Lets say I have a window
"D19473 Present Windows Effect: Allow imprecise match on filtering windows - Mozilla Firefox"

And search for "telegram", I'm going to get all the way to "telegr" before the test fails.
(d19473 presenT windows Effect: aLlow imprEcise match on filterinG windows - mozilla fiRefox)

It's hindering the original goal of making things faster to search.
I fear it's going to appear broken to the end user who can't see why it's showing these false positives, even if this is an option.

In D19473#423221, @davidedmundson wrote:

And search for "telegram", I'm going to get all the way to "telegr" before the test fails.
(d19473 presenT windows Effect: aLlow imprEcise match on filterinG windows - mozilla fiRefox)

It's hindering the original goal of making things faster to search.
I fear it's going to appear broken to the end user who can't see why it's showing these false positives, even if this is an option.

I'm afraid I have to agree.

I also agree with David. For end users it will look broken (unless matched characters are visualized and even in that case I think it's not worth that). Algorithm-wise, I'm not sure whether the proposed fuzzy string matching algortihm is good enough.

In D19473#423221, @davidedmundson wrote:

Lets say I have a window
"D19473 Present Windows Effect: Allow imprecise match on filtering windows - Mozilla Firefox"

And search for "telegram", I'm going to get all the way to "telegr" before the test fails.
(d19473 presenT windows Effect: aLlow imprEcise match on filterinG windows - mozilla fiRefox)

It's hindering the original goal of making things faster to search.
I fear it's going to appear broken to the end user who can't see why it's showing these false positives, even if this is an option.

That's a good counterexample I had not though about :-)

Thank you all for the comments.

One improvement I can think of, as zzag suggested is somehow highlighting the matched characters in the windows titles.

Another one, algorithm-wise maybe not considering non alphanumeric chars as part of the "gaps" between two chars in the filter.

So that "firefox: a windows title page - subtitle" won't match to ffoxsub, but will match to ffox:-sub, as : and - must be present. It would make matching user@machine: /opt/cling/lib/clang/5.0.0/include with something like 'machineinclude' quite complicated, due the number of '/'s in between.

Another one is making the search case sensitive, what would certainly solve the "telegram" issue above mentioned, but would feel awkward as the behaviour is different from the 'precise' filtering, which is case insensitive. I personally would not mind having a case sensitive filter.

Please let me know if you have any idea for improvements or better filtering algorithm.

Cheers,
Leandro