5080 false positives for "Close similar name" with roman numerals

[Amunak]

We have a lot of numbered streets in Prague with only a roman numeral distinguishing them, and Osmose reports them as suspicious (see issues like this).

From reading the source regex it seems like this case should be skipped, but isn't for some reason.

Can you look into it please?

[Famlam]

Interesting, there's more strange errors with those names. For instance "No street with name "Herálecká II" found around" directly next to a street with that name https://www.openstreetmap.org/way/234453526

[Famlam]

Oh, I see, there's a special character used:
Ⅱ and Ⅲ instead of II and III
(The first two are a single character, the latter multiple uppercase i. On osm they look the same)

Is this a common script thing in your alphabet / street numbering? (Especially since the addr:street tagging seems to use regular uppercase i)

[Amunak]

Ahh interesting. I don't think you can even have characters like that in official street names; to me this appears more like a thing where an editor or maybe even the keyboard just automatically replaces the characters with the unicode for the roman numeral (or maybe the user just did it? Idk).

Like, technically it is a more correct representation, but I can also see how it could cause issues. With that being said I think the solution would be to just add the "number forms" unicode range to the regex.

[Famlam]

I would agree it's wrong to warn about "close similar names" indeed, because it's like "street 1" vs "street 2", so we should fix that.

Note to self: \p{N}?