Continuing the discussion from Inappropriate / Obscenity / Profanity Language Filter:
So the profanity filter works well … not that we see it used often.
But is there a way to add regex/patterns so it could block people from putting an email address in a post?
That would be extremely dangerous, though.
“Extremely dangerous” is a strong phrase. Can you say more about why you feel that way?
Etc etc etc
Regex is like a language unto itself. Even a lot of seasoned programmers have trouble with it.
Using it requires not only understanding every possible variation you want to match, but also every possible variation you want to not match.
A lot of people do fairy well with the first, but fail with the second.
For example, using (.)* matches everything, anything, and nothing.
I see it used way too often as a “short cut” to get things to match, but unfortunately it often results in matching what it shouldn’t.
I guess if it were under the “developers only” section it might be enough to scare off Admins that shouldn’t mess with it. But human nature being what it is, give out loaded guns and it’s only a matter of time before someone shoots themselves in their foot.
And as for a valid email regex, it is notoriously difficult to craft a fool-proof one. Many come close and are “good enough” but without additional processing there will likely be problems at some point.
Fair enough, but the filter already exists in Discourse, even if you aren’t using it personally. Also, keep in mind (in response to your blog post) that the Discourse filter doesn’t replace strings, it masks them with squares. 
Regexes are inherently difficult, so I’m not necessarily proposing that you ask everyday users to use them as the mainstream use case. The current system works fine for most cases, but there’s no way to surefire way prevent people from posting most common email addresses. (I am not interested in the debate on the “perfect” email regex.)
Meanwhile, I’m simply blocking some of the most common domain names like @gmail.com, @yahoo.com, etc.
I am not interested in letting perfect get in the way of good here. Just trying to prevent the most common occurrences.
Actually, last I knew it replaces the characters with the box decimal value
eg. blocking “@gmail.com”, “someone@gmail.com” would look like
someone■■■■■■■■■■
and the source would be
someone■■■■■■■■■■
That is what I meant when I finished the sentence with:
Forgive my error of specificity. What I meant was that it doesn’t replace it with other letters to change the word, as described in the blog post above. 
Bonjour,
Je souhaite donc empêcher les utilisateurs de partager leur adresse e-mail dans les discussions publiques de notre communauté, afin de protéger leur vie privée (peut-être que certaines personnes, malgré nos meilleurs efforts, ne réalisent pas que les discussions sont publiques ?).
Cette approche est-elle appropriée ? Ou présente-t-elle des risques importants que je n’aurais pas pris en compte ?
*@*.com
*@*.org
*@*.net
*@*.edu
*@*.info
*@*.biz
Toutes les expressions régulières comportent un risque élevé ; plus elles sont larges, plus le risque est grand. Celles-ci sont… assez risquées.
Mon espoir naïf était que le « @ » et l’inclusion des noms de domaine de premier niveau permettent de restreindre la recherche aux adresses e-mail uniquement. N’y a-t-il aucun moyen de cibler ces éléments ?
Quelque chose comme *.?@gmail.com serait nettement plus sûr. Idéalement, je dirais uniquement des caractères de mots, pas d’astérisque (tous les caractères).