Re: Emacs needs truely useful flex matching

[Le Wang]

On Fri, Mar 22, 2013 at 1:49 AM, Óscar Fuentes <address@hidden> wrote:
> Le Wang <address@hidden> writes:
>
>> Since there is a big thread about a standard way to recognise project roots, 
>> I
>> want to bring up another area in which Emacs is falling behind other
>> Editors (Sublime Text, Textmate, and Vim).
>> Choosing from a very large list of files (or any strings for that matter) 
>> with
>> a minimum of keystrokes.
>>
>> ido-mode has `ido-enable-flex-matching', but that does not do the smart
>> sorting required.
>>
>> Vim has this through the Command-T plugin.  The best way to "get it" is by
>> watching it in action:
>> https://s3.amazonaws.com/s3.wincent.com/command-t/screencasts/command-t-demo.mov
>>
>> Skip to 6 minutes to see it in action in a large project with
>> repetitive path segments.
>>
>> Homepage here: https://wincent.com/products/command-t
>>
>> The matching engine is implemented in C and it interfaces with Vim through 
>> the
>> Ruby bridge.  I think in order to sort a large list of strings (> 10k), this
>> will also have to be implemented in C to be fast enough if done for Emacs.
>>
>> The sorting algorithm is roughly this for a query: "abcd"
>>
>> 1. Get all matches for "a.*b.*c.*c"
>> 2. Calculate score of each match
>>     - contiguous matched chars gets a boost
>>     - matches at word and camelCase boundaries (abbreviation) get a boost
>>     - matches with smallest starting index gets a boost
>> 2. Sort list according to score.
>>
>> A lot of my colleagues use this kind of flex matching to navigate around our
>> large code base in Sublime Text and I'm very jealous that with so few
>> keystrokes the most useful
>> matches just float to the top.
>>
>> This navigation could be implemented with Helm if Emacs had a builtin
>> fast smart flex sorting engine.
>
> IIUC the vim plugin you mention depends on a pre-built list of files. In
> that regard, how is it better than GNU Global, which allows to search
> files with a regex (and much more)? Maybe the algorithm you describe
> can be implemented on GNU Global and then use an Emacs interface.

The type of smart matching + sorting would be useful for files, M-x,
C-h v, all kinds of completion where large lists are involved.  GNU
globals maybe should support it, but Emacs shouldn't depend on
external dependency to offer something this useful.

> I agree that it would be nice to have such a feature native on Emacs,
> though. IIRC someone mentioned a few days ago that a regex of the type
> you descibe has very high execution complexity (-> is slow)

That is true, and speed is of the utmost essence, so emacs-lisp is
probably too slow.

> but obviously it is working for that vim extension.

Textmate has had this for a while, and I see Sublime Text do it very
well every day, of course.  :(

These editors are moving the ball for what people expect a modern
editor to provide.

Honestly when I first tried `ido-enable-flex-matchine', I thought it
was strange,  So it's entirely possible I haven't made the strongest
case for the usefulness of this feature.  Please let me know if you
aren't convinced..


-- 
Le