Compare commits

...

25 Commits

Author SHA1 Message Date
78e10e0895 feat(checker): process frame type definitions 2026-06-24 14:36:53 +02:00
c81e4a9560 feat(cli): add frame type to highlighter 2026-06-24 14:36:53 +02:00
6d0cf1a055 feat(parser): add frame type to midas syntax 2026-06-24 14:36:52 +02:00
cc5e7af143 feat(gen): add support for tuples and dataframes 2026-06-24 14:36:51 +02:00
3bdbc80079 feat(checker): handle setting dataframe column 2026-06-24 14:36:51 +02:00
c1b5284f72 feat(checker): type check subscript on dataframes 2026-06-24 14:36:28 +02:00
5e9ccd4e13 feat(types): add TupleType 2026-06-24 14:36:04 +02:00
cf083fc0c3 fix(types): add str methods to dataframe types 2026-06-24 14:35:31 +02:00
a80da5db2c feat(types): add DataFrameType and ColumnType 2026-06-24 14:35:30 +02:00
f7c43837b5 Merge pull request 'CLI tweaks' (#22) from fix/cli-tweaks into main
Reviewed-on: #22
2026-06-24 12:18:07 +00:00
32ed62a6f1 fix(cli): show summary of diagnostic counts 2026-06-24 14:11:39 +02:00
66f39acec0 fix(cli): show all diagnostics in types command
combine type checker diagnostics with judgements info diagnostics
2026-06-24 14:11:15 +02:00
6c04e2fee4 feat(cli): add compile option to ignore errors 2026-06-24 14:10:30 +02:00
2bb2e0a684 Merge pull request 'Unsafe cast' (#21) from feat/unsafe-cast into main
Reviewed-on: #21
2026-06-24 12:00:03 +00:00
5630320d21 chore: use unsafe_cast in demo script 2026-06-24 13:57:38 +02:00
9f05ba3224 feat: handle unsafe casts 2026-06-24 13:51:14 +02:00
5fbe965919 feat(checker): add typing submodule with cast functions 2026-06-24 13:40:23 +02:00
252a5abdfd Merge pull request 'Static evalution of casts on literals' (#20) from feat/literal-static-constraints into main
Reviewed-on: #20
2026-06-24 09:32:54 +00:00
55fba6a088 tests: update test without evaluated casts 2026-06-24 11:28:44 +02:00
70ce263ea2 feat(gen): skip assertions for evaluated casts
avoid generating a runtime assertion for a cast which has already been checked statically
2026-06-24 11:28:43 +02:00
e1d5eac8b8 feat(checker): evaluate constraints statically on literals 2026-06-24 11:10:09 +02:00
82666a4918 feat(checker): add evaluator
add an evaluator class to evaluate expressions using literal values
2026-06-24 11:08:15 +02:00
45f84a2f23 feat(checker): add debug diagnostics 2026-06-24 11:07:42 +02:00
dedfcb4dbb feat(checker): store builtin python functions in preamble 2026-06-24 11:05:36 +02:00
d9ea6365ea tests: update with cast expression judgement 2026-06-23 16:49:38 +02:00
29 changed files with 893 additions and 89 deletions

View File

@@ -1,6 +1,8 @@
from typing import TypeVar, cast
from typing import TypeVar
from demo_stubs import CHF, EUR, USD, Currency, Price, Discount
from demo_stubs import CHF, EUR, USD, Currency, Discount, Price
from midas.typing import cast, unsafe_cast
T = TypeVar("T", bound=Currency)
@@ -28,3 +30,6 @@ discounted = apply_discount(
)
print(f"Discounted: CHF {discounted}")
large_data = [i * 10 for i in range(100)]
prices = unsafe_cast(list[Price[EUR]], large_data)

View File

@@ -152,4 +152,14 @@ 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
###<

View File

@@ -145,6 +145,7 @@ class LogicalExpr:
class CastExpr:
type: MidasType
expr: Expr
unsafe: bool
class TernaryExpr:

View File

@@ -253,6 +253,9 @@ 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 +314,17 @@ 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)

