bug#20682: 25.0.50; Enable users to combine --script and --quick

bug#20682: 25.0.50; Enable users to combine --script and --quick

[Michael Heerdegen]

Hello!

This is a feature request to enable users to write emacs scripts (à la
--script) that can be executed in a way that --quick does (i.e. avoid
loading site stuff).

AFAIK this is not directly possible currently, because if you try with a
file starting like this:

#!/usr/bin/emacs -Q --script

the "#!" implementation on most Unices will not be able to pass more
than one argument to the interpreter.

The only thing that can be done to force --quick is to define a separate
interpreter like this:

--8<---------------cut here---------------start------------->8---
#!/bin/bash

emacs --quick --script $@
--8<---------------cut here---------------end--------------->8---

(suggested by saint@eng.it in g.e.help).

I have been told in the same thread that if --script would be available
as -s, one could start scripts with "#!/usr/bin/emacs -Qs" and that
would already solve the problem.


These are the main reason why using --quick can be desirable:

 - There are no messages like:

     Loading 00debian-vars...
     Loading /etc/emacs/site-start.d/50a2ps.el (source)...
     Loading /etc/emacs/site-start.d/50autoconf.el (source)...
     Loading /etc/emacs/site-start.d/50cmake-data.el (source)...
     Loading /etc/emacs/site-start.d/50dictionaries-common.el (source)...
     Loading debian-ispell...
     Loading /var/cache/dictionaries-common/emacsen-ispell-default.el (source)...
     Loading /var/cache/dictionaries-common/emacsen-ispell-dicts.el (source)...
     Loading /etc/emacs/site-start.d/51debian-el.el (source)...

   when running a script.  That would be a nice thing if you wanted to
   write something like wc or so.

 - It would speed up startup by a factor of five (for me, 0.01 secs
   vs. 0.002 secs when using the wrapper from above)

 - It might not be necessary to load site stuff.  On the contrary, if
   you want to write portable scripts, it might be more desirable to be
   able to write environment independent scripts


Many thanks,

Michael.




In GNU Emacs 25.0.50.2 (x86_64-unknown-linux-gnu, GTK+ Version 3.14.5)
 of 2015-05-27 on drachen
Windowing system distributor `The X.Org Foundation', version 11.0.11701000
System Description:	Debian GNU/Linux testing (stretch)

Configured features:
XPM JPEG TIFF GIF PNG RSVG IMAGEMAGICK SOUND DBUS GSETTINGS NOTIFY
LIBXML2 FREETYPE XFT ZLIB

bug#20682: 25.0.50; Enable users to combine --script and --quick

[Stefan Monnier]

> This is a feature request to enable users to write emacs scripts (à la
> --script) that can be executed in a way that --quick does (i.e. avoid
> loading site stuff).

I'm not very much in favor of adding yet another ad-hoc arg for that.
OTOH I'd be happy to see a new argument that lets you provide various
args as one.  E.g. "emacs --args=-Q,--script" which would be more
generally useful to deal with the limit of only one arg in shebang.
Given the size limit of shebang lines, we'd also want this extra arg to
be shortish.

An alternative would be to combine --args and --script, so you could
say "emacs --script=-Q" to get what you're asking for.

>  - There are no messages like:

>      Loading 00debian-vars...
>      Loading /etc/emacs/site-start.d/50a2ps.el (source)...
>      Loading /etc/emacs/site-start.d/50autoconf.el (source)...
>      Loading /etc/emacs/site-start.d/50cmake-data.el (source)...
>      Loading /etc/emacs/site-start.d/50dictionaries-common.el (source)...
>      Loading debian-ispell...
>      Loading /var/cache/dictionaries-common/emacsen-ispell-default.el (source)...
>      Loading /var/cache/dictionaries-common/emacsen-ispell-dicts.el (source)...
>      Loading /etc/emacs/site-start.d/51debian-el.el (source)...

>    when running a script.  That would be a nice thing if you wanted to
>    write something like wc or so.

I consider those messages as bugs.

BTW, Ideally, Debian packaging of Emacs packages should move towards
installing those packages as ELPA packages (i.e. install them into one
of package-directory-list and let Emacs load their <pkg>-autoloads.el
file).

>  - It would speed up startup by a factor of five (for me, 0.01 secs
>    vs. 0.002 secs when using the wrapper from above)

We should look into speeding this up, BTW.  Part of it might be
a problem on Debian's side (where those files do too much), but
I suspect that the problem goes deeper and we should figure out how to
improve it.  E.g. package.el's handling of package initialization could
use some optimization for the usual case where the set of packages is
the same as last time.

E.g. Ideally we could cache somewhere the last set of packages seen
along with the resulting set of <pkg>-autoloads.el files that we loaded
and actually keep the concatenation of those files into a single file
which we'd load instead.  Of course, this won't work currently because
those files are "position dependent" (e.g. they use load-file-name or
#$), but we should try and solve those problems.

