Strip spoiler tags for watchlist triggers

[jerbob]

Closes #299

[eivl]

do we want to match

n||F||o||U||t||C||h||K||i||Y||n||O||g||U||

[scragly]

I played around and wrote a possible method to split things up into "interpretations" of text messed up with spoilers:

SPOILER_RE = re.compile(r"\|\|", re.DOTALL)

def process_spoilers(text):
    split_text = SPOILER_RE.split(text)
    public_text = "".join(split_text[0:][::2])
    spoiler_text = "".join(split_text[1:][::2])
    cleaned_text =  "".join(split_text)
    return public_text, spoiler_text, cleaned_text

@eivl's example would give:

('nothing', 'FUCKYOU', 'nFoUtChKiYnOgU')

[jerbob]

Thanks @scragly, I'll add it in

[lemonsaurus]

@AnonGuy what's the status on this? is it done? it's still labeled WIP.

[lemonsaurus]

@AnonGuy Looks like this is a good start, but there are multiple cases that this PR does not resolve, so it feels a bit like a half measure.

This PR does catch the simplest, ||badword||, but beyond that it does not catch most of the stuff discussed in this PR and in the issue this was based on.

For example, have a look at this embed:

Sure, the filter does match it, but the username is spoilered, the location has spoilers, and the original message has spoilers.

It would be better her to display these with the pipes escaped. Can we add a content.replace("|", "\|") in some tactical location so the embeds display correctly?

It also does not catch the cases that @eivl touched on earlier here, stuff like n||B||o||A||t||D||h||W||i||O||n||R||g||D||. Did we decide not to do that? I'll take it up in #dev-core After discussing this in #dev-core, we don't want to look for these obfuscated cases, so don't worry about this.

[eivl]

looks fine, just surprised that there are no more changes then this :D

[lemonsaurus]

hell yeah, let's merge this bad boy