@iBug It would be a good idea to run pip3 install -r user_requirements.txt --upgrade when you have a chance. It's not required in order for SD to continue to run, but it would be convenient.
@Makyen That pattern looks like it's already caught by Pattern-matching website in body and Pattern-matching website in answer; append -force if you really want to do that.
@SmokeDetector They self-cite, but it's disclosed. They provide their email address, but it sounds more like they are willing to have further discussions beyond just the question.
@tripleee That pattern looks like it's already caught by Pattern-matching website in answer and Pattern-matching website in body; append -force if you really want to do that.
@tripleee Yeah, several times I've seen situations where rules requiring more than one thing would be beneficial. We have a few very limited rules along those lines implemented as regular expressions, but they tend to be a bit compute-intensive when implemented as a pure regex, and, obviously, can only look in one block of text at a time (e.g. "buy" within 400 characters of a link, but not in code).
Along those lines, I've been thinking of implementing some pre-processing for each post, where it generates a lists of links, a possible phone numbers, email addresses, etc. and then all detections can use those lists to implement some meta, or just more complex, lookarounds without having to parse out those types of things in each detection.
For the meta stuff, I've thought about having some type of syntax contained in regex comments, which would allow us to have meta criteria within the watch and blacklists. At least to the extent of something like (?#CMD:NO_CODE), or possibly more elaborate: (?#CMD:CONTACT_WITHIN[200]), (?#CMD:LINK_WITHIN[200]), (?#CMD:LENGTH_MAX[250]), or (?#CMD:USERNAME_REGEX[dav+y\d*]).
If done, it would have significant limitations, because regex comments can't contain ), assuming it would need to maintain regex compatibility.
@Makyen the code in findspam.py already has the beginnings of a reusable list of URLs but it's kind of tacky; it would make more sense if it was parsed out at the beginning and made an attribute of the post
e.g. A list that for each <a> has properties like: start_character, end_character, length, href, domain, domain_is_whitelisted, url_is_whitelisted, link_display_html, link_html, etc.
@tripleee That pattern looks like it's already caught by Bad ip for hostname in body and Bad ip for hostname in answer; append -force if you really want to do that.
too lazy to add ownership links; I guess we should automate linking between identical whois registrations at some point (though that will be tricky to solve in a general fashion)
Deleting and then reposting essentially the same question is pushing you towards a question ban. Probably step back and figure out whether you can reduce the problem to multiple smaller questions, and post one at a time, with sufficient time between them to digest the comments and answers you get at each step. Probably also review the help center around How to ask. — tripleee1 min ago
reducing that to \S{0,3} would do what you propose but probably check against the matches in metasmoke to see what it excludes out of genuine hits before committing
the intent of course is to cover stuff like a$zzhole and the problem is that it also matches stuff like alt, answer, etc; maybe restrict the range of characters to just sibilants and punctuation instead?
@Kulfy I can't recall how obvious it is to new users that their answer was deleted, and how to get it undeleted. But yeah, reposting is usually a bad move
