from __future__ import annotations

import re

# Only http(s) is linkified; everything else stays inert plain text.
_URL_PATTERN = re.compile(r'https?://[^\s<>"\'`]+', re.IGNORECASE)

# Prose punctuation that is almost never part of the URL itself.
_TRAILING_PUNCTUATION = '.,;:!?\'"'
_BRACKET_PAIRS = {')': '(', ']': '[', '}': '{'}

def _trim_trailing(url: str) -> str:
  """Give punctuation that belongs to the surrounding sentence back to the text."""
  while url:
    last = url[-1]
    if last in _TRAILING_PUNCTUATION:
      url = url[:-1]
      continue
    opening = _BRACKET_PAIRS.get(last)
    if opening is not None and url.count(opening) < url.count(last):
      url = url[:-1]
      continue
    break
  return url

def tokenize(text: str) -> list[tuple[str, str]]:
  """Split text into ('text', value) and ('url', value) tokens, order preserved."""
  tokens: list[tuple[str, str]] = []
  cursor = 0

  for match in _URL_PATTERN.finditer(text):
    url = _trim_trailing(match.group(0))
    if not url:
      continue

    start = match.start()
    end = start + len(url)
    if start > cursor:
      tokens.append(('text', text[cursor:start]))
    tokens.append(('url', url))
    cursor = end

  if cursor < len(text):
    tokens.append(('text', text[cursor:]))
  return tokens