GNU bug report logs - #45893
[PATCH 0/2] DRAFT: Hint for options.

Previous Next

Full log

View this message in rfc822 format

From: zimoun <zimon.toutoune <at> gmail.com>
To: Ludovic Courtès <ludo <at> gnu.org>
Cc: 45893 <at> debbugs.gnu.org
Subject: [bug#45893] [PATCH 0/2] DRAFT: Hint for options.
Date: Tue, 19 Jan 2021 19:01:24 +0100

Hi,

On Tue, 19 Jan 2021 at 18:38, Ludovic Courtès <ludo <at> gnu.org> wrote:
>
> zimoun <zimon.toutoune <at> gmail.com> skribis:
>
> > * guix/ui.scm (run-guix-command): Add command hint.
>
> [...]
>
> > +    (fold (lambda (name res)
> > +            (if (string-null? res)
> > +                (string-append  "@code{" name "}")
> > +                (string-append "@code{" name "}, " res)))
> > +          ""
> > +          (string-closest (symbol->string guess) command-names)))
>
> Hmm I thought ‘string-closest’ would return a single string, but
> actually it returns a list of strings?
>
> You cannot append strings together like this as this can break i18n.

Hum?  But it is not (G_ "")...

> The proper way would be to write:
>
>   "Do you mean one of these: ~a?"
>
> but then, thinking about it, wouldn’t it be more natural to suggest a
> single command rather than several?

...but the real question is this: one or several hints.

Yeah, if we assume that it is about typo on the command line and the
option names are different enough, which are both 2 reasonable
assumptions, then it should always return one hint.

Well, it depends if we consider the case where the typo is at the
exact same distance as 2 different option names.

> Also, it seems to me that there’s always at least one hit, is that
> correct?  We should make sure that strings above a certain distance are
> ignored, in which case there’s no hint to display.

The hint reports all the options which are at the same minimum
distance; whatever the minimum is.  For instance "guix show --kikoo"
would hint something.

Instead, and including your remarks, maybe if the distance is greater
than a threshold, then return an error as usual.  Make more sense.


> Next up is package names, right?  :-)

Hehe!  I have tried...  But it is not "doable" in practise... well, I
find it too slow.  The natural improvement is to cut down the
levenhstein-distance by stopping if the score is greater than
threshold.  Well, I have not tried yet. :-)


Thanks for the feedback and the comment.

Cheers,
simon

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.