hijack C function, use it transparently with in-between layer

hijack C function, use it transparently with in-between layer

[Emanuel Berg]

The truth is down there!

(defvar old-message)
(fset 'old-message (symbol-function #'message))

(defun message (format-string &rest args)
  (unless (member format-string
                  '(
                    "It was just another system of control."
                    "Mark set"
                    ))
      (apply `(old-message ,format-string ,@args) )))

;; test:
;;
;;  (message "%s - %s" "1" "2")
;;  (message "Break on thru to the other side.")
;;  (message "It was just another system of control.")
;;  (message "Mark set")

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Marcin Borkowski]

On 2016-06-06, at 05:28, Emanuel Berg <embe8573@student.uu.se> wrote:

> The truth is down there!
>
> (defvar old-message)
> (fset 'old-message (symbol-function #'message))
>
> (defun message (format-string &rest args)
>   (unless (member format-string
>                   '(
>                     "It was just another system of control."
>                     "Mark set"
>                     ))
>       (apply `(old-message ,format-string ,@args) )))

Smart!  Tho not what I'd call "good style". ;-)

Best,

-- 
Marcin Borkowski
http://octd.wmi.amu.edu.pl/en/Marcin_Borkowski
Faculty of Mathematics and Computer Science
Adam Mickiewicz University

Re: hijack C function, use it transparently with in-between layer

[Barry Margolin]

In article <868tyjnghn.fsf@student.uu.se>,
 Emanuel Berg <embe8573@student.uu.se> wrote:

> The truth is down there!
> 
> (defvar old-message)
> (fset 'old-message (symbol-function #'message))
> 
> (defun message (format-string &rest args)
>   (unless (member format-string
>                   '(
>                     "It was just another system of control."
>                     "Mark set"
>                     ))
>       (apply `(old-message ,format-string ,@args) )))
> 
> ;; test:
> ;;
> ;;  (message "%s - %s" "1" "2")
> ;;  (message "Break on thru to the other side.")
> ;;  (message "It was just another system of control.")
> ;;  (message "Mark set")

Shouldn't you do this using advice?

-- 
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

Barry Margolin <barmar@alum.mit.edu> writes:

> Shouldn't you do this using advice?

Right, forgot about them...

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Phillip Lord]

Barry Margolin <barmar@alum.mit.edu> writes:

> In article <868tyjnghn.fsf@student.uu.se>,
>  Emanuel Berg <embe8573@student.uu.se> wrote:
>
>> The truth is down there!
>> 
>> (defvar old-message)
>> (fset 'old-message (symbol-function #'message))
>> 
>> (defun message (format-string &rest args)
>>   (unless (member format-string
>>                   '(
>>                     "It was just another system of control."
>>                     "Mark set"
>>                     ))
>>       (apply `(old-message ,format-string ,@args) )))
>> 
>> ;; test:
>> ;;
>> ;;  (message "%s - %s" "1" "2")
>> ;;  (message "Break on thru to the other side.")
>> ;;  (message "It was just another system of control.")
>> ;;  (message "Mark set")
>
> Shouldn't you do this using advice?


Advice only works partially (as I expect this will). The message
function is also called directly in the C layer. Advice will not capture
those cases.

Phil

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

phillip.lord@russet.org.uk (Phillip Lord)
writes:

> Advice only works partially (as I expect this
> will). The message function is also called
> directly in the C layer. Advice will not
> capture those cases.

Good point, any ideas how to get to that
as well?

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Stefan Monnier]

>> Advice only works partially (as I expect this
>> will). The message function is also called
>> directly in the C layer. Advice will not
>> capture those cases.
> Good point, any ideas how to get to that as well?

Report the cases you bump into as bugs.


        Stefan

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

Stefan Monnier <monnier@iro.umontreal.ca>
writes:

>> Good point, any ideas how to get to that
>> as well?
>
> Report the cases you bump into as bugs.

You mean there should be no C to C calls for
stuff that are in Elisp as well?

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Stefan Monnier]

>>> Good point, any ideas how to get to that as well?
>> Report the cases you bump into as bugs.
> You mean there should be no C to C calls for
> stuff that are in Elisp as well?

Just because you report it as a bug doesn't mean the maintainers will
fix it or agree that it should be fixed.  There will sometimes be
overriding concerns.  I think the main 2 reasons why a C-to-C call
should not be replaced by a call "via Elisp" are:
- the call is made in a context where it's unsafe to run arbitrary Elisp
  code (e.g. GC is not allowed, or something like that).
- the performance cost would be too high compared to the benefit of the
  added flexibility.

