Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 949. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 0. Uptime: 00:00:14.4788282.
Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 949. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 754. Uptime: 00:02:26.5793838.
Right, hmm. I'm just thinking, so would it actually be necessary to have redos protection in the first place then? Since all the strings Pham checks never very, very, very rarely consist of a single continuous series of chars.
And (XY)*(XY)* can be ([a-z] )*([a-z] ){1,50}, which is even harder to detect.
Actually, detecting explosive quantifiers are very difficult. Github can't do it, so you have to grep binary data on the repository on a local machine yourself.
@Sam Even if you manage to go through the algorithm to unfold a regex, understanding the pattern is not as simple. Because then you will also have X*XX*, or X*[XY]*Y*, which are all derivatives of `X*X*``.
No, it isn't. I've noticed this bug before, and tried to patch it (which seemed to work). I'll work on it asap after I've finished refactoring the command processor...
I'm still not quite sure where it's getting these massive numbers from. Since newly added terms are set to the average score for that specific filter...
The term's score 9364.7 should be 9.3647 (I think), but when Pat ran Pham (and when all the terms/scores were read during startup) there must've been some sort of culture diff therefore resulting in an incorrect decimal reading.
Although, that doesn't explain why that's the only term with an abnormal score.
@JanDvorak It's a display bug (the post wouldn't have been caught if it weren't escaped). I can also see in the terms files that the actual term is \$\d+.
Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 950. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 774. Uptime: 00:01:00.4314565.
Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 950. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 0. Uptime: 00:00:59.7884198.
Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 950. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 0. Uptime: 00:21:05.0803584.