defface/defcustom question

defface/defcustom question

[Jesper Harder]

In some places Gnus faces are defined both with `defface' _and_
`defcustom'.  E.g.

,----
| (defface gnus-server-opened-face
|   '((((class color) (background light)) (:foreground "Green3" :bold t))
|     (((class color) (background dark)) (:foreground "Green1" :bold t))
|     (t (:bold t)))
|   "Face used for displaying OPENED servers"
|   :group 'gnus-server-visual)
| 
| [...]
| 
| (defcustom gnus-server-opened-face 'gnus-server-opened-face
|   "Face name to use on OPENED servers."
|   :group 'gnus-server-visual
|   :type 'face)
`----

This means that the face shows up twice when you use Customize.  It
looks like a mistake to me ... or is there a reason for doing it?

Re: defface/defcustom question

[Simon Josefsson]

Jesper Harder <harder@myrealbox.com> writes:

> In some places Gnus faces are defined both with `defface' _and_
> `defcustom'.  E.g.
>
> ,----
> | (defface gnus-server-opened-face
> |   '((((class color) (background light)) (:foreground "Green3" :bold t))
> |     (((class color) (background dark)) (:foreground "Green1" :bold t))
> |     (t (:bold t)))
> |   "Face used for displaying OPENED servers"
> |   :group 'gnus-server-visual)
> | 
> | [...]
> | 
> | (defcustom gnus-server-opened-face 'gnus-server-opened-face
> |   "Face name to use on OPENED servers."
> |   :group 'gnus-server-visual
> |   :type 'face)
> `----
>
> This means that the face shows up twice when you use Customize.  It
> looks like a mistake to me ... or is there a reason for doing it?

Aren't variables and faces different things?  The variable identifies
the name of the face to use, and the face define the actual face.  An
extra indirection.  Compare the ':type face' documentation:

`face'
     The value must be a symbol which is a face name, and you can do
     completion with `M-<TAB>'.

Re: defface/defcustom question

[Kai Großjohann]

Simon Josefsson <jas@extundo.com> writes:

> Aren't variables and faces different things?  The variable identifies
> the name of the face to use, and the face define the actual face.  An
> extra indirection.

One might ask: why the indirection?  Maybe one part of the answer is
that variables can be buffer-local, whereas faces, afaik, can't.
-- 
Two cafe au lait please, but without milk.

Re: defface/defcustom question

[Simon Josefsson]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> Aren't variables and faces different things?  The variable identifies
>> the name of the face to use, and the face define the actual face.  An
>> extra indirection.
>
> One might ask: why the indirection?  Maybe one part of the answer is
> that variables can be buffer-local, whereas faces, afaik, can't.

It also makes it possible to express "use the same face for this, that
and this" and then later be able to change the face only once to
change it in all places.  And it simplifies font-lock stuff, compare
gnus-server-font-lock-keywords.

Re: defface/defcustom question

[Kai Großjohann]

Simon Josefsson <jas@extundo.com> writes:

> It also makes it possible to express "use the same face for this, that
> and this" and then later be able to change the face only once to
> change it in all places.

But that's possible with regular faces already?  Just use the same
symbol (face name) in all those places?

> And it simplifies font-lock stuff, compare
> gnus-server-font-lock-keywords.

I looked at it.  It looks simple.  But I can't see how variables made
it simpler.

-- 
Two cafe au lait please, but without milk.

Re: defface/defcustom question

[Simon Josefsson]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> It also makes it possible to express "use the same face for this, that
>> and this" and then later be able to change the face only once to
>> change it in all places.
>
> But that's possible with regular faces already?  Just use the same
> symbol (face name) in all those places?

You mean changing gnus source?  I meant for users.

