better error message

This commit is contained in:
pca006132 2020-12-22 16:53:33 +08:00 committed by pca006132
parent 56215cbac1
commit 00ed5e867d
3 changed files with 81 additions and 48 deletions

View File

@ -1,9 +1,21 @@
import ast
class CustomError(Exception): class CustomError(Exception):
def __init__(self, msg): def __init__(self, msg, node = None):
self.msg = msg self.msg = msg
if node is not None:
self.at(node)
def at(self, node):
self.msg = f'Error at {node.lineno}:{node.col_offset+1}-' \
f'{node.end_lineno}:{node.end_col_offset+1} ' \
f'"{ast.unparse(node)}"\n{self.msg}'
return self
def stringify_subst(subst): def stringify_subst(subst):
if isinstance(subst, str): if isinstance(subst, str):
return subst return subst
elements = [f"{key}: {str(value)}" for key, value in subst.items()] elements = [f"{key}: {str(value)}" for key, value in subst.items()]
return "{" + ', '.join(elements) + "}" return "{" + ', '.join(elements) + "}"

View File

@ -5,7 +5,7 @@ from type_def import *
from inference import * from inference import *
# we assume having the following types: # we assume having the following types:
# bool, int32 with associated operations # bool, int32 and float with associated operations
# not handled now: named expression, type guard # not handled now: named expression, type guard
@ -42,7 +42,7 @@ def parse_expr(ctx: Context,
return parse_if_expr(ctx, sym_table, body) return parse_if_expr(ctx, sym_table, body)
if isinstance(body, ast.ListComp): if isinstance(body, ast.ListComp):
return parse_list_comprehension(ctx, sym_table, body) return parse_list_comprehension(ctx, sym_table, body)
raise CustomError(f'{body} is not yet supported') raise CustomError(f'{body} is not yet supported', body)
def get_unary_op(op): def get_unary_op(op):
@ -73,8 +73,10 @@ def parse_constant(ctx: Context,
return ctx.types['bool'] return ctx.types['bool']
elif isinstance(v, int): elif isinstance(v, int):
return ctx.types['int32'] return ctx.types['int32']
elif isinstance(v, float):
return ctx.types['float']
else: else:
raise CustomError(f'unknown constant {v}') raise CustomError(f'unknown constant {v}', node)
def parse_name(ctx: Context, def parse_name(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -82,7 +84,7 @@ def parse_name(ctx: Context,
if node.id in sym_table: if node.id in sym_table:
return sym_table[node.id] return sym_table[node.id]
else: else:
raise CustomError(f'unbounded variable {node.id}') raise CustomError(f'unbounded variable {node.id}', node)
def parse_list(ctx: Context, def parse_list(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -92,7 +94,7 @@ def parse_list(ctx: Context,
return ListType(BotType()) return ListType(BotType())
for t in types[1:]: for t in types[1:]:
if t != types[0]: if t != types[0]:
raise CustomError(f'inhomogeneous list is not allowed') raise CustomError(f'inhomogeneous list is not allowed', node)
return ListType(types[0]) return ListType(types[0])
def parse_tuple(ctx: Context, def parse_tuple(ctx: Context,
@ -107,7 +109,7 @@ def parse_attribute(ctx: Context,
obj = parse_expr(node.value) obj = parse_expr(node.value)
if node.attr in obj.fields: if node.attr in obj.fields:
return obj.fields[node.attr] return obj.fields[node.attr]
raise CustomError(f'unknown field {node.attr} in {obj}') raise CustomError(f'unknown field {node.attr} in {obj}', node)
def parse_bool_ops(ctx: Context, def parse_bool_ops(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -117,7 +119,7 @@ def parse_bool_ops(ctx: Context,
right = parse_expr(ctx, sym_table, node.values[1]) right = parse_expr(ctx, sym_table, node.values[1])
b = ctx.types['bool'] b = ctx.types['bool']
