Re: new-flex-completion-style

[Óscar Fuentes]

Stefan Monnier <address@hidden> writes:

>> But you didn't answer my question: what specific features of flx
>> are you looking for?
>
> IIUC Le Wang's `flx` has two main features:
> - Its scoring is supposed to be particularly good.

This is all about personal taste. For me it is great, others don't think
so. Also, it takes a while to get used to and adjust your input
according to the scoring method.

> - It's supposed to be particularly fast.

Not quite. The initial implementation was terribly slow. Then it became
usable at the expense of memory usage (caching), which is way higher
than it should be for a feature like this.

It is in my to-do list to implement the algorithm in C, either into
Emacs core or as an extension, and see if the speed improvement is good
enough to dispense with all the caching.

>> I'd be happy to implement them if I have
>> the time, but I have to understand what they are.
>
> Other than using flx.el I don't see any good way to "implement" those
> features (especially the one about speed).
>
> Of course, I have no idea if flx.el's scoring is indeed significantly
> better in practice, nor if it's significantly faster in practice either.
>
> BTW, the issue of scoring quality is not nearly as obvious as one might
> think because the score one should give to a particular match depends on
> the user's expectation and the user's expectation often depend on the
> past behavior of the tool, i.e. on the scoring.

Indeed.