View File

@@ -350,6 +350,25 @@ 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):
@@ -502,6 +521,23 @@ 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,
@@ -757,9 +793,10 @@ 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")

View File

@@ -350,6 +350,7 @@ 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)

View File

@@ -15,7 +15,7 @@ if TYPE_CHECKING:
BUILTIN_SUBTYPES: dict[str, set[str]] = {
"object": {"float", "list", "dict"},
"object": {"float", "list", "dict", "str"},
"float": {"int"},
"int": {"bool"},
}

View File

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

172
midas/checker/evaluator.py Normal file
View File

@@ -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)

133
midas/checker/frames.py Normal file
View File

@@ -0,0 +1,133 @@
from typing import Optional, TypeGuard, cast
from midas.ast.location import Location
from midas.checker.registry import TypesRegistry
from midas.checker.reporter import FileReporter
from midas.checker.types import ColumnType, DataFrameType, TupleType, Type, UnknownType
import midas.ast.python as p
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, types: TypesRegistry) -> None:
self.types: TypesRegistry = types
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]

View File

@@ -14,8 +14,10 @@ from midas.checker.reporter import FileReporter, Reporter
from midas.checker.types import (
AliasType,
AppliedType,
ColumnType,
ComplexType,
ConstraintType,
DataFrameType,
ExtensionType,
Function,
GenericType,
@@ -401,6 +403,18 @@ 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:

View File

@@ -1,4 +1,5 @@
from dataclasses import dataclass
from typing import Callable, Optional
from midas.checker.environment import Environment
from midas.checker.registry import TypesRegistry
@@ -16,16 +17,18 @@ class Preamble(Environment):
def __init__(self, types: TypesRegistry) -> None:
super().__init__()
self._types: TypesRegistry = types
self._python_funcs: dict[str, Callable] = {}
self._def_type_constructor("object")
self._def_type_constructor("float")
self._def_type_constructor("int")
self._def_type_constructor("bool")
self._def_type_constructor("str")
self._def_type_constructor("object", object)
self._def_type_constructor("float", float)
self._def_type_constructor("int", int)
self._def_type_constructor("bool", bool)
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
@@ -33,6 +36,7 @@ class Preamble(Environment):
name="print",
pos=[Param("object", TopType())],
returns=UnitType(),
py_function=print,
)
map_in = TypeVar(name="T", bound=None)
@@ -53,6 +57,7 @@ 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",
@@ -63,12 +68,13 @@ class Preamble(Environment):
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] = 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(
@@ -115,6 +121,7 @@ class Preamble(Environment):
kw: list[Param] = [],
returns: Type = UnitType(),
type_vars: list[TypeVar] = [],
py_function: Optional[Callable] = None,
):
function: Type = self._make_function(
name=name,
@@ -125,3 +132,8 @@ class Preamble(Environment):
type_vars=type_vars,
)
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]:
return self._python_funcs.get(name)

