Re: Characterset for Abbrevation names limited

Re: Characterset for Abbrevation names limited

[Glenn Morris]

Richard Stallman wrote (on Mon, 25 Jun 2007 at 15:53 -0400):

>     > and/or include a similar comment into the function description.
>     > Otherwise an uninformed user might use the unsupported characters like 
>     > [!"#%&/] and register the failure of the function and
>     > interpret it as a bug.
> 
> I suggest making that function give an error if the abbrev contains
> invalid characters.
> 
> Would you like to do that?

I'm going on holiday for a couple of weeks, I can look at it when I
return. Someone else might wish to do it before then.

(This is about define-global-abbrev, continued from bug-gnu-emacs).

Re: Characterset for Abbrevation names limited

[Stefan Monnier]

>> > and/or include a similar comment into the function description.
>> > Otherwise an uninformed user might use the unsupported characters like 
>> > [!"#%&/] and register the failure of the function and
>> > interpret it as a bug.
>> 
>> I suggest making that function give an error if the abbrev contains
>> invalid characters.

I can't recover the original email so I'm not sure what was suggested, but
if the suggestion is to make define-abbrev signal an error if the abbrev
uses chars that are not word constituents, this will be problematic because
the criterion depends on the syntax-table in use and that one may not be the
same during define-abbrev as during expand-abbrev (some code in Emacs even
uses pre-abbrev-expand-hook to change the syntax-table used during
expand-abbrev independently from the syntax-table used otherwise in the
buffer).


        Stefan

Re: Characterset for Abbrevation names limited

[Glenn Morris]

Stefan Monnier wrote:

> I can't recover the original email so I'm not sure what was suggested,

The parent was off-list (for some reason...), but I quoted all of it.
The rest of the thread is in bug-gnu-emacs with the same subject.

> but if the suggestion is to make define-abbrev signal an error if
> the abbrev uses chars that are not word constituents, this will be
> problematic because the criterion depends on the syntax-table in use
> and that one may not be the same during define-abbrev as during
> expand-abbrev (some code in Emacs even uses pre-abbrev-expand-hook
> to change the syntax-table used during expand-abbrev independently
> from the syntax-table used otherwise in the buffer).

Yes, I did think about the define-abbrev case, and didn't see how one
could do much because of the reasons you cite. I think the suggestion
was perhaps just for define-global-abbrev, either to make it error or
warn if using a "non-standard-word-constituent" ([^a-zA-Z0-9] ?),
probably only when called interactively. As it stands you can define
"global" abbrevs that only work in some modes, which is a bit odd. But
I guess one could always construct a mode with a suitably bizarre
syntax that would still mess things up...

Re: Characterset for Abbrevation names limited

[Richard Stallman]

    I can't recover the original email so I'm not sure what was suggested, but
    if the suggestion is to make define-abbrev signal an error if the abbrev
    uses chars that are not word constituents, this will be problematic because
    the criterion depends on the syntax-table in use and that one may not be the
    same during define-abbrev as during expand-abbrev (some code in Emacs even
    uses pre-abbrev-expand-hook to change the syntax-table used during
    expand-abbrev independently from the syntax-table used otherwise in the
    buffer).

We could leave define-abbrev alone and make the user level commands,
define-global-abbrev and define-mode-abbrev, do this check.

That way, ordinary users would be protected from the mistake,
and experts could still do whatever they wish.