from __future__ import annotations

import contextlib

import difflib

import os

import pathlib

import re

import shutil

import sys

import time

from collections import defaultdict

from collections.abc import Callable, Iterable, Iterator

from re import Pattern

from typing import IO, Any

# Exporting Suite as alias to TestCase for backwards compatibility

# TODO: avoid aliasing - import and subclass TestCase directly

from unittest import TestCase

Suite = TestCase # re-exporting

import pytest

import mypy.api as api

import mypy.version

from mypy import defaults

from mypy.main import process_options

from mypy.options import Options

from mypy.test.config import test_data_prefix, test_temp_dir

from mypy.test.data import (

DataDrivenTestCase,

DeleteFile,

UpdateFile,

choose_file_encoding,

fix_cobertura_filename,

)

skip = pytest.mark.skip

# AssertStringArraysEqual displays special line alignment helper messages if

# the first different line has at least this many characters,

MIN_LINE_LENGTH_FOR_ALIGNMENT = 5

def run_mypy(args: list[str]) -> None:

__tracebackhide__ = True

# We must enable site packages even though they could cause problems,

# since stubs for typing_extensions live there.

outval, errval, status = api.run(args + ["--show-traceback", "--no-silence-site-packages"])

if status != 0:

sys.stdout.write(outval)

sys.stderr.write(errval)

pytest.fail(reason="Sample check failed", pytrace=False)

def diff_ranges(

left: list[str], right: list[str]

) -> tuple[list[tuple[int, int]], list[tuple[int, int]]]:

seq = difflib.SequenceMatcher(None, left, right)

# note last triple is a dummy, so don't need to worry

blocks = seq.get_matching_blocks()

i = 0

j = 0

left_ranges = []

right_ranges = []

for block in blocks:

# mismatched range

left_ranges.append((i, block.a))

right_ranges.append((j, block.b))

i = block.a + block.size

j = block.b + block.size

# matched range

left_ranges.append((block.a, i))

right_ranges.append((block.b, j))

return left_ranges, right_ranges

def render_diff_range(

ranges: list[tuple[int, int]],

content: list[str],

*,

colour: str | None = None,

output: IO[str] = sys.stderr,

indent: int = 2,

) -> None:

for i, line_range in enumerate(ranges):

is_matching = i % 2 == 1

lines = content[line_range[0] : line_range[1]]

for j, line in enumerate(lines):

if (

is_matching

# elide the middle of matching blocks

and j >= 3

and j < len(lines) - 3

):

if j == 3:

output.write(" " * indent + "...\n")

continue

if not is_matching and colour:

output.write(colour)

output.write(" " * indent + line)

if not is_matching:

if colour:

output.write("\033[0m")

output.write(" (diff)")

output.write("\n")

def dump_original_errors(errors: list[str]) -> None:

sys.stderr.write("Original errors:\n")

for err in errors:

sys.stderr.write(err + "\n")

def module_order(errors: list[str]) -> list[str]:

result = []

seen = set()

mods = []

for e in errors:

if ":" not in e:

dump_original_errors(errors)

pytest.fail(f"Only module scoped errors are supported, got {e}")

mod, _ = e.split(":", maxsplit=1)

mods.append(mod)

for i, mod in enumerate(mods):

if i > 0:

if mod != mods[i - 1] and mod in seen:

dump_original_errors(errors)

pytest.fail(f"Each module must form a single block, {mod} appears split")

if mod not in seen:

result.append(mod)

seen.add(mod)

return result

def match_module_order(actual: list[str], expected_order: list[str]) -> list[str]:

actual_by_mod = defaultdict(list)

actual_order = module_order(actual)

if set(actual_order) != set(expected_order):

# Different files, give up and show actual errors.

return actual

for a in actual:

mod, _ = a.split(":", maxsplit=1)

actual_by_mod[mod].append(a)

result = []

for mod in expected_order:

result.extend(actual_by_mod[mod])

return result

def assert_string_arrays_equal(

expected: list[str],

actual: list[str],

msg: str,

*,

traceback: bool = False,

ignore_modules_order: bool = False,

) -> None:

"""Assert that two string arrays are equal.