So whether the code can be changed or not will depend on the
specific cases.  Hence report them as bugs and so the maintainers can
look into it.


        Stefan

Re: hijack C function, use it transparently with in-between layer

[Barry Margolin]

In article <mailman.974.1465221532.1216.help-gnu-emacs@gnu.org>,
 Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> >>> Good point, any ideas how to get to that as well?
> >> Report the cases you bump into as bugs.
> > You mean there should be no C to C calls for
> > stuff that are in Elisp as well?
> 
> Just because you report it as a bug doesn't mean the maintainers will
> fix it or agree that it should be fixed.  There will sometimes be
> overriding concerns.  I think the main 2 reasons why a C-to-C call
> should not be replaced by a call "via Elisp" are:
> - the call is made in a context where it's unsafe to run arbitrary Elisp
>   code (e.g. GC is not allowed, or something like that).
> - the performance cost would be too high compared to the benefit of the
>   added flexibility.
> 
> So whether the code can be changed or not will depend on the
> specific cases.  Hence report them as bugs and so the maintainers can
> look into it.

Is that a bug or a feature request? The documentation for advice 
specifically warns that it doesn't have any effect on calls to built-in 
functions from the C layer.

-- 
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***

Re: hijack C function, use it transparently with in-between layer

[Stefan Monnier]

> Is that a bug or a feature request?

Both go through M-x report-emacs-bug anyway, so it doesn't matter.


        Stefan

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

Barry Margolin <barmar@alum.mit.edu> writes:

> Is that a bug or a feature request?

The reason I thought this would be cool to have
is exactly the messaging. I thought it could be
cool to drop certain messages I never care for
anyway (e.g., "Mark set", "Auto-saving...done",
etc.).

I also thought it could be used in Elisp to not
have the message buffer overpopulated. It could
be made into yet another "code scope macro"
like `with-current-buffer' only this time

    (with-messages-inhibited
      (do-stuff-1)
      ;; ...
      (do-stuff-n) )

And then - engage :)

Perhaps it can still be used that way unless
there are two many of them messages coming from
the C layer. But messaging to the user should
belong on top of that, mostly, right?
Unless messaging to the user and logging data
are too much intermingled
implementation-wise by careless colleagues...

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Barry Margolin]

In article <8660tl3oo9.fsf@student.uu.se>,
 Emanuel Berg <embe8573@student.uu.se> wrote:

> Barry Margolin <barmar@alum.mit.edu> writes:
> 
> > Is that a bug or a feature request?
> 
> The reason I thought this would be cool to have
> is exactly the messaging. I thought it could be
> cool to drop certain messages I never care for
> anyway (e.g., "Mark set", "Auto-saving...done",
> etc.).

So maybe you should submit a feature request for a message-filter 
variable containing a list of patterns to filter, and/or message-hook 
that can be called to decide whether to show the message.

-- 
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

Barry Margolin <barmar@alum.mit.edu> writes:

> So maybe you should submit a feature request
> for a message-filter variable containing
> a list of patterns to filter, and/or
> message-hook that can be called to decide
> whether to show the message.

Good plan! I'd do it that way as well.

However I just did this in the Emacs24 source
dir:

    grep 'message\(1\|3\)' *.c | wc -l

only 45 hits!

So perhaps it is not a big deal that some
C-to-C calls can't be hindered. I'll use it for
some time and see.

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[John Mastro]

Emanuel Berg <embe8573@student.uu.se> wrote:
> I also thought it could be used in Elisp to not
> have the message buffer overpopulated. It could
> be made into yet another "code scope macro"
> like `with-current-buffer' only this time
>
>    (with-messages-inhibited

This part you can have today:

(defmacro with-messages-inhibited (&rest body)
  (declare (indent defun))
  `(let ((inhibit-message t)
         (message-log-max nil))
     ,@body))

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

John Mastro <john.b.mastro@gmail.com> writes:

>> I also thought it could be used in Elisp to
>> not have the message buffer overpopulated.
>> It could be made into yet another "code
>> scope macro" like `with-current-buffer' only
>> this time (with-messages-inhibited
>
> This part you can have today
>
> (defmacro ...

Cool. "Do it today - in a different way!"

