Re: Re: Make all tree-sitter modes optional

Re: Re: Make all tree-sitter modes optional

[Pedro Andres Aranda Gutierrez]

[-- Attachment #1: Type: text/plain, Size: 1327 bytes --]

Hi

Disclaimer: no intention to start a flame war, please...

Now: My experience in the last couple of weeks is that there are
tree-sitter modes in the Emacs tree that have non-ts counterparts in MELPA
only. And that's been frustrating. Basically because I don't compile ts
support and despite that, the treesit .el files are installed and used.
Example: dockerfile-mode is in MELPA only. So when I recently opened a
Dockerfile, I got an error saying that Dockerfile-ts-mode couldn't start
because my emacs hasn't ts support compiled in.

I would have prefered to see the dockerfile opened in text mode with no
errors. And then, when I install the Dockerfile-mode package from MELPA,
start working with all the goodies I got from installing the package from
MELPA.

My gut feeling is that if I don't have treesit support compiled in Emacs, I
don't want *-ts-*.el files in my elisp tree nor any side effects related to
these files. Sorry to say, but I feel that the enthusiasm about treesit
some have is a bit pushing for me. And I feel uncomfortable when pushed.

Thanks, /PA
-- 
Fragen sind nicht da um beantwortet zu werden,
Fragen sind da um gestellt zu werden
Georg Kreisler

Headaches with a Juju log:
unit-basic-16: 09:17:36 WARNING juju.worker.uniter.operation we should run
a leader-deposed hook here, but we can't yet

[-- Attachment #2: Type: text/html, Size: 1713 bytes --]

Re: Make all tree-sitter modes optional

[Eli Zaretskii]

> From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> Date: Thu, 19 Jan 2023 17:53:06 +0100
> 
> Now: My experience in the last couple of weeks is that there are tree-sitter modes in the Emacs tree that
> have non-ts counterparts in MELPA only. And that's been frustrating. Basically because I don't compile ts
> support and despite that, the treesit .el files are installed and used. Example: dockerfile-mode is in MELPA
> only. So when I recently opened a Dockerfile, I got an error saying that Dockerfile-ts-mode couldn't start
> because my emacs hasn't ts support compiled in. 
> 
> I would have prefered to see the dockerfile opened in text mode with no errors. And then, when I install the
> Dockerfile-mode package from MELPA, start working with all the goodies I got from installing the package
> from MELPA.
> 
> My gut feeling is that if I don't have treesit support compiled in Emacs, I don't want *-ts-*.el files in my elisp
> tree nor any side effects related to these files. Sorry to say, but I feel that the enthusiasm about treesit some
> have is a bit pushing for me. And I feel uncomfortable when pushed.

Please try the latest emacs-29 branch, where the tree-sitter based
modes have been made entirely optional.

Re: Make all tree-sitter modes optional

[Pedro Andres Aranda Gutierrez]

[-- Attachment #1: Type: text/plain, Size: 3386 bytes --]

Thanks for that :-)
It's a first step. However, the *-ts-mode.el files still go through the .el
compilation, yielding error messages during the compilation.
Additionally, they get installed in the binary distribution:

emacs-29 ➤ find Emacs.app -name \*-ts-\*
Emacs.app/Contents/Resources/lisp/progmodes/c-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/c-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/cmake-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/cmake-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/dockerfile-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/dockerfile-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/go-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/go-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/java-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/java-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/json-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/json-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/ruby-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/ruby-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/rust-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/rust-ts-mode.elc
Emacs.app/Contents/Resources/lisp/progmodes/typescript-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/progmodes/typescript-ts-mode.elc
Emacs.app/Contents/Resources/lisp/textmodes/toml-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/textmodes/toml-ts-mode.elc
Emacs.app/Contents/Resources/lisp/textmodes/yaml-ts-mode.el.gz
Emacs.app/Contents/Resources/lisp/textmodes/yaml-ts-mode.elc

That may be a source of confusion ...

