RE: read-from-minibuffer with default value string

["Drew Adams" <drew.adams@oracle.com>]

> I'm confused by the documentation for read-from-minibuffer. It says
> that the INITIAL-CONTENTS argument is obsolete, but I can't
> duplicate the functionality using DEFAULT-VALUE.

You are correct that it does not _duplicate_ the functionality.

> I'm working on a function that will query the user for input using
> completing-read. If the value entered isn't in the collection used for
> completing-read, then the user is asked if they want to add 
> the value to the collection.

Note what you just said: "asked _if_" (i.e. whether).  To me, that says
`y-or-n-p'.

What you really do however is ask them again to input the (same) KEY (and then
for a corresponding CODE).  You ask them for the (same) KEY as a proxy for
directly asking _whether_ they want to add an entry for KEY.

> 	  (let* ((key (read-from-minibuffer (format "Journal 
> Title <%s>: " key) key))
> 		 (code (read-from-minibuffer "Journal Bibtex Code: ")))
> 	    (push (list key code) collection-list)
> 	    (message "Pass <%s> to another function here" code))))))
> 
> This uses the obsolete INITIAL-CONTENTS argument. If I use the DEFAULT
> value list, I either have to set the READ argument to nil, which means
> that entering an empty string results in key being set to the empty
> string rather than the default value; or I set READ to t, which means
> that any text the user enters is treated as an expression, which
> produces an error when the string contains a space. The strings will
> often contain spaces, so this isn't good.
> 
> How can I rewrite this to work without using INITIAL-CONTENTS?

A few alternatives I can think of:

1. Use `read-string' with KEY as the DEFAULT-VALUE.  This is likely what you
want to do here.


2. Use `read-from-minibuffer' with INITIAL-CONTENTS, but also use a cons as
HIST.  That usage is not considered obsolete AFAIK.  The doc string does suggest
that _any_ use of this parameter is "obsolete" but it also mentions using it
with a cons HIST value.

And the Elisp manual does not say INITIAL-CONTENTS is "obsolete" at all - it
says that it is "mostly deprecated" (equivocation), and it describes a use case
where it is presumably not deprecated:

     Use of INITIAL-CONTENTS is mostly deprecated; we recommend using a
     non-`nil' value only in conjunction with specifying a cons cell
     for HIST.  *Note Initial Input.

That's not the same thing as being obsolete, and the mention of using a cons
HIST seems to be outside of any deprecation/obsolescence.  But see also node
`Initial Input'.


3. Use `y-or-n-p' (or `yes-or-no-p') if you are not really letting users add a
different KEY here from what they entered the first time (in `completing-read')
and you just want to let them say _whether or not_ to add an entry with the
_given_ KEY (you would then still read the CODE).  IOW, you show them KEY in the
`y-or-n-p' prompt, which asks for confirmation to add an entry.


(4. You could even use `completing-read' with a singleton COLLECTION.  You
probably do not want to do that, but it is another alternative here.)


FWIW, _personally_ (but who cares?) I do not consider INITIAL-CONTENTS to be
obsolete.  It provides _different_ behavior from a DEFAULT value (as you
noticed).  In particular, the value is inserted in the minibuffer, and that
might be just what is wanted in some contexts.

Emacs Dev opted for declaring INITIAL-CONTENTS obsolete and putting the default
value inside prompts whenever they want to make it clear to users what the
default is.  Those are not quite the same thing.  Since there is a difference in
behaviors there can be uses for both.

The reason given for this obsolescence/deprecation?  From node `Initial Input':

   *We discourage use of a non-`nil' value for INITIAL*, because
   initial input is an intrusive interface.  History lists and default
   values provide a much more convenient method to offer useful default
   inputs to the user.

Intrusive?  Maybe - clearly it offers you an editable value to start with.  But
convenience is in the eye of the user.  If you have a key that empties the
minibuffer (I use `M-k' - none is provided by default in vanilla Emacs), then
it's pretty much a toss-up whether (a) a value is inserted initially and you
sometimes remove it or (b) nothing is inserted and you sometimes use `M-n' to
retrieve a value.

I personally prefer to have even the DEFAULT value inserted in the minibuffer,
because I often do want to edit it.  I use `M-k' to clear it if I don't want it
at all.  I do that less on average than I would have to use `M-n' to retrieve a
value to edit (or copy+paste from the prompt).

That's me - but everyone is different.  I don't think that one usage pattern
should be considered so superior in general that the other should be deprecated.
IMO, Emacs Dev should not have tried to be so controlling here - there was no
need for that.  But what do I know?


[FWIW, in Icicles a user option controls the treatment of the DEFAULT value
(when INTIAL-CONTENTS is nil or "").  Its possible values:

  nil               - Do not insert default value or add it to prompt.
  t                 - Add default value to prompt.  Do not insert it.
  `insert-start'    - Insert default value and leave cursor at start.
  `insert-end'      - Insert default value and leave cursor at end.
  `preselect-start' - Insert and preselect default value;
                      leave cursor at beginning.
  `preselect-end'   - Insert and preselect default value;
                      leave cursor at end.

My personal preference is `insert-end'.  The default value is `t' (similar to
vanilla Emacs behavior).]