View File

@@ -1,11 +1,14 @@
import ast
import logging
from dataclasses import dataclass
from typing import Optional
from typing import Any, Optional
import midas.ast.python as p
from midas.ast.location import Location
from midas.ast.printer import MidasPrinter
from midas.checker.environment import Environment
from midas.checker.evaluator import Evaluator
from midas.checker.frames import FrameManager
from midas.checker.operators import (
PY_COMPARATOR_METHODS,
PY_OPERATOR_METHODS,
@@ -18,9 +21,14 @@ from midas.checker.resolver import Resolver
from midas.checker.types import (
AliasType,
AppliedType,
BaseType,
ColumnType,
ConstraintType,
DataFrameType,
Function,
GenericType,
OverloadedFunction,
TupleType,
Type,
TypeVar,
UnitType,
@@ -67,10 +75,12 @@ class PythonTyper(
self.logger: logging.Logger = logging.getLogger("PythonTyper")
self.reporter: FileReporter = reporter.for_file(None)
self.types: TypesRegistry = types
self.frame_mgr: FrameManager = FrameManager(self.types)
self.global_env: Environment = Preamble(self.types)
self.env: Environment = self.global_env
self.locals: dict[p.Expr, int] = {}
self.judgements: list[tuple[p.Expr, Type]] = []
self.evaluated_casts: list[p.CastExpr] = []
def process(self, source: str, path: Optional[str]) -> TypedAST:
self.reporter = self.reporter.for_file(path)
@@ -84,10 +94,15 @@ class PythonTyper(
self.env = self.global_env
self.locals = resolver.locals
self.judgements = []
self.evaluated_casts = []
self.check(stmts)
return TypedAST(stmts=stmts, judgements=self.judgements)
return TypedAST(
stmts=stmts,
judgements=self.judgements,
evaluated_casts=self.evaluated_casts,
)
def judge(self, expr: p.Expr, type: Type):
"""Record a typing judgement
@@ -309,9 +324,15 @@ class PythonTyper(
case p.VariableExpr():
self._assign_var(location, target, value_type)
# Allow any kind of object because we disallow creating new attributes
case p.GetExpr(object=object, name=name):
self._assign_attr(location, object, name, value_type)
# Only support variable expressions because modifying
# the underlying value would require reference types
case p.SubscriptExpr(object=p.VariableExpr() as var, index=index):
self._assign_sub(location, var, index, value_type)
case _:
if not isinstance(target, p.VariableExpr):
self.logger.warning(f"Unsupported assignment to {target}")
@@ -350,6 +371,27 @@ class PythonTyper(
f"Cannot assign {value_type} to member '{object_type}.{name}' of type {member}",
)
def _assign_sub(
self,
location: Location,
var: p.VariableExpr,
index: p.Expr,
value_type: Type,
):
var_type: Type = self.type_of(var)
# TODO: what happens if type is an alias of a dataframe type
match var_type:
case DataFrameType() as frame:
new_type: Type = self.frame_mgr.assign(
self.reporter, location, frame, index, value_type
)
self.env.assign(var.name, new_type)
case _:
self.reporter.error(
location,
f"Cannot assign {value_type} to index {index} of {var_type}",
)
def visit_return_stmt(self, stmt: p.ReturnStmt) -> None:
type: Type = self.type_of(stmt.value) if stmt.value is not None else UnitType()
self.env.return_types.append(type)
@@ -538,8 +580,16 @@ class PythonTyper(
return UnknownType()
def visit_cast_expr(self, expr: p.CastExpr) -> Type:
_ = self.type_of(expr.expr)
return self.resolve_type_expr(expr.type)
subject_type: Type = self.type_of(expr.expr)
target_type: Type = self.resolve_type_expr(expr.type)
is_lit, lit_value = self._get_literal(expr.expr)
if is_lit:
evaluated: bool = self._evaluate_cast_statically(
expr, subject_type, target_type, lit_value
)
if evaluated:
self.evaluated_casts.append(expr)
return target_type
def visit_ternary_expr(self, expr: p.TernaryExpr) -> Type:
test_type: Type = self.type_of(expr.test)
@@ -622,6 +672,13 @@ class PythonTyper(
def visit_subscript_expr(self, expr: p.SubscriptExpr) -> Type:
object: Type = self.type_of(expr.object)
unfolded: Type = unfold_type(object)
match unfolded:
case TupleType():