Just my .02cents, /PA

On Fri, 20 Jan 2023 at 09:30, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> > Date: Thu, 19 Jan 2023 17:53:06 +0100
> >
> > Now: My experience in the last couple of weeks is that there are
> tree-sitter modes in the Emacs tree that
> > have non-ts counterparts in MELPA only. And that's been frustrating.
> Basically because I don't compile ts
> > support and despite that, the treesit .el files are installed and used.
> Example: dockerfile-mode is in MELPA
> > only. So when I recently opened a Dockerfile, I got an error saying that
> Dockerfile-ts-mode couldn't start
> > because my emacs hasn't ts support compiled in.
> >
> > I would have prefered to see the dockerfile opened in text mode with no
> errors. And then, when I install the
> > Dockerfile-mode package from MELPA, start working with all the goodies I
> got from installing the package
> > from MELPA.
> >
> > My gut feeling is that if I don't have treesit support compiled in
> Emacs, I don't want *-ts-*.el files in my elisp
> > tree nor any side effects related to these files. Sorry to say, but I
> feel that the enthusiasm about treesit some
> > have is a bit pushing for me. And I feel uncomfortable when pushed.
>
> Please try the latest emacs-29 branch, where the tree-sitter based
> modes have been made entirely optional.
>


-- 
Fragen sind nicht da um beantwortet zu werden,
Fragen sind da um gestellt zu werden
Georg Kreisler

Headaches with a Juju log:
unit-basic-16: 09:17:36 WARNING juju.worker.uniter.operation we should run
a leader-deposed hook here, but we can't yet

[-- Attachment #2: Type: text/html, Size: 4177 bytes --]

Re: Make all tree-sitter modes optional

[Eli Zaretskii]

> From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> Date: Fri, 20 Jan 2023 17:31:44 +0100
> Cc: emacs-devel@gnu.org
> 
> Thanks for that :-) 
> It's a first step. However, the *-ts-mode.el files still go through the .el compilation, yielding error messages
> during the compilation. 

What error messages?  Please show them, we cannot possibly guess what
messages do you see.

And why is it a problem that the *.el files are being compiled as part
of the build?  A build compiles all the *.el files, including ones you
will never use and ones that you cannot use (e.g., because they are
for Emacs running on other OSes).  That is never a problem.

> Additionally, they get installed in the binary distribution:
> 
> emacs-29 ➤ find Emacs.app -name \*-ts-\*
> Emacs.app/Contents/Resources/lisp/progmodes/c-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/c-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/cmake-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/cmake-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/dockerfile-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/dockerfile-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/go-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/go-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/java-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/java-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/json-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/json-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/ruby-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/ruby-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/rust-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/rust-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/progmodes/typescript-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/progmodes/typescript-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/textmodes/toml-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/textmodes/toml-ts-mode.elc
> Emacs.app/Contents/Resources/lisp/textmodes/yaml-ts-mode.el.gz
> Emacs.app/Contents/Resources/lisp/textmodes/yaml-ts-mode.elc
> 
> That may be a source of confusion ...

Why would it confuse?  You also have there stuff like w32-win.el.gz,
which is used only on MS-Windows, and other files that are not
necessarily for your configuration.  This is not a problem, and
shouldn't be one.

Re: Make all tree-sitter modes optional

[Pedro Andres Aranda Gutierrez]

[-- Attachment #1: Type: text/plain, Size: 4321 bytes --]

Hi Eli,

Eli writes:
> Why would it confuse?  You also have there stuff like w32-win.el.gz,
> which is used only on MS-Windows, and other files that are not
> necessarily for your configuration.  This is not a problem, and
> shouldn't be one.

I don't know, maybe for me there is a difference between the OS specific
stuff, tree-sitter and other stuff in Emacs:

tree-sitter modes 'compete' with 'regular' modes and if I don't have
tree-sitter activated at compile time, it can be misleading to see those
files there and 'sub-optimal' (to say the least) to get a message that you
can't use them.
OS related stuff is different in the sense that, well, if I'm on a Linux
system and try to use (you say the OS)-specific features, it is natural
that I get a 'wake-up' message there and stop trying to do things that make
no sense.
As for the rest, having dormant features that _work_ (or are WIP with a
level of maturity enough to be in master) and only wait for me to test them
and activate them if they help me in my day-to-day interactions with Emacs,
of course, put 10^n n->infinity of those in Emacs, no problem.

In that sense, if there was a way to disregard *-ts-*.el files in ELC/ELN
compilation and installation when I compile Emacs _without_ tree-sitter
support, the whole picture would be (once again, IMvHO) much more coherent.