if left != b or right != b: if left != b or right != b:
raise CustomError('operands of bool ops must be booleans') raise CustomError('operands of bool ops must be booleans', node)
return b return b
def parse_bin_ops(ctx: Context, def parse_bin_ops(ctx: Context,
@ -126,7 +128,10 @@ def parse_bin_ops(ctx: Context,
left = parse_expr(ctx, sym_table, node.left) left = parse_expr(ctx, sym_table, node.left)
right = parse_expr(ctx, sym_table, node.right) right = parse_expr(ctx, sym_table, node.right)
op = get_bin_ops(node.op) op = get_bin_ops(node.op)
try:
return resolve_call(left, op, [right], {}, ctx) return resolve_call(left, op, [right], {}, ctx)
except CustomError as e:
raise e.at(node)
def parse_unary_ops(ctx: Context, def parse_unary_ops(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -135,9 +140,12 @@ def parse_unary_ops(ctx: Context,
if isinstance(node.op, ast.Not): if isinstance(node.op, ast.Not):
b = ctx.types['bool'] b = ctx.types['bool']
if t != b: if t != b:
raise CustomError('operands of bool ops must be booleans') raise CustomError('operands of bool ops must be booleans', node)
return b return b
try:
return resolve_call(t, get_unary_op(node.op), [], {}, ctx) return resolve_call(t, get_unary_op(node.op), [], {}, ctx)
except CustomError as e:
raise e.at(node)
def parse_compare(ctx: Context, def parse_compare(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -147,17 +155,21 @@ def parse_compare(ctx: Context,
boolean = ctx.types['bool'] boolean = ctx.types['bool']
ops = [get_bin_ops(v) for v in node.ops] ops = [get_bin_ops(v) for v in node.ops]
for a, b, op in zip(items[:-1], items[1:], ops): for a, b, op in zip(items[:-1], items[1:], ops):
try:
result = resolve_call(a, op, [b], {}, ctx) result = resolve_call(a, op, [b], {}, ctx)
if result != boolean: if result != boolean:
raise CustomError( raise CustomError(
f'result of comparison must be bool instead of {result}') f'result of comparison must be bool instead of {result}')
except CustomError as e:
raise e.at(node)
return boolean return boolean
def parse_call(ctx: Context, def parse_call(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
node): node):
if len(node.keywords) > 0: if len(node.keywords) > 0:
raise CustomError('keyword arguments are not supported') raise CustomError('keyword arguments are not supported', node)
args = [parse_expr(ctx, sym_table, v) for v in node.args] args = [parse_expr(ctx, sym_table, v) for v in node.args]
obj = None obj = None
f = None f = None
@ -166,32 +178,35 @@ def parse_call(ctx: Context,
f = node.func.attr f = node.func.attr
elif isinstance(node.func, ast.Name): elif isinstance(node.func, ast.Name):
f = node.func.id f = node.func.id
try:
return resolve_call(obj, f, args, {}, ctx) return resolve_call(obj, f, args, {}, ctx)
except CustomError as e:
raise e.at(node)
def parse_subscript(ctx: Context, def parse_subscript(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
node): node):
value = parse_expr(ctx, sym_table, node.value) value = parse_expr(ctx, sym_table, node.value)
if not isinstance(value, ListType): if not isinstance(value, ListType):
raise CustomError(f'cannot take index of {value}') raise CustomError(f'cannot take index of {value}', node)
i32 = ctx.types['int32'] i32 = ctx.types['int32']
if isinstance(node.slice, ast.Slice): if isinstance(node.slice, ast.Slice):
if node.slice.lower is not None: if node.slice.lower is not None:
if parse_expr(ctx, sym_table, node.slice.lower) != i32: if parse_expr(ctx, sym_table, node.slice.lower) != i32:
raise CustomError(f'slice index must be int32') raise CustomError(f'slice index must be int32', node.slice.lower)
if node.slice.upper is not None: if node.slice.upper is not None:
