Who uses "complete word" (aka SPC) in the minibuffer and how/why

Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Stefan Monnier]

The subject says it all.

As I'm looking at ways to improve the default completion mechanism, I'm
trying to understand how "complete word" would fit in.  In order to know
how best to adapt it to some of the situations, I'd like to know how
it's used.  I personally never use it and always find TAB to be a much
better choice.

Of course, ideally nobody uses it (except some people who happen to
like to hit SPC but would be just as well served by minibuffer-complete
if it were bound to SPC) and we can get rid of it, but I expect life
isn't that simple.


        Stefan

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Magnus Henoch]

I use it with M-x, especially when I want to run a command that is
shadowed by a shorter one, e.g. customize-variable (shadowed by
customize) and mail-add-payment-async (shadowed by mail-add-payment).
At the end of the common part I hit SPC to get completion of the longer
command.

I could do this with the sequence TAB and '-' too, but that would be one
extra keypress.

Magnus

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Stefan Monnier]

> I use it with M-x, especially when I want to run a command that is
> shadowed by a shorter one, e.g. customize-variable (shadowed by
> customize) and mail-add-payment-async (shadowed by mail-add-payment).
> At the end of the common part I hit SPC to get completion of the longer
> command.

> I could do this with the sequence TAB and '-' too, but that would be one
> extra keypress.

Do you ever use it elsewhere than with M-x ?


        Stefan

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Magnus Henoch]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Do you ever use it elsewhere than with M-x ?

Not that I can think of.

Magnus

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Bastien]

Magnus Henoch <mange@freemail.hu> writes:

> I use it with M-x, especially when I want to run a command that is
> shadowed by a shorter one, e.g. customize-variable (shadowed by
> customize) and mail-add-payment-async (shadowed by mail-add-payment).
> At the end of the common part I hit SPC to get completion of the longer
> command.
>
> I could do this with the sequence TAB and '-' too, but that would be one
> extra keypress.

A reasonable behavior for minibuffer-complete could be:

(0) try to complete
(1) if (0) fails, present the list of possible completions
(2) if the previous TAB was (1), complete by selecting the 
    string which will reduce the number of completions to 
    the larger subset of the list presented in (1)

In the case Magnus describes it would lead to:

M-x cust TAB    [complete => custom]
M-x custom TAB  [display the list of possible completions]
M-x custom TAB  [complete => customize]
M-x custom TAB  [display the list of possible completions]
M-x custom- TAB [display the list of remaining completions]

... 

-- 
Bastien

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Stefan Monnier]

>>>>> "Magnus" == Magnus Henoch <mange@freemail.hu> writes:

> I use it with M-x, especially when I want to run a command that is
> shadowed by a shorter one, e.g. customize-variable (shadowed by
> customize) and mail-add-payment-async (shadowed by mail-add-payment).
> At the end of the common part I hit SPC to get completion of the longer
> command.

> I could do this with the sequence TAB and '-' too, but that would be one
> extra keypress.

>>>>> "Bastien" == Bastien  <bzg@altern.org> writes:

> I hardly use it.  

> My only use is this one: in org-mode, when trying to refile a headline
> under another one, `minibuffer-completion' offers a list displaying all
> headlines, with a slash to indicate the hierarchy:

> Headline with a title
> Headline/Subheadline1
> Headline/Subheadline2
> Headline/Subheadline3
> Headline2/Subheadline2.1
> Headline2/Subheadline2.2
> Headline2/Subheadline2.3

> TAB  after Headline will always offer the complete list, whereas SPC
> will jump to the heading "Heading with a title".

I notice that both uses are in "must-match" completions.
Maybe we should remove the SPC binding from
minibuffer-local-completion-map to minibuffer-local-must-match-map.

In any case, it does help me understand the intention behind
the command.


        Stefan

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[paul r]

> I use it with M-x, especially when I want to run a command that is
>  shadowed by a shorter one, e.g. customize-variable (shadowed by
>  customize) and mail-add-payment-async (shadowed by mail-add-payment).
>  At the end of the common part I hit SPC to get completion of the longer
>  command.
>
>  I could do this with the sequence TAB and '-' too, but that would be one
>  extra keypress.

