Re: fuzzy-matching in quickopen...

Hi,

On 2022-09-16 00:11, Alexander Neundorf wrote:
> Hi,
> 
> I updated my kate (since quite some time) again, and the new fuzzy 
> matching
> makes the quickopen in some cases close to unusable for me.
> 
> 1) files which match and which are already open do not always end up at 
> the top
> after filtering.
> So if I type some letters, now, many files match, including files which 
> I have
> not opened yet, and the one I want (where I was maybe 3 files ago) does 
> not
> even end up on the first page of results, but somewhere deep down. If I 
> have to
> scroll and search with my eyes for the correct file, it is not a 
> "quick" open
> anymore...
> 
> 2) it matches just too much, or at least the scores are not good. Let's 
> say I
> want to open a file named "creategdt.cpp", so I type "gdt". This will 
> match all
> files which contain a "g", a "d" and a "t" somewhere in their name, 
> like let's
> say "globaldestructor.cpp". The files which actually have "gdt" in 
> their name,
> do not necessarily end up at the top of the list, but I have to search 
> and
> scroll.
> As another example, when typing "cmake", it now among others suggest 
> "color-
> themes-gui-breeze-dark-default-text-styles.png." I think this is so far 
> off
> that it is not useful.
> 
> 3) it seems to be somewhat case sensitive. If I enter "c", it does not 
> show
> "CMakeLists.txt". If I enter more than 1 letter, it seems to become 
> case
> insensitive.
> 
> I would actually prefer to revert back to the old matching. 1) IMO 
> needs to be
> restored, files which are open should always be above files which are
> not open.
> Maybe the bonus of 1 point is just too small.
> 3) is less critical, but should be easy.
> 2) ... if I have typed at least 2 characters, filenames which actually 
> contain
> that short string should always get a higher score than filenames which 
> just
> contain those characters somewhere in their name.
> 
> I could spend some time to work in this.
> With the behaviour as it is now, I'm often quicker using the file 
> browser or
> "open documents" tabs.

I use the fuzzy matching there a lot, for me it behaves in most cases 
like I would expect.

But I see your points that one might want to tune some scores.

@waqar might give hints and I would expect some merge request with 
improvements like
higher scored open files and "perfect" matches are welcome.

Greetings
Christoph


> 
> Comments ?
> 
> Alex

-- 
Ignorance is bliss...
https://cullmann.io | https://kate-editor.org