if parse_expr(ctx, sym_table, node.slice.upper) != i32: if parse_expr(ctx, sym_table, node.slice.upper) != i32:
raise CustomError(f'slice index must be int32') raise CustomError(f'slice index must be int32', node.slice.upper)
if node.slice.step is not None: if node.slice.step is not None:
if parse_expr(ctx, sym_table, node.slice.step) != i32: if parse_expr(ctx, sym_table, node.slice.step) != i32:
raise CustomError(f'slice index must be int32') raise CustomError(f'slice index must be int32', node.slice.step)
return value return value
else: else:
s = parse_expr(ctx, sym_table, node.slice) s = parse_expr(ctx, sym_table, node.slice)
if s == i32: if s == i32:
return value.params[0] return value.params[0]
else: else:
raise CustomError(f'index of type {s} is not supported') raise CustomError(f'index of type {s} is not supported', node)
def parse_if_expr(ctx: Context, def parse_if_expr(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -199,11 +214,11 @@ def parse_if_expr(ctx: Context,
b = ctx.types['bool'] b = ctx.types['bool']
t = parse_expr(ctx, sym_table, node.test) t = parse_expr(ctx, sym_table, node.test)
if t != b: if t != b:
raise CustomError(f'type of conditional must be bool instead of {t}') raise CustomError(f'type of conditional must be bool instead of {t}', node)
ty1 = parse_expr(ctx, sym_table, node.body) ty1 = parse_expr(ctx, sym_table, node.body)
ty2 = parse_expr(ctx, sym_table, node.orelse) ty2 = parse_expr(ctx, sym_table, node.orelse)
if ty1 != ty2: if ty1 != ty2:
raise CustomError(f'divergent type for if expression: {ty1} != {ty2}') raise CustomError(f'divergent type for if expression: {ty1} != {ty2}', node)
return ty1 return ty1
def parse_simple_binding(name, ty): def parse_simple_binding(name, ty):
@ -222,7 +237,7 @@ def parse_simple_binding(name, ty):
expected = len(result) + len(binding) expected = len(result) + len(binding)
result |= parse_simple_binding(x, y) result |= parse_simple_binding(x, y)
if len(result) != expected: if len(result) != expected:
raise CustomError('variable name clash') raise CustomError('variable name clash', x)
return result return result
else: else:
raise CustomError(f'binding to {name} is not supported') raise CustomError(f'binding to {name} is not supported')
@ -231,16 +246,20 @@ def parse_list_comprehension(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
node): node):
if len(node.generators) != 1: if len(node.generators) != 1:
raise CustomError('list comprehension with more than 1 for loop is not supported') raise CustomError(
'list comprehension with more than 1 for loop is not supported', node)
if node.generators[0].is_async: if node.generators[0].is_async:
raise CustomError('async list comprehension is not supported') raise CustomError('async list comprehension is not supported', node)
ty = parse_expr(ctx, sym_table, node.generators[0].iter) ty = parse_expr(ctx, sym_table, node.generators[0].iter)
if not isinstance(ty, ListType): if not isinstance(ty, ListType):
raise CustomError(f'unable to iterate over {ty}') raise CustomError(f'unable to iterate over {ty}', node)
try:
sym_table2 = sym_table | parse_simple_binding(node.generators[0].target, ty.params[0]) sym_table2 = sym_table | parse_simple_binding(node.generators[0].target, ty.params[0])
except CustomError as e:
raise e.at(node)
b = ctx.types['bool'] b = ctx.types['bool']
for c in node.generators[0].ifs: for c in node.generators[0].ifs:
if parse_expr(ctx, sym_table2, c) != b: if parse_expr(ctx, sym_table2, c) != b:
raise CustomError(f'condition should be of boolean type') raise CustomError(f'condition should be of boolean type', c)