Same here, and I like it. I find it smart to use for completion a key
that is unlikely to make sens as self-insert. And to be honest, - is a
pain to hit on my keyboard.

-- Paul

PS : Attentive reader may have seen a parallel with Tab for completion
in M-:  ...  :)

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Stefan Monnier]

>> I use it with M-x, especially when I want to run a command that is
>> shadowed by a shorter one, e.g. customize-variable (shadowed by
>> customize) and mail-add-payment-async (shadowed by mail-add-payment).
>> At the end of the common part I hit SPC to get completion of the longer
>> command.
>> 
>> I could do this with the sequence TAB and '-' too, but that would be one
>> extra keypress.

> Same here, and I like it. I find it smart to use for completion a key
> that is unlikely to make sens as self-insert. And to be honest, - is a
> pain to hit on my keyboard.

So, in any case the "word" part of the completion behavior is not
something which you use?

To make it clear: minibuffer-complete-word has 2 main differences with
minibuffer-complete:
1 - if there's no completion, it tries to add a " " or a "-".
2 - it only completes a single word at a time.  E.g.

  M-x fill-region- TAB   => fill-region-as-paragraph
  M-x fill-region- SPC   => fill-region-as-

So you like 1 but you don't seem to be using 2.  Does anyone care about
the part 2 of the behavior of minibuffer-complete-word?

If not, that's great: it's the part that's nasty to implement and I'd be
happy to scrap it (or at least let it fail when it's too difficult to
make it work).


        Stefan

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Paul Rivier]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> If not, that's great: it's the part that's nasty to implement and I'd be
> happy to scrap it (or at least let it fail when it's too difficult to
> make it work).

I use it, although I admit it is no necessary, so I invite you to
carry any change that can suit better your plans.

-- 
Paul

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Stefan Monnier]

>> If not, that's great: it's the part that's nasty to implement and I'd be
>> happy to scrap it (or at least let it fail when it's too difficult to
>> make it work).

> I use it, although I admit it is no necessary, so I invite you to
> carry any change that can suit better your plans.

Could you expand on how you make use of it?


        Stefan

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Paul Rivier]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Could you expand on how you make use of it?

After reading again, nothing incompatible with the change you plan to
make. But thank you for your interest. I really simply use space for
completion in M-x, I just keep hitting space and disambiguating letter
after letter, but actually you are right that the "word-by-word"
completion is of no use here. If you plan to leave regular completion,
and "- when relevant", on space, I think I will not even notice the
change. So, definitely, I do not rely on the "word-by-word" feature.
Thanks.

-- 
Paul R

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Juri Linkov]

> If not, that's great: it's the part that's nasty to implement and I'd be
> happy to scrap it (or at least let it fail when it's too difficult to
> make it work).

I personally don't use word completion, but in Gtk the standard way of
completion is on word-by-word basis.  So it might be natural for Gnome users
to expect Emacs to complete the same way.  Though in Gtk word completion
is bound to TAB, and in Emacs it is bound to SPC, perhaps some users
might want to use SPC or rebind TAB to minibuffer-complete-word.

-- 
Juri Linkov
http://www.jurta.org/emacs/

Re: Who uses "complete word" (aka SPC) in the minibuffer and how/why

[Bastien]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> As I'm looking at ways to improve the default completion mechanism, I'm
> trying to understand how "complete word" would fit in.  In order to know
> how best to adapt it to some of the situations, I'd like to know how
> it's used.  I personally never use it and always find TAB to be a much
> better choice.

I hardly use it.  

My only use is this one: in org-mode, when trying to refile a headline
under another one, `minibuffer-completion' offers a list displaying all
headlines, with a slash to indicate the hierarchy:

Headline with a title
Headline/Subheadline1
Headline/Subheadline2
Headline/Subheadline3
Headline2/Subheadline2.1
Headline2/Subheadline2.2
Headline2/Subheadline2.3

TAB  after Headline will always offer the complete list, whereas SPC
will jump to the heading "Heading with a title".

-- 
Bastien