Display any differences in a human-readable form.

"""

actual = clean_up(actual)

if ignore_modules_order:

expected_module_order = module_order(expected)

actual = match_module_order(actual, expected_module_order)

if expected != actual:

expected_ranges, actual_ranges = diff_ranges(expected, actual)

sys.stderr.write("Expected:\n")

red = "\033[31m" if sys.platform != "win32" else None

render_diff_range(expected_ranges, expected, colour=red)

sys.stderr.write("Actual:\n")

green = "\033[32m" if sys.platform != "win32" else None

render_diff_range(actual_ranges, actual, colour=green)

sys.stderr.write("\n")

first_diff = next(

(i for i, (a, b) in enumerate(zip(expected, actual)) if a != b),

max(len(expected), len(actual)),

)

if 0 <= first_diff < len(actual) and (

len(expected[first_diff]) >= MIN_LINE_LENGTH_FOR_ALIGNMENT

or len(actual[first_diff]) >= MIN_LINE_LENGTH_FOR_ALIGNMENT

):

# Display message that helps visualize the differences between two

# long lines.

show_align_message(expected[first_diff], actual[first_diff])

sys.stderr.write(

"Update the test output using --update-data "

"(implies -n0; you can additionally use the -k selector to update only specific tests)\n"

)

pytest.fail(msg, pytrace=traceback)

def assert_module_equivalence(name: str, expected: Iterable[str], actual: Iterable[str]) -> None:

expected_normalized = sorted(expected)

actual_normalized = sorted(set(actual).difference({"__main__"}))

assert_string_arrays_equal(

expected_normalized,

actual_normalized,

('Actual modules ({}) do not match expected modules ({}) for "[{} ...]"').format(

", ".join(actual_normalized), ", ".join(expected_normalized), name

),

)

def assert_target_equivalence(name: str, expected: list[str], actual: list[str]) -> None:

"""Compare actual and expected targets (order sensitive)."""

assert_string_arrays_equal(

expected,

actual,

('Actual targets ({}) do not match expected targets ({}) for "[{} ...]"').format(

", ".join(actual), ", ".join(expected), name

),

)

def show_align_message(s1: str, s2: str) -> None:

"""Align s1 and s2 so that the their first difference is highlighted.

For example, if s1 is 'foobar' and s2 is 'fobar', display the

following lines:

E: foobar

A: fobar

^

If s1 and s2 are long, only display a fragment of the strings around the

first difference. If s1 is very short, do nothing.

"""

# Seeing what went wrong is trivial even without alignment if the expected

# string is very short. In this case do nothing to simplify output.

if len(s1) < 4:

return

maxw = 72 # Maximum number of characters shown

sys.stderr.write("Alignment of first line difference:\n")

trunc = False

while s1[:30] == s2[:30]:

s1 = s1[10:]

s2 = s2[10:]

trunc = True

if trunc:

s1 = "..." + s1

s2 = "..." + s2

max_len = max(len(s1), len(s2))

extra = ""

if max_len > maxw:

extra = "..."

# Write a chunk of both lines, aligned.

sys.stderr.write(f" E: {s1[:maxw]}{extra}\n")

sys.stderr.write(f" A: {s2[:maxw]}{extra}\n")

# Write an indicator character under the different columns.

sys.stderr.write(" ")

for j in range(min(maxw, max(len(s1), len(s2)))):

if s1[j : j + 1] != s2[j : j + 1]:

sys.stderr.write("^") # Difference

break

else:

sys.stderr.write(" ") # Equal

sys.stderr.write("\n")

def clean_up(a: list[str]) -> list[str]:

"""Remove common directory prefix from all strings in a.

This uses a naive string replace; it seems to work well enough. Also

remove trailing carriage returns.

"""

res = []

pwd = os.getcwd()

driver = pwd + "/driver.py"

for s in a:

prefix = os.sep

ss = s

for p in prefix, prefix.replace(os.sep, "/"):

if p != "/" and p != "//" and p != "\\" and p != "\\\\":

ss = ss.replace(p, "")

# Replace memory address with zeros

if "at 0x" in ss:

ss = re.sub(r"(at 0x)\w+>", r"\g<1>000000000000>", ss)

# Ignore spaces at end of line.

ss = re.sub(" +$", "", ss)

# Remove pwd from driver.py's path

ss = ss.replace(driver, "driver.py")

res.append(re.sub("\\r$", "", ss))

return res

@contextlib.contextmanager

def local_sys_path_set() -> Iterator[None]:

"""Temporary insert current directory into sys.path.

