syntax table entries for comments

syntax table entries for comments

[Arjan Bos]

Hi all,

First of, many thanks to those who answered my question about info for 
major-mode writers. The mode tutorial on emacs wiki was most 
enlightning. In fact it was so helpful, that I decided to completely 
rewrite the major-mode for NetRexx. When it's done I'll post it in 
gnu.emacs.sources.

But before that can happen, I'll need some clues. NetRexx is a 
programming language that contains two types of comments. The first is 
for large blocks and is the familiar /* */ construction. This I can do 
by looking at the examples. The second one is a single-line comment that 
consists of two hyphens, like:
   -- this is a comment.

Based on the major mode tutorial I did put the following code in:

(defvar nrx-mode-syntax-table nil
   "Syntax table in use in NRX-mode buffers.")

(defun nrx-create-syntax-table ()
   (if nrx-mode-syntax-table
       ()
     (setq nrx-mode-syntax-table (make-syntax-table))
     (modify-syntax-entry ?. "." nrx-mode-syntax-table)
     (modify-syntax-entry ?- ". 12b" nrx-mode-syntax-table)
     (modify-syntax-entry ?/ ". 14" nrx-mode-syntax-table)
     (modify-syntax-entry ?* ". 23" nrx-mode-syntax-table)
     (modify-syntax-entry ?\n "> b" nrx-mode-syntax-table)
     (modify-syntax-entry ?\' "\"" nrx-mode-syntax-table))

   (set-syntax-table nrx-mode-syntax-table))

This works but also renders the combination -* and *- as comment start 
and end, which is wrong. Could anyone please tell me what I'm missing?

TIA,

Arjan

Re: syntax table entries for comments

[Stefan Monnier]

> language that contains two types of comments. The first is for large blocks
> and is the familiar /* */ construction. This I can do by looking at the
> examples. The second one is a single-line comment that consists of two
> hyphens, like:
>    -- this is a comment.

Emacs does not properly support this, as you've discovered.

> (defvar nrx-mode-syntax-table nil
>    "Syntax table in use in NRX-mode buffers.")

> (defun nrx-create-syntax-table ()
>    (if nrx-mode-syntax-table
>        ()
>      (setq nrx-mode-syntax-table (make-syntax-table))
>      (modify-syntax-entry ?. "." nrx-mode-syntax-table)
>      (modify-syntax-entry ?- ". 12b" nrx-mode-syntax-table)
>      (modify-syntax-entry ?/ ". 14" nrx-mode-syntax-table)
>      (modify-syntax-entry ?* ". 23" nrx-mode-syntax-table)
>      (modify-syntax-entry ?\n "> b" nrx-mode-syntax-table)
>      (modify-syntax-entry ?\' "\"" nrx-mode-syntax-table))

>    (set-syntax-table nrx-mode-syntax-table))

Could you tell me the place from which this code was inspired so we can
fix it ?  It should look like:

   (defvar nrx-mode-syntax-table
     (let ((st (make-syntax-table)))
       (modify-syntax-entry ...)
       (modify-syntax-entry ...)
       ...
       st))

and the `set-syntax-table' is commonly done implicitly by
`define-derived-mode'.

> This works but also renders the combination -* and *- as comment start and
> end, which is wrong. Could anyone please tell me what I'm missing?

Nothing, really, other than the fact that it's a limitation of current
syntax-tables.  You can either hack on src/syntax.c to add support for
such cases, or use font-lock-syntactic-keywords to recognize `--'
and mark it as a comment starter.


        Stefan

Re: syntax table entries for comments

[Arjan Bos]

Stefan Monnier wrote:

>> The second one is a single-line comment that consists of two
>>hyphens, like:
>>   -- this is a comment.
> 
> 
> Emacs does not properly support this, as you've discovered.
To bad. At least I won't have to look for it in the docs anymore.

> 
> 
>>(defvar nrx-mode-syntax-table nil
>>   "Syntax table in use in NRX-mode buffers.")
> 
> 
>>(defun nrx-create-syntax-table ()
>>   (if nrx-mode-syntax-table
>>       ()
>>     (setq nrx-mode-syntax-table (make-syntax-table))
>>     (modify-syntax-entry ?. "." nrx-mode-syntax-table)
>>     (modify-syntax-entry ?- ". 12b" nrx-mode-syntax-table)
>>     (modify-syntax-entry ?/ ". 14" nrx-mode-syntax-table)
>>     (modify-syntax-entry ?* ". 23" nrx-mode-syntax-table)
>>     (modify-syntax-entry ?\n "> b" nrx-mode-syntax-table)
>>     (modify-syntax-entry ?\' "\"" nrx-mode-syntax-table))
> 
> 
>>   (set-syntax-table nrx-mode-syntax-table))
> 
> 
> Could you tell me the place from which this code was inspired so we can
> fix it ?  It should look like:
> 
>    (defvar nrx-mode-syntax-table
>      (let ((st (make-syntax-table)))
>        (modify-syntax-entry ...)
>        (modify-syntax-entry ...)
>        ...
>        st))
> 
> and the `set-syntax-table' is commonly done implicitly by
> `define-derived-mode'.

