Big diff: use lower-case list and dict

[Akuli]

Created with quick and dirty scripts very similarly to #5872

main script

import sys
import pathlib
import re


def parse_list(string):
    assert string.startswith("[")
    depth = 1
    parts = []
    part_start = i = 1
    while depth > 0:
        if depth == 1 and string[i:].startswith(","):
            parts.append(string[part_start:i].strip())
            part_start = i + 1
        depth += string[i] == "["
        depth -= string[i] == "]"
        i += 1
    if string[part_start : i - 1].strip():
        parts.append(string[part_start : i - 1].strip())
    assert parts, string[:100]
    assert "" not in parts
    return (parts, i)


assert parse_list("[x] bruh") == (["x"], 3)
assert parse_list("[x, y[z]] bruh") == (["x", "y[z]"], 9)

for path in list(map(pathlib.Path, sys.argv[1:])) or (
    list(pathlib.Path("stdlib").rglob("*.pyi")) + list(pathlib.Path("stubs").rglob("*.pyi"))
):
    if "@python2" in path.parts:
        # no real reason to skip, I just don't like python 2
        continue

    print(path)
    content = path.read_text()

    while True:
        try:
            match = next(
                m for m in re.finditer(
                    r"^.*?\b(List|Dict|typing\.List|typing\.Dict)\[", content, flags=re.MULTILINE
                )
                # skip type aliases and base classes
                if not re.match("[A-Za-z_][A-Za-z0-9_]* = |class ", m.group(0).lstrip(" "))
            )
        except StopIteration:
            break

        parts, list_length = parse_list(content[match.end() - 1 :])
        start = match.start(1)
        end = match.end() - 1 + list_length
        if match.group(1).endswith("List"):
            [p] = parts
            content = content[:start] + f"list[{p}]" + content[end:]
        else:
            [k, v] = parts
            content = content[:start] + f"dict[{k}, {v}]" + content[end:]

    path.write_text(content)

import cleaner (can do weird things, needs manual review)

import pathlib
import re


for path in list(pathlib.Path("stdlib").rglob("*.pyi")) + list(pathlib.Path("stubs").rglob("*.pyi")):
    if "@python2" in path.parts:
        # no real reason to skip, I just don't like python 2
        continue

    print(path)
    content = path.read_text()

    for old in ["List", "Dict"]:
        if old + "[" not in content:
            content = re.sub(rf", \b{old}\b", "", content)
            content = re.sub(rf"\b{old}\b,", "", content)
            content = content.replace(r"from typing import " + old + "\n", "")
    path.write_text(content)

reverting script

#!/bin/bash
set -ex
[ -f $1 ]
git diff --exit-code
git diff HEAD --exit-code
git show origin/master:$1 > $1
git add $1
git commit -m "Revert $1 to origin/master"

Files that need manual work (checked = separate PR created):

• stdlib/imaplib.pyi
• stdlib/multiprocessing/managers.pyi
• stdlib/nntplib.pyi
• stdlib/os/__init__.pyi
• stdlib/plistlib.pyi
• stdlib/poplib.pyi
• stdlib/pydoc.pyi
• stdlib/tarfile.pyi
• stdlib/xmlrpc/server.pyi
• stubs/decorator/decorator.pyi
• stubs/docutils/docutils/parsers/rst/roles.pyi (nothing to do)
• stubs/frozendict/frozendict.pyi
• stubs/Jinja2/jinja2/nodes.pyi
• stubs/Pygments/pygments/token.pyi
• stubs/requests/requests/models.pyi
• stubs/Werkzeug/werkzeug/http.pyi

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

[srittau]

LGTM