rewriting y-or-n-p to use read-char-choice

rewriting y-or-n-p to use read-char-choice

[Ted Zlatanov]

Stefan Monnier asked me to look at rewriting `y-or-n-p' to use
`read-char-choice'.  This looked easy at first but I realized it's not
such a good idea because:

- `y-or-n-p' uses `query-replace-map' which could be modified
  externally.  So it would have to gather the key choices from the key
  map, and even that's not possible because it takes the default, so any
  character can potentially be meaningful.

- `read-char-choice' has special code for keyboard macros which I think
  will break `y-or-n-p'; `y-or-n-p' OTOH calls `discard-input'

- the amount of duplicated code is not much at all

So I think a better way would be to write a
`read-char-choice-from-keymap' function which can be passed
`query-replace-map' or something else AND a prompt and a list of
acceptable return values.

This function then calls `lookup-key' on the keyboard input and returns
a symbolic value.  This function could also build the a/b/c part of the
prompt dynamically based on the acceptable key choices.  But if
`y-or-n-p' is the only consumer for this functionality, it seems
pointless to add it.

Ted

Re: rewriting y-or-n-p to use read-char-choice

[Edward O'Connor]

Hi,

> Stefan Monnier asked me to look at rewriting `y-or-n-p' to use
> `read-char-choice'.  This looked easy at first but I realized it's not
> such a good idea because:
[snip]
> So I think a better way would be to write a
> `read-char-choice-from-keymap' function which can be passed
> `query-replace-map' or something else AND a prompt and a list of
> acceptable return values.

I don't know if this is at all useful, but I had implemented the
functionality that `read-char-choice' provides a couple of years ago:

http://edward.oconnor.cx/elisp/read-char-spec.el

I've signed papers, so feel free to pilfer anything useful out of there.


Ted

Re: rewriting y-or-n-p to use read-char-choice

[Ted Zlatanov]

On Thu, 10 Mar 2011 11:10:06 -0800 "Edward O'Connor" <hober0@gmail.com> wrote: 

EO> Hi,
>> Stefan Monnier asked me to look at rewriting `y-or-n-p' to use
>> `read-char-choice'.  This looked easy at first but I realized it's not
>> such a good idea because:
EO> [snip]
>> So I think a better way would be to write a
>> `read-char-choice-from-keymap' function which can be passed
>> `query-replace-map' or something else AND a prompt and a list of
>> acceptable return values.

EO> I don't know if this is at all useful, but I had implemented the
EO> functionality that `read-char-choice' provides a couple of years ago:

EO> http://edward.oconnor.cx/elisp/read-char-spec.el

EO> I've signed papers, so feel free to pilfer anything useful out of there.

Thanks.  I think at least the ?? help behavior should be in
`read-char-choice', and appending the mapconcat-ed list of choices is a
nice feature (I did something similar for `auth-source-read-char-choice').

I really think, however, that a keymap is a necessity in this kind of
prompting function.  We want it to be flexible multilingually and we
very rarely actually want the char value of the key pressed.  We usually
want a symbolic return.  It's also nice to accept more than literal
characters.

Ted

RE: rewriting y-or-n-p to use read-char-choice

[Drew Adams]

> Stefan Monnier asked me to look at rewriting `y-or-n-p' to use
> `read-char-choice'.

Completely missing from your exposition of implementation issues is why.  Why
rewrite `y-or-n-p'?  What advantages are sought/expected?

Re: rewriting y-or-n-p to use read-char-choice

[Ted Zlatanov]

On Thu, 10 Mar 2011 14:35:49 -0800 "Drew Adams" <drew.adams@oracle.com> wrote: 

>> Stefan Monnier asked me to look at rewriting `y-or-n-p' to use
>> `read-char-choice'.

DA> Completely missing from your exposition of implementation issues is why.  Why
DA> rewrite `y-or-n-p'?  What advantages are sought/expected?

See bug #8050.  The original need was, I believe, to reuse code.

Ted

RE: rewriting y-or-n-p to use read-char-choice

[Drew Adams]

> > Completely missing from your exposition of implementation 
> > issues is why.  Why rewrite `y-or-n-p'?  What advantages
> > are sought/expected?
> 
> See bug #8050.  The original need was, I believe, to reuse code.

I see.  Thanks (anyway).  Might I suggest that next time you briefly describe
the problem you are trying to solve, in the thread where you invite discussion
of possible solutions?  That might help you as well as others.

Re: rewriting y-or-n-p to use read-char-choice

[Stefan Monnier]

> So I think a better way would be to write a
> `read-char-choice-from-keymap' function which can be passed
> `query-replace-map' or something else AND a prompt and a list of
> acceptable return values.

Using a keymap sounds good.  It also means that the right way to
implement it is by using read-key-sequence (temporarily overshadowing
all the normal keymaps with the one provided).


        Stefan