fix(classify): do not crash on invalid user-supplied regex

[fanxing11]

Small drive-by fix for ActivityWatch/activitywatch#1340.

What happened

Rule.__init__ (in aw_transform/classify.py) compiles user-supplied
regex patterns eagerly with re.compile(...). If the pattern is
invalid, re.error is raised straight out of __init__, bubbles up
through the query engine, and aw-server responds with 500 Internal
Server Error for the whole query. The web UI then shows nothing for
any categorize()/tag() chart, not just the rule that was broken.

The reproducer in the issue is Notepad++ (the user probably wanted
Notepad\+\+). On Python <= 3.10 this raises re.error: multiple repeat at position 8. On 3.11+ ++ happens to be a valid possessive
quantifier so that particular pattern no longer breaks, but the class
of bug still exists: any user typo like *foo or (unclosed will
take down the whole categorize query.

Fix

Catch re.error in Rule.__init__, log a warning, and set self.regex = None so that rule matches nothing. Other rules keep working;
events that would have matched the broken rule fall through to
Uncategorized instead of tanking the query. The webui can still
provide inline validation on top of this later, but the server should
not 500 on user input.

Tests

Added two tests in tests/test_transforms.py:

• test_rule_invalid_regex_does_not_raise — invalid pattern gives
  regex=None, match(event) returns False.
• test_categorize_survives_invalid_regex — a bad rule mixed with a
  good rule: good rule still categorizes, unrelated events go to
  Uncategorized (instead of the whole call raising).

Existing test_categorize / test_tags unchanged and still pass.

Fixes ActivityWatch/activitywatch#1340

[greptile-apps]

Non-string regex errors escape

When a categorize or tag rule contains a truthy non-string regex value such as 123, re.compile raises TypeError, which bypasses this handler and still aborts the entire query with an internal server error.

Suggested change

	except re.error as e:
	except (re.error, TypeError) as e:

[greptile-apps]

Greptile Summary

This PR prevents syntactically invalid string regexes from aborting classification queries.

• Catches re.error while compiling category and tag rules.
• Logs invalid patterns and disables those rules so other rules continue matching.
• Adds direct and categorization-level regression tests.

Confidence Score: 3/5

The PR should not merge until malformed non-string regex values are prevented from escaping the new recovery path and aborting classification queries.

The change handles re.error correctly for malformed strings, but reachable query dictionaries can contain truthy non-string regex values for which re.compile raises an uncaught TypeError.

Files Needing Attention: aw_transform/classify.py

Important Files Changed

Filename	Overview
aw_transform/classify.py	Adds graceful handling for regex syntax errors, but truthy non-string rule values still escape as TypeError and abort the query.
tests/test_transforms.py	Covers invalid regex syntax and mixed good/bad rules, but does not cover malformed non-string regex values accepted by the query representation.

_{Reviews (1): Last reviewed commit: "fix(classify): do not crash on invalid u..." | Re-trigger Greptile}