0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-27 03:03:01 +00:00
yt-dlp/devscripts/tomlparse.py

189 lines
4.7 KiB
Python
Executable file

#!/usr/bin/env python3
"""
Simple parser for spec compliant toml files
A simple toml parser for files that comply with the spec.
Should only be used to parse `pyproject.toml` for `install_deps.py`.
IMPORTANT: INVALID FILES OR MULTILINE STRINGS ARE NOT SUPPORTED!
"""
from __future__ import annotations
import datetime as dt
import json
import re
WS = r'(?:[\ \t]*)'
STRING_RE = re.compile(r'"(?:\\.|[^\\"\n])*"|\'[^\'\n]*\'')
SINGLE_KEY_RE = re.compile(rf'{STRING_RE.pattern}|[A-Za-z0-9_-]+')
KEY_RE = re.compile(rf'{WS}(?:{SINGLE_KEY_RE.pattern}){WS}(?:\.{WS}(?:{SINGLE_KEY_RE.pattern}){WS})*')
EQUALS_RE = re.compile(rf'={WS}')
WS_RE = re.compile(WS)
_SUBTABLE = rf'(?P<subtable>^\[(?P<is_list>\[)?(?P<path>{KEY_RE.pattern})\]\]?)'
EXPRESSION_RE = re.compile(rf'^(?:{_SUBTABLE}|{KEY_RE.pattern}=)', re.MULTILINE)
LIST_WS_RE = re.compile(rf'{WS}((#[^\n]*)?\n{WS})*')
LEFTOVER_VALUE_RE = re.compile(r'[^,}\]\t\n#]+')
def parse_key(value: str):
for match in SINGLE_KEY_RE.finditer(value):
if match[0][0] == '"':
yield json.loads(match[0])
elif match[0][0] == '\'':
yield match[0][1:-1]
else:
yield match[0]
def get_target(root: dict, paths: list[str], is_list=False):
target = root
for index, key in enumerate(paths, 1):
use_list = is_list and index == len(paths)
result = target.get(key)
if result is None:
result = [] if use_list else {}
target[key] = result
if isinstance(result, dict):
target = result
elif use_list:
target = {}
result.append(target)
else:
target = result[-1]
assert isinstance(target, dict)
return target
def parse_enclosed(data: str, index: int, end: str, ws_re: re.Pattern):
index += 1
if match := ws_re.match(data, index):
index = match.end()
while data[index] != end:
index = yield True, index
if match := ws_re.match(data, index):
index = match.end()
if data[index] == ',':
index += 1
if match := ws_re.match(data, index):
index = match.end()
assert data[index] == end
yield False, index + 1
def parse_value(data: str, index: int):
if data[index] == '[':
result = []
indices = parse_enclosed(data, index, ']', LIST_WS_RE)
valid, index = next(indices)
while valid:
index, value = parse_value(data, index)
result.append(value)
valid, index = indices.send(index)
return index, result
if data[index] == '{':
result = {}
indices = parse_enclosed(data, index, '}', WS_RE)
valid, index = next(indices)
while valid:
valid, index = indices.send(parse_kv_pair(data, index, result))
return index, result
if match := STRING_RE.match(data, index):
return match.end(), json.loads(match[0]) if match[0][0] == '"' else match[0][1:-1]
match = LEFTOVER_VALUE_RE.match(data, index)
assert match
value = match[0].strip()
for func in [
int,
float,
dt.time.fromisoformat,
dt.date.fromisoformat,
dt.datetime.fromisoformat,
{'true': True, 'false': False}.get,
]:
try:
value = func(value)
break
except Exception:
pass
return match.end(), value
def parse_kv_pair(data: str, index: int, target: dict):
match = KEY_RE.match(data, index)
if not match:
return None
*keys, key = parse_key(match[0])
match = EQUALS_RE.match(data, match.end())
assert match
index = match.end()
index, value = parse_value(data, index)
get_target(target, keys)[key] = value
return index
def parse_toml(data: str):
root = {}
target = root
index = 0
while True:
match = EXPRESSION_RE.search(data, index)
if not match:
break
if match.group('subtable'):
index = match.end()
path, is_list = match.group('path', 'is_list')
target = get_target(root, list(parse_key(path)), bool(is_list))
continue
index = parse_kv_pair(data, match.start(), target)
assert index is not None
return root
def main():
import argparse
from pathlib import Path
parser = argparse.ArgumentParser()
parser.add_argument('infile', type=Path, help='The TOML file to read as input')
args = parser.parse_args()
with args.infile.open('r', encoding='utf-8') as file:
data = file.read()
def default(obj):
if isinstance(obj, (dt.date, dt.time, dt.datetime)):
return obj.isoformat()
print(json.dumps(parse_toml(data), default=default))
if __name__ == '__main__':
main()