try-completion not showing all possible completions?

try-completion not showing all possible completions?

[Chris Mears]

Hello all,

Executing the following code:

    (let ((table (make-vector 1021 0)))
      (intern "johnny jones" table)
      (intern "joe blow" table)
      (intern "jimi hendrix" table)
      (try-completion "j" table (lambda (s) (princ (format "%s\n" s)))))

I would expect the output to be all three names in the table.  However,
in the current CVS Emacs it produces:

    johnny jones
    jimi hendrix
    "j"

"joe blow" has been omitted.  I tested the same code in Emacs 21.4.1 and
it does what I expected:

    johnny jones
    jimi hendrix
    joe blow
    "j"

Can anyone shed some light on this?

Thanks,
Chris Mears

Re: try-completion not showing all possible completions?

[Stefan Monnier]

>     (let ((table (make-vector 1021 0)))
>       (intern "johnny jones" table)
>       (intern "joe blow" table)
>       (intern "jimi hendrix" table)
>       (try-completion "j" table (lambda (s) (princ (format "%s\n" s)))))

> I would expect the output to be all three names in the table.  However,
> in the current CVS Emacs it produces:

>     johnny jones
>     jimi hendrix
>     "j"

> "joe blow" has been omitted.  I tested the same code in Emacs 21.4.1 and
> it does what I expected:

>     johnny jones
>     jimi hendrix
>     joe blow
>     "j"

> Can anyone shed some light on this?

Simple optimization: after seeing johnny jones and jimi hendrix,
try-completion already knows that the return value will be "j", so it
doesn't look any further.


        Stefan

Re: try-completion not showing all possible completions?

[Chris Mears]

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

>>     (let ((table (make-vector 1021 0)))
>>       (intern "johnny jones" table)
>>       (intern "joe blow" table)
>>       (intern "jimi hendrix" table)
>>       (try-completion "j" table (lambda (s) (princ (format "%s\n" s)))))
>
>> I would expect the output to be all three names in the table.  However,
>> in the current CVS Emacs it produces:
>
>>     johnny jones
>>     jimi hendrix
>>     "j"
[...]
> Simple optimization: after seeing johnny jones and jimi hendrix,
> try-completion already knows that the return value will be "j", so it
> doesn't look any further.

Thank you, that makes perfect sense.  The documentation for
try-completion is a little misleading; it says:

"Each car of each element of ALIST (or each element if it is not a cons
cell) is tested to see if it begins with STRING."

I understood this to mean that the predicate (try-completion's third
argument) would be run on *each and every* element of ALIST.

Re: try-completion not showing all possible completions?

[Stefan Monnier]

> Thank you, that makes perfect sense.  The documentation for
> try-completion is a little misleading; it says:

> "Each car of each element of ALIST (or each element if it is not a cons
> cell) is tested to see if it begins with STRING."

> I understood this to mean that the predicate (try-completion's third
> argument) would be run on *each and every* element of ALIST.

The predicate is definitely not used on each and every element of ALIST.
In older Emacsen it was called on each and every element that started with
STRING (tho it might also depend on completion-regexp-list) and now it's
called in slightly fewer cases.
Basically it's called whenever it's needed.  Notice that the piece of
docstring you quote doesn't mention the predicate at all.


        Stefan