BUG: MatchRegex() "\b" doesn't work

Bugs & Issues
1

In a Formula field, a regex with \b never matches anything.

image
image1441×1220 71.9 KB

The above Formula should match these entities, but doesn’t match any:

image

2

It seems you’re right, but I don’t know why :confused:
In the meantime, for your specific case, could you just use
[\W^]PM[\W$]
instead?

1 Like
3

Maybe a Unicode-related bug somewhere? Or escaping/quoting? SQL?

4

I think it would need to be something more like
(^|\W)PM($|^W)
because you can’t put ^ $ in a character class - it’s just a literal there.

5

You’re probably right - regex is not my strong point to be honest!

6

Hi! I just came across this issue, and it still does not work i.e. word boundaries in Fibery Regex formula functions somehow do not get passed correctly to the Regex engine.

7

Since this issue still persists, I figured I would share the other workaround using negative lookbehind & lookahead assertions to replace the \b functionality:

Example: ReplaceRegex(Text, "(?<!\w)YOUR_WORD_REGEX(?!\w)", "YOUR_REPLACEMENT")

How it works:

  • (?<!X): this ensures the the match (i.e. YOUR_WORD_REGEX) is not preceded by X. In our example above the X is \w, which means letters, numbers and _ in ASCII*.
  • (?!X): this ensures the the match is not followed by X.

* Unicode versus ASCII
I have tested this with Unicode characters, and the \w matches exactly as expected. For example ReplaceRegex("and😀 and €and 𝐀and andé", "(?<!\w)and(?!\w)", "#") results in #😀 # €# 𝐀and andé as expected.

If you somehow still want to expand the definition of your word boundaries, you can convert the \w into a character class and add to it. For example (?<![\w-·]) would stop considering the dash and middle-dot boundaries, too.

1 Like