>  - It might not be necessary to load site stuff.  On the contrary, if
>    you want to write portable scripts, it might be more desirable to be
>    able to write environment independent scripts

I tend to consider that any <pkg>-autoloads.el file should be 100%
harmless, or at least that the kind of changes they cause should be no
worse than the kinds of changes you experience when going from
Emacs-A.B to Emacs-A.C or Emacs-A+1.C


        Stefan


PS: Just to clarify, this should not be taken as an argument against
providing a way to get both -Q and --script as a single arg.

bug#20682: 25.0.50; Enable users to combine --script and --quick

[Glenn Morris]

Stefan Monnier wrote:

> OTOH I'd be happy to see a new argument that lets you provide various
> args as one.  E.g. "emacs --args=-Q,--script" which would be more
> generally useful to deal with the limit of only one arg in shebang.
> Given the size limit of shebang lines, we'd also want this extra arg to
> be shortish.
>
> An alternative would be to combine --args and --script, so you could
> say "emacs --script=-Q" to get what you're asking for.

Apparently it's somewhat conventional (perl, ruby) to use "-x" for this
purpose. A nice write-up at

https://github.com/smikes/node/blob/minus-x-switch/doc/Minus-X-Switch-Proposal.md

Amazingly, "-x" is available as an Emacs command-line switch...


Personally, I'd probably just use a trivial shell wrapper script.

bug#25671: Feature request: emacs -Q --script as a single binary

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 1138 bytes --]

Hi bug-gnu-emacs,

Since version 22 Emacs lets scripts begin with a shebang line:

  #!/usr/bin/emacs --script
  ;; Some ELisp code here

This breaks if Emacs isn't installed in /usr/bin, though.  Thus, one would want to write this:

  #!/usr/bin/env emacs --script
  ;; Some ELisp code here

But this doesn't work (env doesn't split its arguments), and it also breaks even if one is happy with hardcoding /usr/bin/emacs but tries to pass more than one argument (think -Q and --script).  Thus one writes this:

  #!/bin/sh
  ":"; exec emacs -Q --script "$0" "$@" # -*- mode: emacs-lisp; lexical-binding: t; -*-

(See http://stackoverflow.com/questions/6238331/emacs-shell-scripts-how-to-put-initial-options-into-the-script/6259330#6259330)

This isn't pretty.  Could we set up emacs to assume --script and possibly -Q when started as "elisp", or "emacs-script", or some other name (that is, could we make it so that invoking emacs with argv[0] = ".../elisp" starts emacs -Q --script)? (And ideally Emacs would ignore all options after the script's name, to let the script itself process them).

Thanks!
Clément.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#25671: Feature request: emacs -Q --script as a single binary

[Glenn Morris]

This seems like a duplicate of https://debbugs.gnu.org/20682 .

(IMO a big, unwieldy text editor that is never going to be a sensible
alternative to bash, perl, python etc for scripting.)

bug#25671: Feature request: emacs -Q --script as a single binary

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 433 bytes --]

On 2017-02-09 16:27, Glenn Morris wrote:
> This seems like a duplicate of https://debbugs.gnu.org/20682 .
> 
> (IMO a big, unwieldy text editor that is never going to be a sensible
> alternative to bash, perl, python etc for scripting.)

Partly; thanks for the pointer! That thread still assumes that you can pass an argument to Emacs, though, which isn't always the case (e.g. if you want to use /usr/bin/env).

Clément.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#25671: Feature request: emacs -Q --script as a single binary

[Eli Zaretskii]

> From: Clément Pit--Claudel <clement.pitclaudel@live.com>
> Date: Thu, 9 Feb 2017 16:14:41 -0500
> 
> Could we set up emacs to assume --script and possibly -Q when started as "elisp", or "emacs-script", or some other name (that is, could we make it so that invoking emacs with argv[0] = ".../elisp" starts emacs -Q --script)? (And ideally Emacs would ignore all options after the script's name, to let the script itself process them).

I believe changing the program behavior depending on how it was named
in the command that invoked it is against GNU coding standards.

bug#25671: Feature request: emacs -Q --script as a single binary

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 701 bytes --]

On 2017-02-10 02:37, Eli Zaretskii wrote:
>> From: Clément Pit--Claudel <clement.pitclaudel@live.com>
>> Date: Thu, 9 Feb 2017 16:14:41 -0500
>>
>> Could we set up emacs to assume --script and possibly -Q when started as "elisp", or "emacs-script", or some other name (that is, could we make it so that invoking emacs with argv[0] = ".../elisp" starts emacs -Q --script)? (And ideally Emacs would ignore all options after the script's name, to let the script itself process them).
> 
> I believe changing the program behavior depending on how it was named
> in the command that invoked it is against GNU coding standards.