return self._visit_tuple_subscript(unfolded, expr)
case DataFrameType():
return self._visit_frame_subscript(unfolded, expr)
operation: Optional[Type] = self.types.lookup_member(object, "__getitem__")
if operation is None:
self.reporter.error(
@@ -659,13 +716,26 @@ class PythonTyper(
self.reporter.warning(node.location, "ConstraintType not yet supported")
return UnknownType()
def visit_frame_column(self, node: p.FrameColumn) -> Type:
self.reporter.warning(node.location, "FrameColumn not yet supported")
return UnknownType()
def visit_frame_column(self, node: p.FrameColumn) -> ColumnType:
return ColumnType(
type=(
self.resolve_type_expr(node.type)
if node.type is not None
else UnknownType()
)
)
def visit_frame_type(self, node: p.FrameType) -> Type:
self.reporter.warning(node.location, "FrameType not yet supported")
return UnknownType()
return DataFrameType(
columns=[
DataFrameType.Column(
index=i,
name=column.name,
type=self.visit_frame_column(column),
)
for i, column in enumerate(node.columns)
]
)
def _get_call_result(
self,
@@ -1108,3 +1178,113 @@ class PythonTyper(
return p.BaseType(location=location, base=name, param=None)
case _:
raise NotImplementedError
def _get_literal(self, expr: p.Expr) -> tuple[bool, Any]:
match expr:
case p.LiteralExpr(value=value):
return True, value
case p.ListExpr(items=items):
values: list[Any] = []
for item in items:
is_lit, value = self._get_literal(item)
if not is_lit:
return False, None
values.append(value)
return True, values
case p.DictExpr(keys=keys, values=values):
pairs: list[tuple[Any, Any]] = []
for key, value in zip(keys, values):
key_val = None
if key is not None:
is_lit, key_val = self._get_literal(key)
if not is_lit:
return False, None
is_lit, value_val = self._get_literal(value)
if not is_lit:
return False, None
if key is None:
# TODO: check that value is always a dict
assert isinstance(value_val, dict)
pairs.extend(value_val.items())
else:
pairs.append((key_val, value_val))
return True, dict(pairs)
case _:
return False, None
def _evaluate_cast_statically(
self, expr: p.CastExpr, subject_type: Type, target_type: Type, lit_value: Any
) -> bool:
match target_type:
case AliasType(type=base):
return self._evaluate_cast_statically(
expr, subject_type, base, lit_value
)
case AppliedType(body=body):
return self._evaluate_cast_statically(
expr, subject_type, body, lit_value
)
case ConstraintType(type=base, constraint=constraint):
evaluated: bool = True
if not self._evaluate_cast_statically(
expr, subject_type, base, lit_value
):
evaluated = False
evaluator = Evaluator(self.types)
evaluator.set_value("_", lit_value)
res = evaluator.evaluate(constraint)
if not res:
printer = MidasPrinter()
constraint_str: str = printer.print(constraint)
self.reporter.error(
expr.location,
f"Value {lit_value!r} does not fit constraint '{constraint_str}'",
)
evaluated = False
return evaluated
case BaseType():
# TODO: do we want to allow cast(float, int)? would require runtime conversion
if not self.types.is_subtype(
subject_type, target_type
) or not self.types.is_subtype(target_type, subject_type):
self.reporter.error(
expr.location,
f"Value {lit_value!r} of type {subject_type} cannot be cast as {target_type}",
)
return False
return True
case _:
self.reporter.info(
expr.location, f"Cannot evaluate cast to {target_type} statically"
)
return False
def _visit_tuple_subscript(self, tup: TupleType, expr: p.SubscriptExpr) -> Type:
match expr.index:
case p.LiteralExpr(value=int() as index):
if index < 0 or index >= len(tup.items):
self.reporter.error(
expr.location, f"Index {index} out of range for tuple {tup}"
)
return UnknownType()
return tup.items[index]
case _:
self.reporter.error(
expr.location, f"Invalid index type {expr.index} on {tup}"
)
return UnknownType()
def _visit_frame_subscript(
self, frame: DataFrameType, expr: p.SubscriptExpr
) -> Type:
return self.frame_mgr.get(self.reporter, expr.location, frame, expr.index)

