RE: SPC in completing minibuffers?

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

> I have hit two different instances where I needed to use C-q SPC
> to insert a blank in the minibuffer recently.
>
> One is with `insert-char', if you try to complete a unicode name
> with spaces (almost all have one).  The other is with `ert'
> when prompted for the selector.  When I try to give a tag based
> selector like (tag :foo) I need to use C-q SPC to insert the space.
> If I type a normal space, I get [No match].
>
> As a long time Emacs user I know how to work around these issues,
> but I was thinking: Aren't they really bugs?  Is it expected in
> some situations that the user needs to escape SPC in the minibuffer,
> or do we consider these as bugs?

FWIW, yes, `SPC' should self-insert in the minibuffer - as should
`?' and `C-j' (newline).  This should be a no-brainer, IMHO.

Minibuffer input with completion can be used for *all kinds*
of things.  Thinking that in most cases such input cannot contain
`SPC' etc. is overly narrow.

I proposed such self-insertion many moon ago (Icicles has had it
for a decade), but it was rejected.  I have never seen a good
argument for not self-inserting such chars.

After a long while, `SPC' was finally allowed to self-insert for
file/dir-name completion (only), though IIRC some felt that
supporting this amounts to caving in to MS Windows users.

Let's hope that your reviving the suggestion will have more luck
this time around.

As a longtime user, you know that you can easily overcome this
limitation for own use by binding `SPC' etc. in minibuffer keymaps
to whatever commands you like.  But yes, it would be better to
make Emacs behave nicely for everyone, out of the box.