Re: set-fontset-font is slow if the fontset is specified by a long name

[Kenichi Handa]

In article <address@hidden>, YAMAMOTO Mitsuharu <address@hidden> writes:

>>>>>>  On Mon, 25 Oct 2004 16:25:16 +0900 (JST), Kenichi Handa 
>>>>>> <address@hidden> said:
>>>  How about consulting Vfontset_table literally before doing regexp
>>>  matches?

>>  Such a change may make what matches with NAME different from the
>>  current code (though only in vary rare case such as you have a
>>  fontset "-adove-courier-*-*-*-*-*-*-*-*-*-*-fontset-default").

> So a fontset is always specified indirectly, via pattern or alias,
> then.

Yes, at least in set-fontset-font.

>>>  The scenario I'm considering is something like the following:

>>>  (let ((fontset (create-fontset-from-ascii-font "7x14")))
>>>    (while ...
>>>      (set-fontset-font fontset ...)
>>>      ...)
>>>    (set-frame-font fontset))

>>  Isn't the latest code fast enough?

> Yes, it's way faster now.  Thanks.

> Before your change, I've used short names to avoid the slow execution
> of set-fontset-font:

>    (if (not (string-match "^fontset-" fontset))
>        (setq fontset
>            (concat "fontset-" (aref (x-decompose-font-name fontset)
>                                     xlfd-regexp-encoding-subnum))))

> I thought this might specify another fontset that has the same alias
> entry, and the long format could identify the fontset directly.

Unfortunately even the long format can't identify the
fontset directly.  I admit that this is a mistake of the
design.  The basic problem is that a fontset name can
contain '*'.  But, fixing that requires rather big changes,
and I'd like to postpone it until the next release as far as
the recent change gives reasonable speed of processing.

---
Ken'ichi HANDA
address@hidden