Fix: use a PostgreSQL-compatible regex for scheduled workflow filename filters

This commit is contained in:
Trenton H committed 2026-11-15 13:26:42 -08:00
1 parent 4610171a94
commit 8f029f30a3
1 file changed
+45 -3
+45 -3
View File
@@ -3,7 +3,6 @@ from __future__ import annotations
import logging
import re
from fnmatch import fnmatch
from fnmatch import translate as fnmatch_translate
from typing import TYPE_CHECKING
from rest_framework import serializers
@@ -548,6 +547,48 @@ def existing_document_matches_workflow(
return (True, None)
def _glob_to_portable_regex(pattern: str) -> str:
"""
Translate an fnmatch-style glob into an anchored regex which is accepted by
Python, PostgreSQL (ARE), and MariaDB (PCRE).
fnmatch.translate() can't be used, as its output (`(?s:...)\\Z`) uses syntax
PostgreSQL rejects. Supports `*`, `?`, `[seq]` and `[!seq]`.
"""
out: list[str] = []
i = 0
n = len(pattern)
while i < n:
c = pattern[i]
i += 1
if c == "*":
out.append(".*")
elif c == "?":
out.append(".")
elif c == "[":
j = i
if j < n and pattern[j] == "!":
j += 1
if j < n and pattern[j] == "]":
j += 1
while j < n and pattern[j] != "]":
j += 1
if j >= n:
# No closing bracket, treat as a literal
out.append(r"\[")
else:
body = pattern[i:j]
i = j + 1
negate = body.startswith("!")
if negate:
body = body[1:]
body = "".join(f"\\{ch}" if ch in "\\[^" else ch for ch in body)
out.append(f"[{'^' if negate else ''}{body}]")
else:
out.append(re.escape(c))
return f"^{''.join(out)}$"
def prefilter_documents_by_workflowtrigger(
documents: QuerySet[Document],
trigger: WorkflowTrigger,
@@ -628,8 +669,9 @@ def prefilter_documents_by_workflowtrigger(
documents = documents.annotate(**annotations).filter(custom_field_q)
if trigger.filter_filename:
regex = fnmatch_translate(trigger.filter_filename).lstrip("^").rstrip("$")
documents = documents.filter(original_filename__iregex=regex)
documents = documents.filter(
original_filename__iregex=_glob_to_portable_regex(trigger.filter_filename),
)
return documents