return ListType(parse_expr(ctx, sym_table2, node.elt)) return ListType(parse_expr(ctx, sym_table2, node.elt))

View File

@ -26,7 +26,7 @@ def parse_stmts(ctx: Context,
elif isinstance(node, ast.Break) or isinstance(node, ast.Continue): elif isinstance(node, ast.Break) or isinstance(node, ast.Continue):
continue continue
else: else:
raise CustomError(f'{node} is not supported yet') raise CustomError(f'{node} is not supported yet', node)
sym_table2 |= a sym_table2 |= a
used_sym_table2 |= b used_sym_table2 |= b
if returned: if returned:
@ -40,24 +40,24 @@ def get_target_type(ctx: Context,
if isinstance(target, ast.Subscript): if isinstance(target, ast.Subscript):
t = get_target_type(ctx, sym_table, used_sym_table, target.value) t = get_target_type(ctx, sym_table, used_sym_table, target.value)
if not isinstance(t, ListType): if not isinstance(t, ListType):
raise CustomError(f'cannot index through type {t}') raise CustomError(f'cannot index through type {t}', target)
if isinstance(target.slice, ast.Slice): if isinstance(target.slice, ast.Slice):
raise CustomError(f'assignment to slice is not supported') raise CustomError(f'assignment to slice is not supported', target)
i = parse_expr(ctx, sym_table, target.slice) i = parse_expr(ctx, sym_table, target.slice)
if i != ctx.types['int32']: if i != ctx.types['int32']:
raise CustomError(f'index must be int32') raise CustomError(f'index must be int32', target.slice)
return t.params[0] return t.params[0]
elif isinstance(target, ast.Attribute): elif isinstance(target, ast.Attribute):
t = get_target_type(ctx, sym_table, used_sym_table, target.value) t = get_target_type(ctx, sym_table, used_sym_table, target.value)
if target.attr not in t.fields: if target.attr not in t.fields:
raise CustomError(f'{t} has no field {target.attr}') raise CustomError(f'{t} has no field {target.attr}', target)
return t.fields[target.attr] return t.fields[target.attr]
elif isinstance(target, ast.Name): elif isinstance(target, ast.Name):
if target.id not in sym_table: if target.id not in sym_table:
raise CustomError(f'unbounded {target.id}') raise CustomError(f'unbounded {target.id}', target)
return sym_table[target.id] return sym_table[target.id]
else: else:
raise CustomError(f'assignment to {target} is not supported') raise CustomError(f'assignment to {target} is not supported', target)
def parse_stmt_binding(ctx: Context, def parse_stmt_binding(ctx: Context,
sym_table: dict[str, Type], sym_table: dict[str, Type],
@ -67,15 +67,17 @@ def parse_stmt_binding(ctx: Context,
if isinstance(target, ast.Name): if isinstance(target, ast.Name):
if target.id in used_sym_table: if target.id in used_sym_table:
if used_sym_table[target.id] != ty: if used_sym_table[target.id] != ty:
raise CustomError('inconsistent type') raise CustomError(f'inconsistent type, {target.id} was ' \
f'defined to be {used_sym_table[target.id]}, ' \
f'but is now {ty}', target)
if target.id == '_': if target.id == '_':
return {} return {}
return {target.id: ty} return {target.id: ty}
elif isinstance(target, ast.Tuple): elif isinstance(target, ast.Tuple):
if not isinstance(ty, TupleType): if not isinstance(ty, TupleType):
raise CustomError(f'cannot pattern match over {ty}') raise CustomError(f'cannot pattern match over {ty}', target)
if len(target.elts) != len(ty.params): if len(target.elts) != len(ty.params):
raise CustomError(f'pattern matching length mismatch') raise CustomError(f'pattern matching length mismatch', target)
result = {} result = {}
for x, y in zip(target.elts, ty.params): for x, y in zip(target.elts, ty.params):
new = parse_stmt_binding(ctx, sym_table, used_sym_table, x, y) new = parse_stmt_binding(ctx, sym_table, used_sym_table, x, y)