>> And it simplifies font-lock stuff, compare
>> gnus-server-font-lock-keywords.
>
> I looked at it.  It looks simple.  But I can't see how variables made
> it simpler.

font-lock-keywords uses face names, not faces.  Compare
info://Elisp/Faces%20for%20Font%20Lock

Re: defface/defcustom question

[Jesper Harder]

[-- Attachment #1: Type: text/plain, Size: 1069 bytes --]

Simon Josefsson <jas@extundo.com> writes:

> kai.grossjohann@gmx.net (Kai Großjohann) writes:
>
>> Simon Josefsson <jas@extundo.com> writes:
>>
>>> It also makes it possible to express "use the same face for this, that
>>> and this" and then later be able to change the face only once to
>>> change it in all places.
>>
>> But that's possible with regular faces already?  Just use the same
>> symbol (face name) in all those places?
>
> You mean changing gnus source?  I meant for users.

I think it's harder to understand for users.  It's confusing to have
two customization options with the same name (and very similar
purpose).

>>> And it simplifies font-lock stuff, compare
>>> gnus-server-font-lock-keywords.
>>
>> I looked at it.  It looks simple.  But I can't see how variables made
>> it simpler.
>
> font-lock-keywords uses face names, not faces.  Compare
> info://Elisp/Faces%20for%20Font%20Lock

I don't see your point.  This just means that you need to quote the
name of the face -- I thinks that's a lot simpler than an extra layer
of indirection.  E.g.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: gnus-srvr.el.diff --]
[-- Type: text/x-patch, Size: 1598 bytes --]

--- gnus/lisp/gnus-srvr.el	Tue Jun 24 07:22:43 2003
+++ cvsgnus/lisp/gnus-srvr.el	Sat Aug  9 15:50:22 2003
@@ -202,38 +202,13 @@
   "Face used for displaying OFFLINE servers"
   :group 'gnus-server-visual)
 
-(defcustom gnus-server-agent-face 'gnus-server-agent-face
-  "Face name to use on AGENTIZED servers."
-  :group 'gnus-server-visual
-  :type 'face)
-
-(defcustom gnus-server-opened-face 'gnus-server-opened-face
-  "Face name to use on OPENED servers."
-  :group 'gnus-server-visual
-  :type 'face)
-
-(defcustom gnus-server-closed-face 'gnus-server-closed-face
-  "Face name to use on CLOSED servers."
-  :group 'gnus-server-visual
-  :type 'face)
-
-(defcustom gnus-server-denied-face 'gnus-server-denied-face
-  "Face name to use on DENIED servers."
-  :group 'gnus-server-visual
-  :type 'face)
-
-(defcustom gnus-server-offline-face 'gnus-server-offline-face
-  "Face name to use on OFFLINE servers."
-  :group 'gnus-server-visual
-  :type 'face)
-
 (defvar gnus-server-font-lock-keywords
   (list
-   '("(\\(agent\\))" 1 gnus-server-agent-face)
-   '("(\\(opened\\))" 1 gnus-server-opened-face)
-   '("(\\(closed\\))" 1 gnus-server-closed-face)
-   '("(\\(offline\\))" 1 gnus-server-offline-face)
-   '("(\\(denied\\))" 1 gnus-server-denied-face)))
+   '("(\\(agent\\))" 1 'gnus-server-agent-face)
+   '("(\\(opened\\))" 1 'gnus-server-opened-face)
+   '("(\\(closed\\))" 1 'gnus-server-closed-face)
+   '("(\\(offline\\))" 1 'gnus-server-offline-face)
+   '("(\\(denied\\))" 1 'gnus-server-denied-face)))
 
 (defun gnus-server-mode ()
   "Major mode for listing and editing servers.

Re: defface/defcustom question

[Alan Shutko]

Simon Josefsson <jas@extundo.com> writes:

> You mean changing gnus source?  I meant for users.

Well, users can customize all the Gnus faces to inherit from the same
face.

-- 
Alan Shutko <ats@acm.org> - I am the rocks.
Last night our sex was so good, the neighbors lit cigarettes.

Re: defface/defcustom question

[Simon Josefsson]

Alan Shutko <ats@acm.org> writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> You mean changing gnus source?  I meant for users.
>
> Well, users can customize all the Gnus faces to inherit from the same
> face.

