last-abbrev-location

last-abbrev-location

[Luc Teirlinck]

Unlike `abbrev-start-location', which is a marker,
`last-abbrev-location' is just a buffer position, an integer.  Is
there a reason for this?  It seems to cause problems.

Do `emacs -q'.  Define an abbrev `exp' expanding to `expansion'.

In the *scratch* buffer type exp in the following position (no
trailing whitespace):

;; This buffer is for notes you don't want to save, and for Lisp evaluation.
;; If you want to create a file, visit that file with C-x C-f,
;; then enter the text in that file's own buffer.

exp

Then expand exp.  Go to the previous line and type 12345:

;; This buffer is for notes you don't want to save, and for Lisp evaluation.
;; If you want to create a file, visit that file with C-x C-f,
;; then enter the text in that file's own buffer.
12345
expansion


M-x unexpand-abbrev results in:

;; This buffer is for notes you don't want to save, and for Lisp evaluation.
;; If you want to create a file, visit that file with C-x C-f,
;; then enter the text in that file's own buffer.
1expnsion

This does not seem right and I have the impression that if
`last-abbrev-location' were a marker instead of an integer, one would
get:

;; This buffer is for notes you don't want to save, and for Lisp
   evaluation.
;; If you want to create a file, visit that file with C-x C-f,
;; then enter the text in that file's own buffer.
12345
exp

It would seem from the docstring of `unexpand-abbrev' that it is still
supposed to work orrectly after some minor editing of the type
described above:

Undo the expansion of the last abbrev that expanded.
This differs from ordinary undo in that other editing done since then
is not undone.

Sincerely,

Luc.

Re: last-abbrev-location

[Luc Teirlinck]

>From my previous message:

   Unlike `abbrev-start-location', which is a marker,
   `last-abbrev-location' is just a buffer position, an integer.  Is
   there a reason for this?  It seems to cause problems.

Maybe, on second thought, it might give compatibility problems if one
tried to change this now.  Maybe `unexpand-abbrev' is only intended to
be used after having typed a few extra characters _after_ the epanded
abbrev, in which case the problems described in my previous message do
not occur.  Maybe the only problem is that its docstring does not make
that sufficiently clear.

Sincerely,

Luc.

Re: last-abbrev-location

[Richard Stallman]

    Maybe, on second thought, it might give compatibility problems if one
    tried to change this now.

What compatibility problems?  That variable is obscure; changing it
is ok if it fixes something.

Re: last-abbrev-location

[Luc Teirlinck]

Richard Stallman wrote:

   What compatibility problems?  That variable is obscure; changing it
   is ok if it fixes something.

It would require some accompanying changes in idlwave.el, but those
would not appear to be very extensive.

Sincerely,

Luc.