@ -83,12 +85,12 @@ def parse_stmt_binding(ctx: Context,
result |= new result |= new
used_sym_table |= new used_sym_table |= new
if len(result) != old_len + len(new): if len(result) != old_len + len(new):
raise CustomError(f'variable name clash') raise CustomError(f'variable name clash', target)
return result return result
else: else:
t = get_target_type(ctx, sym_table, used_sym_table, target) t = get_target_type(ctx, sym_table, used_sym_table, target)
if ty != t: if ty != t:
raise CustomError(f'inconsistent type') raise CustomError(f'type mismatch', target)
return {} return {}
def parse_assign(ctx: Context, def parse_assign(ctx: Context,
@ -113,7 +115,7 @@ def parse_if_stmt(ctx: Context,
node): node):
test = parse_expr(ctx, sym_table, node.test) test = parse_expr(ctx, sym_table, node.test)
if test != ctx.types['bool']: if test != ctx.types['bool']:
raise CustomError(f'condition must be bool instead of {test}') raise CustomError(f'condition must be bool instead of {test}', node.test)
a, b, r = parse_stmts(ctx, sym_table, used_sym_table, return_ty, node.body) a, b, r = parse_stmts(ctx, sym_table, used_sym_table, return_ty, node.body)
used_sym_table |= b used_sym_table |= b
a1, b1, r1 = parse_stmts(ctx, sym_table, used_sym_table, return_ty, node.orelse) a1, b1, r1 = parse_stmts(ctx, sym_table, used_sym_table, return_ty, node.orelse)
@ -127,12 +129,12 @@ def parse_for_stmt(ctx: Context,
node): node):
ty = parse_expr(ctx, sym_table, node.iter) ty = parse_expr(ctx, sym_table, node.iter)
if not isinstance(ty, ListType): if not isinstance(ty, ListType):
raise CustomError('only iteration over list is supported') raise CustomError('only iteration over list is supported', node.iter)
binding = parse_simple_binding(node.target, ty.params[0]) binding = parse_simple_binding(node.target, ty.params[0])
for key, value in binding.items(): for key, value in binding.items():
if key in used_sym_table: if key in used_sym_table:
if value != used_sym_table[key]: if value != used_sym_table[key]:
raise CustomError('inconsistent type') raise CustomError('inconsistent type', node)
# more sophisticated return analysis is needed... # more sophisticated return analysis is needed...
a, b, _ = parse_stmts(ctx, sym_table | binding, used_sym_table | binding, a, b, _ = parse_stmts(ctx, sym_table | binding, used_sym_table | binding,
return_ty, node.body) return_ty, node.body)
@ -148,7 +150,7 @@ def parse_while_stmt(ctx: Context,
node): node):
ty = parse_expr(ctx, sym_table, node.test) ty = parse_expr(ctx, sym_table, node.test)
if ty != ctx.types['bool']: if ty != ctx.types['bool']:
raise CustomError('condition must be bool') raise CustomError('condition must be bool', node.test)
# more sophisticated return analysis is needed... # more sophisticated return analysis is needed...
a, b, _ = parse_stmts(ctx, sym_table, used_sym_table, return_ty, node.body) a, b, _ = parse_stmts(ctx, sym_table, used_sym_table, return_ty, node.body)
a1, b1, _ = parse_stmts(ctx, sym_table, used_sym_table | b, a1, b1, _ = parse_stmts(ctx, sym_table, used_sym_table | b,
@ -163,10 +165,10 @@ def parse_return_stmt(ctx: Context,
node): node):
if return_ty is None: if return_ty is None:
if node.value is not None: if node.value is not None:
raise CustomError('no return value is allowed') raise CustomError('no return value is allowed', node)
return {}, {}, True return {}, {}, True
ty = parse_expr(ctx, sym_table, node.value) ty = parse_expr(ctx, sym_table, node.value)
if ty != return_ty: if ty != return_ty:
raise CustomError(f'expected returning {return_ty} but got {ty}') raise CustomError(f'expected returning {return_ty} but got {ty}', node)
return {}, {}, True return {}, {}, True