bug#5936: 23.1.95; fix a bug in abbrev--before-point

bug#5936: 23.1.95; fix a bug in abbrev--before-point

[Leo]

If you use the :regexp property in an abbrev table, all abbrevs in that
table will fail except those with only one char. This is because of a
regexp match in abbrev--before-point which only looks back one char.

BTW, the doc string of looking-back seems irrelevant if by default it
only looks back one char, that can't be slower than looking-at.

diff --git a/lisp/abbrev.el b/lisp/abbrev.el
index b72bdbb..d3b95cb 100644
--- a/lisp/abbrev.el
+++ b/lisp/abbrev.el
@@ -681,7 +681,7 @@ then ABBREV is looked up in that table only."
                              (setq start (point))
                              (forward-word 1)
                              (setq end (min (point) lim)))
-                         (when (looking-back re (line-beginning-position))
+                         (when (looking-back re (line-beginning-position) t)
                            (setq start (match-beginning 1))
                            (setq end   (match-end 1)))))
                      (setq name  (buffer-substring start end))

bug#5936: 23.1.95; fix a bug in abbrev--before-point

[Stefan Monnier]

> If you use the :regexp property in an abbrev table, all abbrevs in that
> table will fail except those with only one char.

Not at all.  But you may get such a result if you use a naive regexp, yes.

> This is because of a regexp match in abbrev--before-point which only
> looks back one char.

No, this is because the regexp you have to use must match part of the
text *before* the abbrev.  That's why the doc for define-abbrev-table
gives as example \<\(\w+\)\W* rather than just \(\w+\)\W*.


        Stefan

bug#5936: 23.1.95; fix a bug in abbrev--before-point

[Leo]

On 2010-04-12 19:25 +0100, Stefan Monnier wrote:
>> If you use the :regexp property in an abbrev table, all abbrevs in that
>> table will fail except those with only one char.
>
> Not at all.  But you may get such a result if you use a naive regexp, yes.
>
>> This is because of a regexp match in abbrev--before-point which only
>> looks back one char.
>
> No, this is because the regexp you have to use must match part of the
> text *before* the abbrev.  That's why the doc for define-abbrev-table
> gives as example \<\(\w+\)\W* rather than just \(\w+\)\W*.

Thank you for the clarification.

The bug can be closed.

>         Stefan

Leo