Revert "Move `glob_to_regex` and `re_word_boundary` to `matrix-python-common` (#11505) (#11527)

This reverts commit a77c36989785c0d5565ab9a1169f4f88e512ce8a.

1 files changed, 58 insertions, 1 deletions

diff --git a/synapse/util/__init__.py b/synapse/util/__init__.py
index f157132210..95f23e27b6 100644
--- a/synapse/util/__init__.py
+++ b/synapse/util/__init__.py
@@ -14,8 +14,9 @@
 
 import json
 import logging
+import re
 import typing
-from typing import Any, Callable, Dict, Generator, Optional
+from typing import Any, Callable, Dict, Generator, Optional, Pattern
 
 import attr
 from frozendict import frozendict
@@ -34,6 +35,9 @@ if typing.TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
+_WILDCARD_RUN = re.compile(r"([\?\*]+)")
+
+
 def _reject_invalid_json(val: Any) -> None:
     """Do not allow Infinity, -Infinity, or NaN values in JSON."""
     raise ValueError("Invalid JSON value: '%s'" % val)
@@ -181,3 +185,56 @@ def log_failure(
     if not consumeErrors:
         return failure
     return None
+
+
+def glob_to_regex(glob: str, word_boundary: bool = False) -> Pattern:
+    """Converts a glob to a compiled regex object.
+
+    Args:
+        glob: pattern to match
+        word_boundary: If True, the pattern will be allowed to match at word boundaries
+           anywhere in the string. Otherwise, the pattern is anchored at the start and
+           end of the string.
+
+    Returns:
+        compiled regex pattern
+    """
+
+    # Patterns with wildcards must be simplified to avoid performance cliffs
+    # - The glob `?**?**?` is equivalent to the glob `???*`
+    # - The glob `???*` is equivalent to the regex `.{3,}`
+    chunks = []
+    for chunk in _WILDCARD_RUN.split(glob):
+        # No wildcards? re.escape()
+        if not _WILDCARD_RUN.match(chunk):
+            chunks.append(re.escape(chunk))
+            continue
+
+        # Wildcards? Simplify.
+        qmarks = chunk.count("?")
+        if "*" in chunk:
+            chunks.append(".{%d,}" % qmarks)
+        else:
+            chunks.append(".{%d}" % qmarks)
+
+    res = "".join(chunks)
+
+    if word_boundary:
+        res = re_word_boundary(res)
+    else:
+        # \A anchors at start of string, \Z at end of string
+        res = r"\A" + res + r"\Z"
+
+    return re.compile(res, re.IGNORECASE)
+
+
+def re_word_boundary(r: str) -> str:
+    """
+    Adds word boundary characters to the start and end of an
+    expression to require that the match occur as a whole word,
+    but do so respecting the fact that strings starting or ending
+    with non-word characters will change word boundaries.
+    """
+    # we can't use \b as it chokes on unicode. however \W seems to be okay
+    # as shorthand for [^0-9A-Za-z_].
+    return r"(^|\W)%s(\W|$)" % (r,)