I found it via Emacs Wiki. On page 
http://www.emacswiki.org/cgi-bin/wiki/ModeTutorial, they link to the 
following mode tutorial:

http://two-wugs.net/emacs/mode-tutorial.html
which I followed and used as an inspiration.

Could you please enlighten me as to why your way is better? I'm no 
(e)lisp expert, but I'm doing allright with a bit of voodoo programming[1].
> 
> 
>>This works but also renders the combination -* and *- as comment start and
>>end, which is wrong. Could anyone please tell me what I'm missing?
> 
> 
> Nothing, really, other than the fact that it's a limitation of current
> syntax-tables.  You can either hack on src/syntax.c to add support for
> such cases, or use font-lock-syntactic-keywords to recognize `--'
> and mark it as a comment starter.
I tried the latter, but then strings within the `--' comment will undo 
the comment highlighting. So I'll try to voodoo hack src/syntax.c


Thanks a lot,

Arjan Bos

[1] voodoo programming: Change something and see if it works. Keep on 
changing things until it does what you want.
This way I can change most sources with only a minimum of knowledge 
about the language it's written in, by repeating the constructs already 
used. Downside is of course that you don't learn new language 
constructs. ;-)

Re: syntax table entries for comments

[Stefan Monnier]

>>> (defvar nrx-mode-syntax-table nil
>>> "Syntax table in use in NRX-mode buffers.")
>> 
>>> (defun nrx-create-syntax-table ()
>>> (if nrx-mode-syntax-table
>>> ()
>>> (setq nrx-mode-syntax-table (make-syntax-table))
>>> (modify-syntax-entry ?. "." nrx-mode-syntax-table)
>>> (modify-syntax-entry ?- ". 12b" nrx-mode-syntax-table)
>>> (modify-syntax-entry ?/ ". 14" nrx-mode-syntax-table)
>>> (modify-syntax-entry ?* ". 23" nrx-mode-syntax-table)
>>> (modify-syntax-entry ?\n "> b" nrx-mode-syntax-table)
>>> (modify-syntax-entry ?\' "\"" nrx-mode-syntax-table))
>> 
>>> (set-syntax-table nrx-mode-syntax-table))
>> Could you tell me the place from which this code was inspired so we can
>> fix it ?  It should look like:
>> (defvar nrx-mode-syntax-table
>> (let ((st (make-syntax-table)))
>> (modify-syntax-entry ...)
>> (modify-syntax-entry ...)
>> ...
>> st))
>> and the `set-syntax-table' is commonly done implicitly by
>> `define-derived-mode'.

> I found it via Emacs Wiki. On page
> http://www.emacswiki.org/cgi-bin/wiki/ModeTutorial, they link to the
> following mode tutorial:

> http://two-wugs.net/emacs/mode-tutorial.html
> which I followed and used as an inspiration.

> Could you please enlighten me as to why your way is better? I'm no (e)lisp
> expert, but I'm doing allright with a bit of voodoo programming[1].

Advantages are:
- shorter.
- less memory used since the code that sets up the table can be discarded
  after the table is setup, whereas in your case, the function
  nrx-create-syntax-table needs to be kept around in case someone wants
  to call it.
- no temporary stage where nrx-mode-syntax-table holds a value that is
  not a syntax-table (or that is a syntax-table but that is not yet
  properly initialized): either it's there and initialized or it's not
  there at all.

>>> This works but also renders the combination -* and *- as comment start and
>>> end, which is wrong. Could anyone please tell me what I'm missing?
>> Nothing, really, other than the fact that it's a limitation of current
>> syntax-tables.  You can either hack on src/syntax.c to add support for
>> such cases, or use font-lock-syntactic-keywords to recognize `--'
>> and mark it as a comment starter.
> I tried the latter, but then strings within the `--' comment will undo the
> comment highlighting.

I suspect you used font-lock-keywords rather than
font-lock-syntactic-keywords.

> So I'll try to voodoo hack src/syntax.c

That would be wonderful.


        Stefan

Re: syntax table entries for comments

[Arjan Bos]

Stefan Monnier wrote:

<snip>code snippet</snip>

>>Could you please enlighten me as to why your way is better? I'm no (e)lisp
>>expert, but I'm doing allright with a bit of voodoo programming[1].
> 
> 
> Advantages are:
> - shorter.
> - less memory used since the code that sets up the table can be discarded
>   after the table is setup, whereas in your case, the function
>   nrx-create-syntax-table needs to be kept around in case someone wants
>   to call it.
> - no temporary stage where nrx-mode-syntax-table holds a value that is
>   not a syntax-table (or that is a syntax-table but that is not yet
>   properly initialized): either it's there and initialized or it's not
>   there at all.

Sounds good to me, thanks!

> 
> 
<snip/>

>>
>>I tried the latter, but then strings within the `--' comment will undo the
>>comment highlighting.
> 
> 
> I suspect you used font-lock-keywords rather than
> font-lock-syntactic-keywords.
You're right, I'll look up the docs for that then.

> 
> 
>>So I'll try to voodoo hack src/syntax.c
> 
> 
> That would be wonderful.
> 
Don't get your hopes up to high yet. The amount of constants in syntax.c 
and my lack of c knowlegde will make this hard. (Were it COBOL though... 
:-). I do happen to have a fairly recent CVS snapshot though, so I'll 
try it.

Arjan