Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T255502 Goal: Save Timing median back under 1 second | |||
| Open | None | T289923 Let AbuseFilter cache its parser AST result | |||
| Open | None | T290454 Create a new RuleChecker class to transform string rules into result objects |
Event Timeline
Comment Actions
I haven't investigated this and don't plan on doing that soon, hence unlicking. A few things that come to mind:
- An AST can be pretty big, so there could be a lot of data being cached if there are e.g. 200 active filters.
- Currently, only CachingParser knows about the cache, so we'd need an intermediary between the parser and its current callers that would take care of caching. This might also be a good time to rename CachingParser to Evaluator (or something) and have it not deal with the actual parser (AFPTreeParser) inside.
- The cache should be invalidated if any filter is created or modified.
Comment Actions
r716040 (WIP) goes in that direction, introducing such an intermediary class that controls the ASTs cache, and making the Evaluator not know about the previous steps (lexing, parsing, static checking). The new RuleChecker could also depend on FilterLookup and have a method for checking all active filters.