True.  Hm.  Using both a variable and a face seem to be common
practice though, I wonder if there are other reasons.  Maybe face
inheritance is a recent feature?  I dunno.

Re: defface/defcustom question

[Alan Shutko]

Simon Josefsson <jas@extundo.com> writes:

> True.  Hm.  Using both a variable and a face seem to be common
> practice though, I wonder if there are other reasons.  Maybe face
> inheritance is a recent feature?  I dunno.

I believe it is.  The custom attribute is only from 2000, and I know
these faces are a lot older than that.  I don't know if there's a big
reason to change it, but it might make M-x customize-group gnus a bit
cleaner.  


-- 
Alan Shutko <ats@acm.org> - I am the rocks.
extraneous text! ... You may find (as I have) that Ron Janorkar's Tag-X

Re: defface/defcustom question

[Kai Großjohann]

Simon Josefsson <jas@extundo.com> writes:

> kai.grossjohann@gmx.net (Kai Großjohann) writes:
>
>> Simon Josefsson <jas@extundo.com> writes:
>>
>>> It also makes it possible to express "use the same face for this, that
>>> and this" and then later be able to change the face only once to
>>> change it in all places.
>>
>> But that's possible with regular faces already?  Just use the same
>> symbol (face name) in all those places?
>
> You mean changing gnus source?  I meant for users.

I don't understand.  The default font-lock setup is to use different
faces for all the elements in the server buffer.  If you want to
change the look of the faces, set-face-foreground and friends seem to
be sufficient.

Obviously, they can be used to make different faces look the same.

And equally obviously, if font-lock-keywords specifies the same face
for construct A and B, then they will always look the same,
regardless of whether a variable is used or a face is used
directly -- a variable can only have one value after all...

I think I'm misunderstanding you completely; I hope the above helps
to show you what my weak mind is thinking so that you can guide me
out of the fog.  I apologize for being so dense :-/
-- 
Two cafe au lait please, but without milk.

Re: defface/defcustom question

[Kai Großjohann]

Alan Shutko <ats@acm.org> writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> True.  Hm.  Using both a variable and a face seem to be common
>> practice though, I wonder if there are other reasons.  Maybe face
>> inheritance is a recent feature?  I dunno.
>
> I believe it is.  The custom attribute is only from 2000, and I know
> these faces are a lot older than that.  I don't know if there's a big
> reason to change it, but it might make M-x customize-group gnus a bit
> cleaner.  

IIRC, font-lock was changed several times between using variables and
not using variables.  I have no idea, why.
-- 
Two cafe au lait please, but without milk.

Re: defface/defcustom question

[Simon Josefsson]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> kai.grossjohann@gmx.net (Kai Großjohann) writes:
>>
>>> Simon Josefsson <jas@extundo.com> writes:
>>>
>>>> It also makes it possible to express "use the same face for this, that
>>>> and this" and then later be able to change the face only once to
>>>> change it in all places.
>>>
>>> But that's possible with regular faces already?  Just use the same
>>> symbol (face name) in all those places?
>>
>> You mean changing gnus source?  I meant for users.
>
> I don't understand.  The default font-lock setup is to use different
> faces for all the elements in the server buffer.  If you want to
> change the look of the faces, set-face-foreground and friends seem to
> be sufficient.
>
> Obviously, they can be used to make different faces look the same.
>
> And equally obviously, if font-lock-keywords specifies the same face
> for construct A and B, then they will always look the same,
> regardless of whether a variable is used or a face is used
> directly -- a variable can only have one value after all...
>
> I think I'm misunderstanding you completely; I hope the above helps
> to show you what my weak mind is thinking so that you can guide me
> out of the fog.  I apologize for being so dense :-/

You are right in all that, and I wasn't clear.  Consider a user that
customizes some face to RED, and then wants the server buffer to use
that face for everything.  She can then customize all the server
buffers to have a RED face.  But, the point of having a variable would
be if the user then changes her mind on the original RED face, and
wants it to be BLUE.  Initially, instead of changing the faces, she
would only have changed the server buffer face names to point at the
original RED face name.  So if she changes the original RED face to
BLUE, then all server buffer faces also changes to BLUE.