Thanks for your patience, /PA

On Fri, 20 Jan 2023 at 20:13, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> > Date: Fri, 20 Jan 2023 17:31:44 +0100
> > Cc: emacs-devel@gnu.org
> >
> > Thanks for that :-)
> > It's a first step. However, the *-ts-mode.el files still go through the
> .el compilation, yielding error messages
> > during the compilation.
>
> What error messages?  Please show them, we cannot possibly guess what
> messages do you see.
>
> And why is it a problem that the *.el files are being compiled as part
> of the build?  A build compiles all the *.el files, including ones you
> will never use and ones that you cannot use (e.g., because they are
> for Emacs running on other OSes).  That is never a problem.
>
> > Additionally, they get installed in the binary distribution:
> >
> > emacs-29 ➤ find Emacs.app -name \*-ts-\*
> > Emacs.app/Contents/Resources/lisp/progmodes/c-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/c-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/cmake-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/cmake-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/dockerfile-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/dockerfile-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/go-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/go-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/java-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/java-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/json-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/json-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/ruby-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/ruby-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/rust-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/rust-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/progmodes/typescript-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/progmodes/typescript-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/textmodes/toml-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/textmodes/toml-ts-mode.elc
> > Emacs.app/Contents/Resources/lisp/textmodes/yaml-ts-mode.el.gz
> > Emacs.app/Contents/Resources/lisp/textmodes/yaml-ts-mode.elc
> >
> > That may be a source of confusion ...
>
> Why would it confuse?  You also have there stuff like w32-win.el.gz,
> which is used only on MS-Windows, and other files that are not
> necessarily for your configuration.  This is not a problem, and
> shouldn't be one.
>


-- 
Fragen sind nicht da um beantwortet zu werden,
Fragen sind da um gestellt zu werden
Georg Kreisler

Headaches with a Juju log:
unit-basic-16: 09:17:36 WARNING juju.worker.uniter.operation we should run
a leader-deposed hook here, but we can't yet

[-- Attachment #2: Type: text/html, Size: 5351 bytes --]

Re: Make all tree-sitter modes optional

[Eli Zaretskii]

> From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> Date: Sat, 21 Jan 2023 12:30:26 +0100
> Cc: emacs-devel@gnu.org
> 
> Eli writes:
> > Why would it confuse?  You also have there stuff like w32-win.el.gz,
> > which is used only on MS-Windows, and other files that are not
> > necessarily for your configuration.  This is not a problem, and
> > shouldn't be one.
> 
> I don't know, maybe for me there is a difference between the OS specific stuff, tree-sitter and other stuff in
> Emacs:

It is nothing new in Emacs: we provide many packages, some of which
are specific to platforms other than yours, some others need optional
libraries that you don't necessarily have, etc.

> tree-sitter modes 'compete' with 'regular' modes and if I don't have tree-sitter activated at compile time, it
> can be misleading to see those files there

I disagree it should mislead anyone, and let's leave it at that.

> and 'sub-optimal' (to say the least) to get a message that you
> can't use them.

What message?  I asked to show these messages before, and I didn't see
your answer to that question.  We don't intend to have Emacs show any
such messages just because you don't have tree-sitter installed.

> OS related stuff is different in the sense that, well, if I'm on a Linux system and try to use (you say the
> OS)-specific features, it is natural that I get a 'wake-up' message there and stop trying to do things that make
> no sense. 

Same thing if you use a Lisp package which requires an optional
library you don't have installed or if you use Emacs which wasn't
built with support for that library.  Examples: GnuTLS, HarfBuzz,
librsvg, sqlite3, etc.

> As for the rest, having dormant features that _work_ (or are WIP with a level of maturity enough to be in
> master) and only wait for me to test them and activate them if they help me in my day-to-day interactions
> with Emacs, of course, put 10^n n->infinity of those in Emacs, no problem.
> 
> In that sense, if there was a way to disregard *-ts-*.el files in ELC/ELN compilation and installation when I
> compile Emacs _without_ tree-sitter support, the whole picture would be (once again, IMvHO) much more
> coherent.

We don't disregard any Lisp files.  When Emacs builds, it compiles all
the files in the source tree.  A release tarball comes with *.elc
files already compiled, and *.eln files will only be produced if you
actually load the corresponding *.el package.  So I see no problem
here.

Re: Make all tree-sitter modes optional

[Pedro Andres Aranda Gutierrez]

[-- Attachment #1: Type: text/plain, Size: 3666 bytes --]

Eli writes:
> What message?  I asked to show these messages before, and I didn't see
> your answer to that question.  We don't intend to have Emacs show any
> such messages just because you don't have tree-sitter installed.

Just try M-x c-ts-mode and you'l get this message *four* times (side
question, isn't one enough?)

⛔ Warning (treesit): Cannot activate tree-sitter, because tree-sitter
library is not compiled with Emacs

Another question: If I can't use the command, wouldn't it be better if I
had no access to it?