View File

@@ -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,
)

View File

@@ -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}")

View File

@@ -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
@@ -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(
@@ -165,6 +196,13 @@ 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
@@ -250,10 +288,26 @@ def substitute_typevars(type: Type, substitutions: dict[str, Type]) -> Type:
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
@@ -317,6 +371,15 @@ 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)
@@ -342,4 +405,7 @@ Type = (
| GenericType
| AppliedType
| ConstraintType
| TupleType
| ColumnType
| DataFrameType
)

View File

@@ -19,9 +19,11 @@ 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("--ignore-errors", is_flag=True)
def compile(
file: TextIO,
types: tuple[TextIO],
ignore_errors: bool,
):
source: str = file.read()
source_path: Path = Path(file.name).resolve()
@@ -35,7 +37,9 @@ def compile(
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)

View File

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

View File

@@ -350,6 +350,14 @@ class MidasHighlighter(
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"

View File

@@ -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
@@ -55,11 +78,7 @@ class DiagnosticPrinter:
before: str = line[:start_offset]
after: str = line[end_offset:]
color: int = {
DiagnosticType.ERROR: Ansi.RED,
DiagnosticType.WARNING: Ansi.YELLOW,
DiagnosticType.INFO: Ansi.CYAN,
}.get(diagnostic.type, Ansi.WHITE)
color: int = self.COLORS.get(diagnostic.type, Ansi.WHITE)
subject: str = Ansi.FG(color) + line[start_offset:end_offset] + Ansi.RESET
cursor: str = (

View File

@@ -1,4 +1,5 @@
import ast
import logging
import shutil
from dataclasses import dataclass, field
from pathlib import Path
@@ -13,13 +14,16 @@ from midas.checker.types import (
AliasType,
AppliedType,
BaseType,
ColumnType,
ComplexType,
ConstraintType,
DataFrameType,
ExtensionType,
Function,
GenericType,
OverloadedFunction,
TopType,
TupleType,
Type,
TypeVar,
UnitType,
@@ -40,10 +44,12 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
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
@@ -131,6 +137,10 @@ 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)
@@ -327,6 +337,19 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
if bound is not None:
self._make_cast_asserts(src_location, expr, bound)
case TupleType(items=items):
self._add_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):
self._make_cast_asserts(src_location, item, item_type)
case (
TopType()
| Function()
@@ -334,8 +357,10 @@ class Generator(p.Stmt.Visitor[ast.stmt], p.Expr.Visitor[ast.expr]):
| ComplexType()
| ExtensionType()
| GenericType()
| ColumnType()
| DataFrameType()
):
raise NotImplementedError(f"Can't make assertion for type {type}")
self.logger.warning(f"Can't make assertion for type {type}")
# Ensure exhaustiveness
case _:

View File

@@ -7,13 +7,16 @@ from midas.checker.types import (
AliasType,
AppliedType,
BaseType,
ColumnType,
ComplexType,
ConstraintType,
DataFrameType,
ExtensionType,
Function,
GenericType,
OverloadedFunction,
TopType,
TupleType,
Type,
TypeVar,
UnitType,
@@ -30,6 +33,7 @@ class StubsGenerator:
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
@@ -38,6 +42,7 @@ class StubsGenerator:
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
@@ -53,7 +58,7 @@ class StubsGenerator:
continue
self.generate_stub(name, type)
imports = [
imports: list[ast.stmt] = [
ast.ImportFrom(
module="__future__",
names=[ast.alias(name="annotations")],
@@ -70,6 +75,17 @@ class StubsGenerator:
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):
@@ -231,6 +247,31 @@ class StubsGenerator:
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)

View File

@@ -9,6 +9,7 @@ from midas.ast.midas import (
Expr,
ExtendStmt,
ExtensionType,
FrameType,
FunctionType,
GenericType,
GetExpr,
@@ -204,8 +205,10 @@ 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 +227,7 @@ 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 +262,32 @@ 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

View File

@@ -49,6 +49,7 @@ 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] = []
@@ -423,6 +424,9 @@ 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)
@@ -527,16 +531,19 @@ 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:

34
midas/typing.py Normal file
View File

@@ -0,0 +1,34 @@
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**_
"""

View File

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

View File

@@ -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",
@@ -16,7 +29,8 @@
"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",
@@ -40,7 +67,8 @@
"expr": {
"_type": "LiteralExpr",
"value": 6.7
}
},
"unsafe": false
},
"type": {
"name": "Second",

View File

@@ -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__')),
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')])),
value=Constant(value=123.45)),
Assign(
targets=[
Name(id='time')],
value=Name(id='__midas_a1__')),
Delete(
targets=[
Name(id='__midas_a1__')]),
value=Constant(value=6.7)),
Assign(
targets=[
Name(id='speed')],

View File

@@ -263,6 +263,7 @@ 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: