Feature request: specifying allowed characters when fuzzy-matching

[mrabarnett]

Original report by Anonymous.

---

It would be a neat feature if one, in addition to the type of fuzziness, could specify character classes when performing a fuzzy match. For instance, trying to fuzzy match a word with spelling errors it hardly makes sense to allow introduction of [^a-z] characters.

[mrabarnett]

Original comment by Matthew Barnett (Bitbucket: mrabarnett, GitHub: mrabarnett).

---

Could you mock up an example?

[mrabarnett]

Original comment by Peter Holm (Bitbucket: [Peter Holm](https://bitbucket.org/Peter Holm), ).

---

Say I have a large corpus of news articles, blogposts, etc. and I am looking for misspellings of "matthew mcconaughey". His last name is notoriously difficult to get right, so it is expected that a lot of variations will pop up. With a regex like (?i)matthew (mcconaughey){1<=e<=5} i would match stuff like mcconaughey. and _mcconaughey which isn't misspellings. One could imagine extending the syntax of the fuzzy matching to specify which kind of characters are allowed to be inserted/replaced/substituted. One idea of a potential syntax: (?i)matthew (mcconaughey){1<=e<=5|[a-z]} .

[mrabarnett]

Original comment by Matthew Barnett (Bitbucket: mrabarnett, GitHub: mrabarnett).

---

Your example looks wrong because it has [^a-z] where I'd expect [a-z].

[mrabarnett]

Original comment by Peter Holm (Bitbucket: [Peter Holm](https://bitbucket.org/Peter Holm), ).

---

Sorry, you’re correct I’ve edited it now :)

[mrabarnett]

Original comment by Matthew Barnett (Bitbucket: mrabarnett, GitHub: mrabarnett).

---

Done in regex 2019.08.19.

I've chosen to use ":" instead of a "|", so your example would be (?i)matthew (mcconaughey){1<=e<=5:[a-z]}, and it's not limited to a character set, but can also be, say, a property like \d or \p{Digit}.

[mrabarnett]

Original comment by Peter Holm (Bitbucket: [Peter Holm](https://bitbucket.org/Peter Holm), ).

---

Awesome! I like that solution a lot.