__')],
-                keywords=[]),
-              attr='__name__'),
-            conversion=-1),
-          Constant(value=' to float')])),
     Assign(
       targets=[
         Name(id='distance')],
-      value=Name(id='__midas_a0__')),
+      value=Constant(value=123.45)),
-    Delete(
-      targets=[
-        Name(id='__midas_a0__')]),
-    Assign(
-      targets=[
-        Name(id='__midas_a1__')],
-      value=Constant(value=6.7)),
-    Assert(
-      test=Call(
-        func=Name(id='isinstance'),
-        args=[
-          Name(id='__midas_a1__'),
-          Name(id='float')],
-        keywords=[]),
-      msg=JoinedStr(
-        values=[
-          Constant(value='01_simple_types.py:L4:16: CastError: Cannot cast '),
-          FormattedValue(
-            value=Attribute(
-              value=Call(
-                func=Name(id='type'),
-                args=[
-                  Name(id='__midas_a1__')],
-                keywords=[]),
-              attr='__name__'),
-            conversion=-1),
-          Constant(value=' to float')])),
     Assign(
       targets=[
         Name(id='time')],
-      value=Name(id='__midas_a1__')),
+      value=Constant(value=6.7)),
-    Delete(
-      targets=[
-        Name(id='__midas_a1__')]),
     Assign(
       targets=[
         Name(id='speed')],

tests/cases/python-parser/01_simple_types.py.ref.json

@@ -16,7 +16,7 @@
             "type": {
               "_type": "BaseType",
               "base": "bool",
-              "param": null
+              "args": []
             }
           },
           {
@@ -25,7 +25,7 @@
             "type": {
               "_type": "BaseType",
               "base": "int",
-              "param": null
+              "args": []
             }
           },
           {
@@ -36,7 +36,7 @@
               "type": {
                 "_type": "BaseType",
                 "base": "float",
-                "param": null
+                "args": []
               },
               "constraint": "(_ > 0) + (_ < 250)"
             }
@@ -47,7 +47,7 @@
             "type": {
               "_type": "BaseType",
               "base": "str",
-              "param": null
+              "args": []
             }
           },
           {
@@ -56,7 +56,7 @@
             "type": {
               "_type": "BaseType",
               "base": "datetime",
-              "param": null
+              "args": []
             }
           },
           {
@@ -65,7 +65,7 @@
             "type": {
               "_type": "BaseType",
               "base": "float",
-              "param": null
+              "args": []
             }
           },
           {
@@ -79,7 +79,7 @@
             "type": {
               "_type": "BaseType",
               "base": "_",
-              "param": null
+              "args": []
             }
           }
         ]

tests/cases/python-parser/02_custom_types.py.ref.json

@@ -16,7 +16,7 @@
             "type": {
               "_type": "BaseType",
               "base": "GeoLocation",
-              "param": null
+              "args": []
             }
           }
         ]
@@ -28,11 +28,13 @@
       "type": {
         "_type": "BaseType",
         "base": "Column",
-        "param": {
+        "args": [
-          "_type": "BaseType",
+          {
-          "base": "GeoLocation",
+            "_type": "BaseType",
-          "param": null
+            "base": "GeoLocation",
-        }
+            "args": []
+          }
+        ]
       }
     },
     {
@@ -65,11 +67,13 @@
       "type": {
         "_type": "BaseType",
         "base": "Column",
-        "param": {
+        "args": [
-          "_type": "BaseType",
+          {
-          "base": "GeoLocation",
+            "_type": "BaseType",
-          "param": null
+            "base": "GeoLocation",
-        }
+            "args": []
+          }
+        ]
       }
     },
     {
@@ -117,7 +121,7 @@
       "type": {
         "_type": "BaseType",
         "base": "Latitude",
-        "param": null
+        "args": []
       }
     },
     {
@@ -146,7 +150,7 @@
       "type": {
         "_type": "BaseType",
         "base": "Latitude",
-        "param": null
+        "args": []
       }
     },
     {
@@ -175,11 +179,13 @@
       "type": {
         "_type": "BaseType",
         "base": "Difference",
-        "param": {
+        "args": [
-          "_type": "BaseType",
+          {
-          "base": "Latitude",
+            "_type": "BaseType",
-          "param": null
+            "base": "Latitude",
-        }
+            "args": []
+          }
+        ]
       }
     },
     {
@@ -217,7 +223,7 @@
               "type": {
                 "_type": "BaseType",
                 "base": "int",
-                "param": null
+                "args": []
               },
               "constraint": "_ >= 0"
             }
@@ -230,7 +236,7 @@
               "type": {
                 "_type": "BaseType",
                 "base": "float",
-                "param": null
+                "args": []
               },
               "constraint": "_ >= 0"
             }
@@ -252,7 +258,7 @@
               "type": {
                 "_type": "BaseType",
                 "base": "int",
-                "param": null
+                "args": []
               },
               "constraint": "Positive"
             }
@@ -265,7 +271,7 @@
               "type": {
                 "_type": "BaseType",
                 "base": "float",
-                "param": null
+                "args": []
               },
               "constraint": "Positive"
             }

