URL detection is too aggressive

[kuiperzone]

From what I can see, Folio auto detects a link in the text if merely contains a . with no whitespace. As a result, it also characterises simple file names and namespaces as links and my notes are littered with false positive "links".

See screenie.

May I suggest for purposes of auto-link detection, a text fragment must contain (but not start or end with) both:

://, .

where the first occurrence of . must be after the :// the scheme separator.

Or as well: a fragment may start with www. instead of containing ::/, though few use www any more.

This obviously shouldn't apply to links explicitly designated with ()[].

[toolstack]

I'm not surprise there are a few false positives, url detection is challenging in unstructured text.

I had it checking for :// but that excluded too much.

I'll take another look and see if it can be tuned a little more, otherwise I'm thinking that erring on the side of too many matches is better than too few.

[kuiperzone]

Hey thanks for the reply!

While I don't want to suggest an option for every problem...

Perhaps there could be an option for auto-detect links with 3 states:

1. Disabled
2. Cautious/Normal (using ://)
3. Aggressive/Verbose (as current)

?

[toolstack]

Perhaps, but I really want to rewrite the regex processor anyway at some point so don't want to put a whole lot of effort into playing whack a mole with the existing one.

[kuiperzone]

If we cannot have a robust URL detector, can we at least disable the auto URL detection. I don't actually see it as useful, but understand others might.

I guess this is content specific, but if you are keeping IT related notes than more than half of "URLs" will be false positives.

It falsely detects the following as URLs:

ISO times
filenames
version numbers
namespaces
Anything with a period in it

[toolstack]

I was thinking of adding a disable url detection option, so I don't see that being an issue.

Unfortunately url detection is a messy business in free form text, so it's never going to be perfect.

I might do a three way selector; aggressive, strict, disable.

Strict would require a proper protocol part to exist (aka https://, etc.) before it was identified as a url.

[kuiperzone]

Good stuff!

Aggressive and strict? That doesn't sound fun. I'll be selecting disabled then. :)