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Phillip Lord]

Emanuel Berg <embe8573@student.uu.se> writes:

> Barry Margolin <barmar@alum.mit.edu> writes:
>
>> Is that a bug or a feature request?
>
> The reason I thought this would be cool to have
> is exactly the messaging. I thought it could be
> cool to drop certain messages I never care for
> anyway (e.g., "Mark set", "Auto-saving...done",
> etc.).

The messaging framework of Emacs is a little under-powered at the
moment -- one which bugs me, for instance, is Emacs has now (lisp level)
ability to write to standard out. Unless it's running in batch, when it
always does so.

It's possible, of course, to write something better, but most packages
use message directly. The ability to filter, suppress and change various
messages would be good. Probably the best way to achieve this would be
to pull message out of C and into lisp.

It's all doable, just takes time. If you want it, why not fork from
master, and start working on it?

Phil

Re: hijack C function, use it transparently with in-between layer

[Eli Zaretskii]

> From: phillip.lord@russet.org.uk (Phillip Lord)
> Date: Tue, 07 Jun 2016 12:12:12 +0100
> 
> The messaging framework of Emacs is a little under-powered at the
> moment -- one which bugs me, for instance, is Emacs has now (lisp level)
> ability to write to standard out.

??? What's wrong with /dev/stdout (as long as you are on a Posix
system)?  As in

   (append-to-file "Hi, there!\n" nil "/dev/stdout")

> Probably the best way to achieve this would be to pull message out
> of C and into lisp.

'message' is just a thin wrapper around format-message, which I don't
see being rewritten in Lisp any time soon.

In any case, if you want more flexibility in the available
destinations, the way to achieve that is by adding new kinds of
streams, so that all the print functions could benefit from that, not
just 'message'.

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

phillip.lord@russet.org.uk (Phillip Lord)
writes:

> It's all doable, just takes time. If you want
> it, why not fork from master, and start
> working on it?

I want it, but not enough to do it the correct
way unless my hacks work to say 80% or more.

Unless of course, there is a general sense this
would be useful. If so, I'd do it, but not in
a heartbeat, because I think it'd take somewhat
longer than that.

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Emanuel Berg]

phillip.lord@russet.org.uk (Phillip Lord)
writes:

>> Is that a bug or a feature request?
>> The reason I thought this would be cool to
>> have is exactly the messaging. I thought it
>> could be cool to drop certain messages
>> I never care for anyway (e.g., "Mark set",
>> "Auto-saving...done", etc.).
>
> The messaging framework of Emacs is a little
> under-powered at the moment -- one which bugs
> me, for instance, is Emacs has now (lisp
> level) ability to write to standard out.
> Unless it's running in batch, when it always
> does so.
>
> It's possible, of course, to write something
> better, but most packages use message
> directly. The ability to filter, suppress and
> change various messages would be good.

Indeed.

> Probably the best way to achieve this would
> be to pull message out of C and into lisp.
>
> It's all doable, just takes time. If you want
> it, why not fork from master, and start
> working on it?

Good idea!

-- 
underground experts united .... http://user.it.uu.se/~embe8573
Emacs Gnus Blogomatic ......... http://user.it.uu.se/~embe8573/blogomatic
                   - so far: 48 Blogomatic articles -                   

Re: hijack C function, use it transparently with in-between layer

[Barry Margolin]

In article <mailman.967.1465212644.1216.help-gnu-emacs@gnu.org>,
 phillip.lord@russet.org.uk (Phillip Lord) wrote:

> Barry Margolin <barmar@alum.mit.edu> writes:
> 
> > In article <868tyjnghn.fsf@student.uu.se>,
> >  Emanuel Berg <embe8573@student.uu.se> wrote:
> >
> >> The truth is down there!
> >> 
> >> (defvar old-message)
> >> (fset 'old-message (symbol-function #'message))
> >> 
> >> (defun message (format-string &rest args)
> >>   (unless (member format-string
> >>                   '(
> >>                     "It was just another system of control."
> >>                     "Mark set"
> >>                     ))
> >>       (apply `(old-message ,format-string ,@args) )))
> >> 
> >> ;; test:
> >> ;;
> >> ;;  (message "%s - %s" "1" "2")
> >> ;;  (message "Break on thru to the other side.")
> >> ;;  (message "It was just another system of control.")
> >> ;;  (message "Mark set")
> >
> > Shouldn't you do this using advice?
> 
> 
> Advice only works partially (as I expect this will). The message
> function is also called directly in the C layer. Advice will not capture
> those cases.

The same is true for redefining it as above. Nothing you do in Emacs 
Lisp will affect what happens for C-to-C calls.

-- 
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***