[lomeliDev]

Addresses issue #5 (#5); the issue stays open until a maintainer decides to merge and close it.

Root cause

The modern algorithm validates its caller-supplied arguments up front in
_modern_slugify (max_length, separator, text, replacement_stage,
backend), but the neighboring regex_pattern argument was skipped. It is
consumed much later by re.sub, so a bad value fails cryptically deep in the
pipeline — after entity decoding, transliteration and pre-replacements have
already run — with a message that names neither the argument nor the accepted
types:

>>> slugify('abc', algorithm='modern', regex_pattern=123)
TypeError: first argument must be string or compiled pattern

Fix (minimal)

Add the same guard the sibling arguments already have, right after the
separator check in slugify/slugify.py:

if regex_pattern is not None and not isinstance(regex_pattern, (str, re.Pattern)):
    raise TypeError(f"regex_pattern must be str, re.Pattern or None, not {type(regex_pattern).__name__}")

None and empty strings keep their documented behavior (empty string falls
back to the default pattern). Legacy (slugify/_legacy.py) is frozen and
untouched.

Lesson propagation

The issue is itself the propagation of the PR un33k#196 "validate up front" lesson
to the one neighboring argument that was missed. Scope is intentionally limited
to regex_pattern, as requested; other caller-supplied arguments that the
modern pipeline also consumes late (stopwords, replacements) are out of
scope for this issue and were left unchanged.

Files

• slugify/slugify.py
• tests/test_modern_unicode.py (new)

Evidencia

Vista roja (fix removed, test kept)

$ find . -name '__pycache__' -type d -exec rm -rf {} + ; find . -name '*.pyc' -delete
$ python3 -m pytest -p no:cacheprovider tests/test_modern_unicode.py -v
...
tests/test_modern_unicode.py::test_regex_pattern_type_is_validated_up_front FAILED
...
E       AssertionError: Regex pattern did not match.
E         Expected regex: 'regex_pattern'
E         Actual message: 'first argument must be string or compiled pattern'
1 failed, 1 passed

Vista verde (fix restored, test kept)

$ python3 -m pytest -p no:cacheprovider tests/test_modern_unicode.py -v
...
tests/test_modern_unicode.py::test_regex_pattern_type_is_validated_up_front PASSED
tests/test_modern_unicode.py::test_regex_pattern_accepts_str_and_compiled_pattern PASSED
2 passed

Suite completa (fix in place)

$ python3 -m pytest -p no:cacheprovider tests -q
20 failed, 114 passed, 8 skipped, 127 subtests passed

The 20 failures are pre-existing and unrelated to this change: every one is a
ModuleNotFoundError: No module named 'text_unidecode' (plus one CLI subprocess
failure that is the same missing dependency surfaced via python -m slugify).
text-unidecode is the mandatory base dependency but is not installed in this
sandbox; pip install is blocked by the externally-managed (PEP 668)
environment, and installing/downloading tooling is out of scope for this
worker. The 2 new tests in this PR both pass.

Quality gate (CI)

Reproduced against the taller-ci workflow (.github/workflows/taller-ci.yml),
which runs flake8, pycodestyle and pytest on a matrix of Python 3.11 and
3.12:

• pytest: reproduced on the installed Python 3.11.2 (see above).
• flake8 / pycodestyle: not reproduced — neither binary is installed in
  the sandbox and installing them is disallowed for this worker. The diff adds
  only simple, existing-style Python (no trailing whitespace, no new imports
  beyond the already-used re in slugify.py; the new test file imports only
  re, pytest and slugify). No risky modern syntax was introduced.
• Python 3.12: not installed here; not reproduced (matches the 3.11 run for a
  syntax-neutral change).

Diff scope

The diff is confined to slugify/slugify.py and the new
tests/test_modern_unicode.py; no lockfiles, generated artifacts, or unrelated
files. (The issue suggested tests/test_modern_unicode.py, which did not yet
exist, so it is created here.)

🚀 Generated with Dojo ⛩️

[lomeliDev]

Revisión adversarial independiente — APROBABLE

Revisado por quien no escribió el código, en worktree propio (origin/pr/8), import resuelto al worktree (slugify/__init__.py del worktree, no el editable del clon principal), Python 3.11.2.

Puerta fiel al CI (taller-ci.yml). Reproduje los tres pasos exactos del workflow:

$ flake8 --ignore=E501,F403,F401,E241,E225,E128 --exclude=legacy_reference.py slugify setup.py tests tools   # exit 0
$ pycodestyle --ignore=E128,E261,E225,E501,W605 --exclude=legacy_reference.py slugify tests setup.py tools   # exit 0
$ python -m pytest -p no:cacheprovider tests -q        # 135 passed, 5 skipped, 136 subtests passed

(La base del fork era 133 passed; +2 por los tests nuevos, sin regresiones.) CI del fork verde en 3.11 y 3.12 (gh pr checks).

Vista roja/verde. Revertí slugify/slugify.py a origin/master dejando el test:

# SIN el guard:
FAILED test_regex_pattern_type_is_validated_up_front - AssertionError: Regex pattern did not match ('first argument must be string or compiled pattern')
test_regex_pattern_accepts_str_and_compiled_pattern PASSED
# CON el guard:
2 passed

Efectos secundarios / alcance: el guard es regex_pattern is not None and not isinstance(regex_pattern, (str, re.Pattern)). None y cadena vacía conservan su comportamiento documentado: '' pasa el guard y luego regex_pattern or (DEFAULT) recae en el patrón por defecto (cubierto por el 2º test). Sólo toca slugify/slugify.py y el test nuevo; _legacy.py intacto (la paridad diferencial del test_release pasa). Conforme a Archivos: de la issue #5.

Sin reparos.

— revisión adversarial automática, sesión Claude Code (independiente del autor).