This can be used by test cases that do runtime imports, for example

by the stubgen tests.

"""

old_sys_path = sys.path.copy()

if not ("" in sys.path or "." in sys.path):

sys.path.insert(0, "")

try:

yield

finally:

sys.path = old_sys_path

def testfile_pyversion(path: str) -> tuple[int, int]:

if m := re.search(r"python3([0-9]+)\.test$", path):

# For older unsupported version like python38,

# default to that earliest supported version.

return max((3, int(m.group(1))), defaults.PYTHON3_VERSION_MIN)

else:

return defaults.PYTHON3_VERSION_MIN

def normalize_error_messages(messages: list[str]) -> list[str]:

"""Translate an array of error messages to use / as path separator."""

a = []

for m in messages:

a.append(m.replace(os.sep, "/"))

return a

def retry_on_error(func: Callable[[], Any], max_wait: float = 1.0) -> None:

"""Retry callback with exponential backoff when it raises OSError.

If the function still generates an error after max_wait seconds, propagate

the exception.

This can be effective against random file system operation failures on

Windows.

"""

t0 = time.time()

wait_time = 0.01

while True:

try:

func()

return

except OSError:

wait_time = min(wait_time * 2, t0 + max_wait - time.time())

if wait_time <= 0.01:

# Done enough waiting, the error seems persistent.

raise

time.sleep(wait_time)

def good_repr(obj: object) -> str:

if isinstance(obj, str):

if obj.count("\n") > 1:

bits = ["'''\\"]

for line in obj.split("\n"):

# force repr to use ' not ", then cut it off

bits.append(repr('"' + line)[2:-1])

bits[-1] += "'''"

return "\n".join(bits)

return repr(obj)

def assert_equal(a: object, b: object, fmt: str = "{} != {}") -> None:

__tracebackhide__ = True

if a != b:

raise AssertionError(fmt.format(good_repr(a), good_repr(b)))

def typename(t: type) -> str:

if "." in str(t):

return str(t).split(".")[-1].rstrip("'>")

else:

return str(t)[8:-2]

def assert_type(typ: type, value: object) -> None:

__tracebackhide__ = True

if type(value) != typ:

raise AssertionError(f"Invalid type {typename(type(value))}, expected {typename(typ)}")

def parse_options(

program_text: str, testcase: DataDrivenTestCase, incremental_step: int

) -> Options:

"""Parse comments like '# flags: --foo' in a test case."""

options = Options()

flags = re.search("# flags: (.*)$", program_text, flags=re.MULTILINE)

if incremental_step > 1:

flags2 = re.search(f"# flags{incremental_step}: (.*)$", program_text, flags=re.MULTILINE)

if flags2:

flags = flags2

if flags:

flag_list = flags.group(1).split()

flag_list.append("--no-site-packages") # the tests shouldn't need an installed Python

targets, options = process_options(flag_list, require_targets=False)

if targets:

# TODO: support specifying targets via the flags pragma

raise RuntimeError("Specifying targets via the flags pragma is not supported.")

if "--show-error-codes" not in flag_list:

options.hide_error_codes = True

else:

flag_list = []

options = Options()

options.error_summary = False

options.hide_error_codes = True

# Allow custom python version to override testfile_pyversion.

if all(flag.split("=")[0] != "--python-version" for flag in flag_list):

options.python_version = testfile_pyversion(testcase.file)

if testcase.config.getoption("--mypy-verbose"):

options.verbosity = testcase.config.getoption("--mypy-verbose")

if testcase.config.getoption("--mypy-num-workers"):

options.num_workers = testcase.config.getoption("--mypy-num-workers")

return options

def split_lines(*streams: bytes) -> list[str]:

"""Returns a single list of string lines from the byte streams in args."""

