Ignore code blocks for English-text-on-localized-site

[user12986714]

According to issue #3844

Changed regex so any code block will be excluded when evaluating whether there is English text on localized site.

Warning: the updated regex will exclude all pattern "sometext>". May be exploitable by spammers, by appending a ">" to the blacklisted word to circumvent blacklisting.

[makyen]

Just FYI: the following regex will match valid HTML tags in HTML received from SE (and then pre-processed by SD):

/<((?:\/?(?:b|blockquote|code|del|dd|dl|dt|em|h[123]|i|kbd|li|p|s|sup|sub|strong|strike|ul|br|hr))|(?:\/(?:a|div|img|ol|pre|span))|(?:a|div|img|ol|pre|span)\b.*?)\s*\/?>/

Note that this is a JavaScript based regex, so you'll want to drop the / at the start and end.

The above regex could be used to strip the HTML tags from the text prior to running the detection. You may need to look at the interaction of how we strip code blocks. There was an issue in how that was done for another detection. I'd need to double check...

[user12986714]

There are some issues with my second commit. The regex won't work as intended and the line was too long. Decided to revert the second commit.

[user12986714]

@makyen Looks like this regex won't work with negative lookahead. Yeah as you have said some magic may be needed in stripping, but a temporary solution may be disabling some phrases ended with '>'. (Commit 1c89 won't work with urls, they are too long so <a href="blah blah blah very long"> won't be disabled)

[user12986714]

After checking the code base, it is not a trivial task for me to use that regex in stripping without major refactoring of the entire stripping/spam detection code logic. This may be possible when resolving issue #2500, in which refactoring is proposed. (@makyen )

[makyen]

Closed by request per chat.