tests/cases/python-parser/03_functions.py.ref.json

@@ -14,15 +14,17 @@
           "type": {
             "_type": "BaseType",
             "base": "Column",
-            "param": {
+            "args": [
-              "_type": "ConstraintType",
+              {
-              "type": {
+                "_type": "ConstraintType",
-                "_type": "BaseType",
+                "type": {
-                "base": "float",
+                  "_type": "BaseType",
-                "param": null
+                  "base": "float",
-              },
+                  "args": []
-              "constraint": "0 <= _ <= 1"
+                },
-            }
+                "constraint": "0 <= _ <= 1"
+              }
+            ]
           },
           "default": null
         },
@@ -31,15 +33,17 @@
           "type": {
             "_type": "BaseType",
             "base": "Column",
-            "param": {
+            "args": [
-              "_type": "ConstraintType",
+              {
-              "type": {
+                "_type": "ConstraintType",
-                "_type": "BaseType",
+                "type": {
-                "base": "float",
+                  "_type": "BaseType",
-                "param": null
+                  "base": "float",
-              },
+                  "args": []
-              "constraint": "0 <= _ <= 1"
+                },
-            }
+                "constraint": "0 <= _ <= 1"
+              }
+            ]
           },
           "default": null
         }
@@ -50,15 +54,17 @@
       "returns": {
         "_type": "BaseType",
         "base": "Column",
-        "param": {
+        "args": [
-          "_type": "ConstraintType",
+          {
-          "type": {
+            "_type": "ConstraintType",
-            "_type": "BaseType",
+            "type": {
-            "base": "float",
+              "_type": "BaseType",
-            "param": null
+              "base": "float",
-          },
+              "args": []
-          "constraint": "0 <= _ <= 2"
+            },
-        }
+            "constraint": "0 <= _ <= 2"
+          }
+        ]
       },
       "body": [
         {
@@ -67,15 +73,17 @@
           "type": {
             "_type": "BaseType",
             "base": "Column",
-            "param": {
+            "args": [
-              "_type": "ConstraintType",
+              {
-              "type": {
+                "_type": "ConstraintType",
-                "_type": "BaseType",
+                "type": {
-                "base": "float",
+                  "_type": "BaseType",
-                "param": null
+                  "base": "float",
-              },
+                  "args": []
-              "constraint": "0 <= _ <= 2"
+                },
-            }
+                "constraint": "0 <= _ <= 2"
+              }
+            ]
           }
         },
         {
@@ -117,7 +125,7 @@
           "type": {
             "_type": "BaseType",
             "base": "int",
-            "param": null
+            "args": []
           },
           "default": null
         }
@@ -128,7 +136,7 @@
           "type": {
             "_type": "BaseType",
             "base": "float",
-            "param": null
+            "args": []
           },
           "default": null
         }
@@ -140,7 +148,7 @@
           "type": {
             "_type": "BaseType",
             "base": "str",
-            "param": null
+            "args": []
           },
           "default": null
         }

tests/generator.py

@@ -46,7 +46,8 @@ class GeneratorTester(Tester):
 
         if not any(d.type == DiagnosticType.ERROR for d in checker.diagnostics):
             generator = Generator(workdir=path.parent, types=checker.types)
-            result.compiled_ast = generator.generate_ast(typed_ast, path)
+            generator.set_src_path(path)
+            result.compiled_ast = generator.generate_ast(typed_ast)
 
         return result
 

tests/serializer/midas.py

@@ -1,6 +1,7 @@
 from typing import Optional, Sequence
 
 from midas.ast.midas import (
+    AliasStmt,
     BinaryExpr,
     CallExpr,
     ComplexType,
@@ -8,6 +9,7 @@ from midas.ast.midas import (
     Expr,
     ExtendStmt,
     ExtensionType,
+    FrameType,
     FunctionType,
     GenericType,
     GetExpr,
@@ -60,6 +62,13 @@ class MidasAstJsonSerializer(
             "bound": self._serialize_optional(param.bound),
         }
 
+    def visit_alias_stmt(self, stmt: AliasStmt) -> dict:
+        return {
+            "_type": "AliasStmt",
+            "name": stmt.name.lexeme,
+            "type": stmt.type.accept(self),
+        }
+
     def visit_member_stmt(self, stmt: MemberStmt) -> dict:
         return {
             "_type": "MemberStmt",
@@ -197,3 +206,15 @@ class MidasAstJsonSerializer(
             "base": type.base.accept(self),
             "extension": type.extension.accept(self),
         }
+
+    def visit_frame_type(self, type: FrameType) -> dict:
+        return {
+            "_type": "FrameType",
+            "columns": [self._serialize_column(col) for col in type.columns],
+        }
+
+    def _serialize_column(self, column: FrameType.Column):
+        return {
+            "name": column.name.lexeme,
+            "type": column.type.accept(self),
+        }

tests/serializer/python.py

@@ -30,6 +30,7 @@ from midas.ast.python import (
     Stmt,
     SubscriptExpr,
     TernaryExpr,
+    TupleExpr,
     TypeAssign,
     UnaryExpr,
     VariableExpr,
@@ -98,7 +99,7 @@ class PythonAstJsonSerializer(
         return {
             "_type": "BaseType",
             "base": node.base,
-            "param": self._serialize_optional(node.param),
+            "args": self._serialize_list(node.args),
         }
 
     def visit_constraint_type(self, node: ConstraintType) -> dict:
@@ -263,6 +264,7 @@ class PythonAstJsonSerializer(
             "_type": "CastExpr",
             "type": expr.type.accept(self),
             "expr": expr.expr.accept(self),
+            "unsafe": expr.unsafe,
         }
 
     def visit_ternary_expr(self, expr: TernaryExpr) -> dict:
@@ -301,6 +303,12 @@ class PythonAstJsonSerializer(
             "step": self._serialize_optional(expr.step),
         }
 
+    def visit_tuple_expr(self, expr: TupleExpr) -> dict:
+        return {
+            "_type": "TupleExpr",
+            "items": [item.accept(self) for item in expr.items],
+        }
+
     def visit_raw_expr(self, expr: RawExpr) -> dict:
         return {
             "_type": "RawExpr",