From: Tomi Ollila <tomi.ollila@iki.fi>
To: Christopher Wellons <wellons@nullprogram.com>, notmuch@notmuchmail.org
Subject: Re: [PATCH] Switch to DEF from INITIAL-INPUT in completing-read.
In-Reply-To: <m2r4dmrgwg.fsf@guru.guru-group.fi>
References: <1377103488-7425-1-git-send-email-wellons@nullprogram.com>
	<m2r4dmrgwg.fsf@guru.guru-group.fi>
User-Agent: Notmuch/0.16+3~g340c058 (http://notmuchmail.org) Emacs/24.3.1
	(x86_64-unknown-linux-gnu)
Date: Thu, 22 Aug 2013 20:08:46 +0300
Message-ID: <m21u5lbqld.fsf@guru.guru-group.fi>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: list

On Wed, Aug 21 2013, Tomi Ollila <tomi.ollila@iki.fi> wrote:

> On Wed, Aug 21 2013, Christopher Wellons <wellons@nullprogram.com> wrote:
>
>> As stated in the Emacs documentation, the initial-input argument is
>> deprecated because it presents a poor interface to the user. In fact,
>> with my setup where ido replaces completing-read, it's nearly unusable
>> with initial-input.
>>
>> http://www.gnu.org/software/emacs/manual/html_node/elisp/Initial-Input.html
>> ---
>
> I agree that this behaves better, but the empty content with completing-read
> looks a bit confusing (if arrow keys or tab aren't pressed)
>
> Although the completing-read documentation also deprecates (STRING . POSITION)
> format IMO I found that to work best there: i.e.
>
> -   prompt collection nil nil initial-input 'notmuch-address-history))
> +   prompt collection nil nil (cons initial-input 0) 'notmuch-address-history))

Actually, I don't know this anymore. Yesterday the current worked a bit
badly but today while I was testing something else the current worked ok...

... but you're right, with ido-completing-read the current is unusable:

(load-library "ido")
(ido-completing-read "foo" '("foo" "bar" "baz") nil nil "foo")

and especially

(ido-completing-read "foo" '("foo" "bar" "baz") nil nil '("foo" . 0))

and, 

(ido-completing-read "foo" '("foo" "bar" "baz") nil nil nil nil "foo")

surely works the best of the above...

we could also drop the default/initial input:

(ido-completing-read "foo" '("foo" "bar" "baz"))
(completing-read "foo" '("foo" "bar" "baz"))

So that, with completing-read, user presses enter on empty content,
empty string is returned... and doesn't affect how ido-completing-read
does it...

>
> I don't know how ido-replaced completing-read (just ido-completing-read ?)
> works there, though.
>
> Better get other thoughts/opinions too -- I use this:
>
> https://github.com/domo141/nottoomuch/blob/dogfood/selection-menu.rst
>
> for choosing completions -- it ignores INITIAL-INPUT (/DEF) argument
> altogether. 
>
> Tomi

Tomi

>
>
>>  emacs/notmuch-address.el | 6 +++---
>>  1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/emacs/notmuch-address.el b/emacs/notmuch-address.el
>> index fa65cd5..e89200b 100644
>> --- a/emacs/notmuch-address.el
>> +++ b/emacs/notmuch-address.el
>> @@ -42,11 +42,11 @@ to know how address selection is made by default."
>>    :group 'notmuch-send
>>    :group 'notmuch-external)
>>  
>> -(defun notmuch-address-selection-function (prompt collection initial-input)
>> +(defun notmuch-address-selection-function (prompt collection def)
>>    "Call (`completing-read'
>> -      PROMPT COLLECTION nil nil INITIAL-INPUT 'notmuch-address-history)"
>> +      PROMPT COLLECTION nil nil nil 'notmuch-address-history DEF)"
>>    (completing-read
>> -   prompt collection nil nil initial-input 'notmuch-address-history))
>> +   prompt collection nil nil nil 'notmuch-address-history def))
>>  
>>  (defvar notmuch-address-message-alist-member
>>    '("^\\(Resent-\\)?\\(To\\|B?Cc\\|Reply-To\\|From\\|Mail-Followup-To\\|Mail-Copies-To\\):"
>> -- 
>> 1.8.4.rc3