Now it seems customize-face supports inheritance, so this can be
achieved anyway, but perhaps that feature didn't exist when the face
name idea was implemented.

So I don't know what the real reason for still supporting face names
is.  Backwards compatibility?  Does XEmacs support face inheritance?
Maybe we should ask on emacs-devel, if nobody else here knows, since
several other parts of Emacs also uses face names.

Re: defface/defcustom question

[Kai Großjohann]

Simon Josefsson <jas@extundo.com> writes:

> You are right in all that, and I wasn't clear.  Consider a user that
> customizes some face to RED, and then wants the server buffer to use
> that face for everything.  She can then customize all the server
> buffers to have a RED face.  But, the point of having a variable would
> be if the user then changes her mind on the original RED face, and
> wants it to be BLUE.  Initially, instead of changing the faces, she
> would only have changed the server buffer face names to point at the
> original RED face name.  So if she changes the original RED face to
> BLUE, then all server buffer faces also changes to BLUE.

Ah, I see.  An extra level of indirection can be good, I see now.
-- 
Two cafe au lait please, but without milk.

Re: defface/defcustom question

[Alex Schroeder]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> Ah, I see.  An extra level of indirection can be good, I see now.

Well, it often feels like overengineering to me.  People really
understand face customization.

Consider how I customize the look of things.  I look at the buffer
and notice an ugly piece of text.  I try to discover the face used at
that point.  And then I customize it.  Straightforward. 

Now, how would I find out about the variable used for the indirection
we have?  I believe I would have to read the source for that.

Therefore I think that indirection via variables used to be a way to
keep the number of faces small.  Emacs Example: apropos-label-face's
value is italic.  In order to avoid creating a new apropos-label-face,
the author decided to reuse an existing face.  The same is true for
dired.  dired-font-lock-keywords contains lots of references to plain
font-lock faces such as font-lock-type-face, font-lock-constant-face,
etc.

The alternative is to create many new faces for each "item".  This is
why XEmacs has lots of dired faces.

So only one small unexplained item remains.  Why are there also variables
called font-lock-type-face and similar?  I think they are useless.
The reason is this: You cannot use M-x customize-option to change its
value.  You can use M-x customize-face to customize font-lock faces.

Therefore, having a variable font-lock-type-face pointing to the face
font-lock-type-face must be something old -- kept for backwards
compatibility.

For users, there should be only one thing to change.  And based on
how font-lock-type-face works, I suggest we allow users to customize
faces, not set variables.  This is also how I personally customize
Emacs, as I said at the beginning.

If at all, I therefore propose to remove the indirections via
variables.

Alex.
-- 
http://www.emacswiki.org/alex/
There is no substitute for experience.

Re: defface/defcustom question

[Per Abrahamsen]

Alan Shutko <ats@acm.org> writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> True.  Hm.  Using both a variable and a face seem to be common
>> practice though, I wonder if there are other reasons.  Maybe face
>> inheritance is a recent feature?  I dunno.
>
> I believe it is.  The custom attribute is only from 2000, and I know
> these faces are a lot older than that. 

Faces are oldest, customize in between, and face inheritance is
newest.

Emacs never really been consistent about whether to use variables to
hold the faces, but before customize there was a great incentive to do
so, as the pre-customize face API sucked.

Often one would just do

(defvar some-package-face 'bold
  "Face used by some package.")

in order to avoid the problem of creating a face.  If the user wanted
some package to use italic instead, he'd 

(setq some-package-face 'italic)

It had a nice feature that if bold looked ugly with his favorite face,
he could change the bold face to something else, and many occurrences
of bold in Emacs would adjust.

It is probably simpler to just use a defface for new options these
days, but if we remove the old variables we will break user
customizations with setq like the above.