return [s for stream in streams for s in stream.decode("utf8").splitlines()]

def write_and_fudge_mtime(content: str, target_path: str) -> None:

# In some systems, mtime has a resolution of 1 second which can

# cause annoying-to-debug issues when a file has the same size

# after a change. We manually set the mtime to circumvent this.

# Note that we increment the old file's mtime, which guarantees a

# different value, rather than incrementing the mtime after the

# copy, which could leave the mtime unchanged if the old file had

# a similarly fudged mtime.

new_time = None

if os.path.isfile(target_path):

new_time = os.stat(target_path).st_mtime + 1

dir = os.path.dirname(target_path)

os.makedirs(dir, exist_ok=True)

with open(target_path, "w", encoding=choose_file_encoding(target_path)) as target:

target.write(content)

if new_time:

os.utime(target_path, times=(new_time, new_time))

def perform_file_operations(operations: list[UpdateFile | DeleteFile]) -> None:

for op in operations:

if isinstance(op, UpdateFile):

# Modify/create file

write_and_fudge_mtime(op.content, op.target_path)

else:

# Delete file/directory

if os.path.isdir(op.path):

# Sanity check to avoid unexpected deletions

assert op.path.startswith("tmp")

shutil.rmtree(op.path)

else:

# Use retries to work around potential flakiness on Windows (AppVeyor).

path = op.path

retry_on_error(lambda: os.remove(path))

def check_test_output_files(

testcase: DataDrivenTestCase, step: int, strip_prefix: str = ""

) -> None:

for path, expected_content in testcase.output_files:

path = path.removeprefix(strip_prefix)

if not os.path.exists(path):

raise AssertionError(

"Expected file {} was not produced by test case{}".format(

path, " on step %d" % step if testcase.output2 else ""

)

)

with open(path, encoding="utf8") as output_file:

actual_output_content = output_file.read()

if isinstance(expected_content, Pattern):

if expected_content.fullmatch(actual_output_content) is not None:

continue

raise AssertionError(

"Output file {} did not match its expected output pattern\n---\n{}\n---".format(

path, actual_output_content

)

)

normalized_output = normalize_file_output(

actual_output_content.splitlines(), os.path.abspath(test_temp_dir)

)

# We always normalize things like timestamp, but only handle operating-system

# specific things if requested.

if testcase.normalize_output:

if testcase.suite.native_sep and os.path.sep == "\\":

normalized_output = [fix_cobertura_filename(line) for line in normalized_output]

normalized_output = normalize_error_messages(normalized_output)

if os.path.basename(testcase.file) == "reports.test":

normalized_output = normalize_report_meta(normalized_output)

assert_string_arrays_equal(

expected_content.splitlines(),

normalized_output,

"Output file {} did not match its expected output{}".format(

path, " on step %d" % step if testcase.output2 else ""

),

)

def normalize_file_output(content: list[str], current_abs_path: str) -> list[str]:

"""Normalize file output for comparison."""

timestamp_regex = re.compile(r"\d{10}")

result = [x.replace(current_abs_path, "$PWD") for x in content]

version = mypy.version.__version__

result = [re.sub(r"\b" + re.escape(version) + r"\b", "$VERSION", x) for x in result]

# We generate a new mypy.version when building mypy wheels that

# lacks base_version, so handle that case.

base_version = getattr(mypy.version, "base_version", version)

result = [re.sub(r"\b" + re.escape(base_version) + r"\b", "$VERSION", x) for x in result]

result = [timestamp_regex.sub("$TIMESTAMP", x) for x in result]

return result

def normalize_report_meta(content: list[str]) -> list[str]:

# libxml 2.15 and newer emits the "modern" version of this <meta> element.

# Normalize the old style to look the same.

html_meta = '<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">'

return ['<meta charset="UTF-8">' if x == html_meta else x for x in content]

def find_test_files(pattern: str, exclude: list[str] | None = None) -> list[str]:

return [

path.name

for path in (pathlib.Path(test_data_prefix).rglob(pattern))

if path.name not in (exclude or [])

]

def remove_typevar_ids(a: list[str]) -> list[str]:

return [re.sub(r"`-?\d+", "", line) for line in a]