Best, /PA

PS: BTW, at this point, removing the 'Warning (treesit)' would leave a
coherent message WRT the no entry sign, but that's another thread ;-)


On Sat, 21 Jan 2023 at 12:48, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> > Date: Sat, 21 Jan 2023 12:30:26 +0100
> > Cc: emacs-devel@gnu.org
> >
> > Eli writes:
> > > Why would it confuse?  You also have there stuff like w32-win.el.gz,
> > > which is used only on MS-Windows, and other files that are not
> > > necessarily for your configuration.  This is not a problem, and
> > > shouldn't be one.
> >
> > I don't know, maybe for me there is a difference between the OS specific
> stuff, tree-sitter and other stuff in
> > Emacs:
>
> It is nothing new in Emacs: we provide many packages, some of which
> are specific to platforms other than yours, some others need optional
> libraries that you don't necessarily have, etc.
>
> > tree-sitter modes 'compete' with 'regular' modes and if I don't have
> tree-sitter activated at compile time, it
> > can be misleading to see those files there
>
> I disagree it should mislead anyone, and let's leave it at that.
>
> > and 'sub-optimal' (to say the least) to get a message that you
> > can't use them.
>
> What message?  I asked to show these messages before, and I didn't see
> your answer to that question.  We don't intend to have Emacs show any
> such messages just because you don't have tree-sitter installed.
>
> > OS related stuff is different in the sense that, well, if I'm on a Linux
> system and try to use (you say the
> > OS)-specific features, it is natural that I get a 'wake-up' message
> there and stop trying to do things that make
> > no sense.
>
> Same thing if you use a Lisp package which requires an optional
> library you don't have installed or if you use Emacs which wasn't
> built with support for that library.  Examples: GnuTLS, HarfBuzz,
> librsvg, sqlite3, etc.
>
> > As for the rest, having dormant features that _work_ (or are WIP with a
> level of maturity enough to be in
> > master) and only wait for me to test them and activate them if they help
> me in my day-to-day interactions
> > with Emacs, of course, put 10^n n->infinity of those in Emacs, no
> problem.
> >
> > In that sense, if there was a way to disregard *-ts-*.el files in
> ELC/ELN compilation and installation when I
> > compile Emacs _without_ tree-sitter support, the whole picture would be
> (once again, IMvHO) much more
> > coherent.
>
> We don't disregard any Lisp files.  When Emacs builds, it compiles all
> the files in the source tree.  A release tarball comes with *.elc
> files already compiled, and *.eln files will only be produced if you
> actually load the corresponding *.el package.  So I see no problem
> here.
>


-- 
Fragen sind nicht da um beantwortet zu werden,
Fragen sind da um gestellt zu werden
Georg Kreisler

Headaches with a Juju log:
unit-basic-16: 09:17:36 WARNING juju.worker.uniter.operation we should run
a leader-deposed hook here, but we can't yet

[-- Attachment #2: Type: text/html, Size: 4822 bytes --]

Re: Make all tree-sitter modes optional

[Eli Zaretskii]

> From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> Date: Sun, 22 Jan 2023 07:23:17 +0100
> Cc: emacs-devel@gnu.org
> 
> Eli writes:
> > What message?  I asked to show these messages before, and I didn't see
> > your answer to that question.  We don't intend to have Emacs show any
> > such messages just because you don't have tree-sitter installed.
> 
> Just try M-x c-ts-mode and you'l get this message *four* times (side question, isn't one enough?)
> 
> ⛔ Warning (treesit): Cannot activate tree-sitter, because tree-sitter library is not compiled with Emacs

Thanks.  This is an issue we know about, it will be fixed soon,
probably today or tomorrow.

> Another question: If I can't use the command, wouldn't it be better if I had no access to it?

The way Emacs implements the "you have no access to the command" is by
signaling an error when/if you try invoking it.  So we are okay in
that department.

Re: Make all tree-sitter modes optional

[Eli Zaretskii]

> Date: Sun, 22 Jan 2023 08:38:31 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> > From: Pedro Andres Aranda Gutierrez <paaguti@gmail.com>
> > Date: Sun, 22 Jan 2023 07:23:17 +0100
> > Cc: emacs-devel@gnu.org
> > 
> > ⛔ Warning (treesit): Cannot activate tree-sitter, because tree-sitter library is not compiled with Emacs
> 
> Thanks.  This is an issue we know about, it will be fixed soon,
> probably today or tomorrow.

Should be already fixed on the emacs-29 branch.