What about providing elisp or emacs-script as a separate binary?


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#25671: Feature request: emacs -Q --script as a single binary

[Eli Zaretskii]

> Cc: 25671@debbugs.gnu.org
> From: Clément Pit--Claudel <clement.pitclaudel@live.com>
> Date: Fri, 10 Feb 2017 09:17:00 -0500
> 
> > I believe changing the program behavior depending on how it was named
> > in the command that invoked it is against GNU coding standards.
> 
> What about providing elisp or emacs-script as a separate binary?

What about it?

bug#25671: Feature request: emacs -Q --script as a single binary

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 703 bytes --]

On 2017-02-10 10:02, Eli Zaretskii wrote:
>> Cc: 25671@debbugs.gnu.org
>> From: Clément Pit--Claudel <clement.pitclaudel@live.com>
>> Date: Fri, 10 Feb 2017 09:17:00 -0500
>>
>>> I believe changing the program behavior depending on how it was named
>>> in the command that invoked it is against GNU coding standards.
>>
>> What about providing elisp or emacs-script as a separate binary?
> 
> What about it?

Would that be OK? Could it work?  I thought it would make implementation simpler to just make the emacs binary ct differently based on how its invoked, but if that's not OK then could we distribute a separate "elisp" or "emacs-script" binary that can be used in a shebang line?


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#25671: Feature request: emacs -Q --script as a single binary

[Eli Zaretskii]

> Cc: 25671@debbugs.gnu.org
> From: Clément Pit--Claudel <clement.pitclaudel@live.com>
> Date: Fri, 10 Feb 2017 11:45:55 -0500
> 
> >>> I believe changing the program behavior depending on how it was named
> >>> in the command that invoked it is against GNU coding standards.
> >>
> >> What about providing elisp or emacs-script as a separate binary?
> > 
> > What about it?
> 
> Would that be OK? Could it work?

Yes, it will.  Although distributing two large binaries might be
overkill.  (All that just to make shell scripts slightly simpler?)

bug#25671: Feature request: emacs -Q --script as a single binary

[Glenn Morris]

Clément Pit--Claudel wrote:

>>> What about providing elisp or emacs-script as a separate binary?
>> 
>> What about it?
>
> Would that be OK?

I would not like installing Emacs to pollute my PATH with trivial things
like that. In https://debbugs.gnu.org/20682#11 I mentioned a
standard solution for this issue as used by perl, ruby, etc;
namely a -x switch.

bug#25671: Feature request: emacs -Q --script as a single binary

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 577 bytes --]

On 2017-02-10 12:32, Glenn Morris wrote:
> Clément Pit--Claudel wrote:
> 
>>>> What about providing elisp or emacs-script as a separate binary?
>>>
>>> What about it?
>>
>> Would that be OK?
> 
> I would not like installing Emacs to pollute my PATH with trivial things
> like that. In https://debbugs.gnu.org/20682#11 I mentioned a
> standard solution for this issue as used by perl, ruby, etc;
> namely a -x switch.

Can you explain how that works wrt the problem I mentioned in my original post? "#!/usr/bin/env emacs -x" doesn't work, does it?

Clément


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#25671: Feature request: emacs -Q --script as a single binary

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 1113 bytes --]

On 2017-02-10 11:58, Eli Zaretskii wrote:
>> Cc: 25671@debbugs.gnu.org
>> From: Clément Pit--Claudel <clement.pitclaudel@live.com>
>> Date: Fri, 10 Feb 2017 11:45:55 -0500
>>
>>>>> I believe changing the program behavior depending on how it was named
>>>>> in the command that invoked it is against GNU coding standards.
>>>>
>>>> What about providing elisp or emacs-script as a separate binary?
>>>
>>> What about it?
>>
>> Would that be OK? Could it work?
> 
> Yes, it will.  Although distributing two large binaries might be
> overkill.  (All that just to make shell scripts slightly simpler?)

Right; that's why I thought the "two possible names for the same binary" approach was nice :/
It would make it possible to run elisp command line programs in Windows without a batch script wrapper, and it would make these scripts nicer in GNU/Linux and macOS.  But arguably that's still a limited benefit.  If there's no way to reuse the Emacs binary (that is, if there's no way to make "elisp" just be a wrapper around "emacs"), then I think we should close this issue.

Cheers,
Clément.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#20682: bug#25671: Feature request: emacs -Q --script as a single binary

[Lars Ingebrigtsen]

Glenn Morris <rgm@gnu.org> writes:

>> OTOH I'd be happy to see a new argument that lets you provide various
>> args as one.  E.g. "emacs --args=-Q,--script" which would be more
>> generally useful to deal with the limit of only one arg in shebang.
>> Given the size limit of shebang lines, we'd also want this extra arg to
>> be shortish.
>>
>> An alternative would be to combine --args and --script, so you could
>> say "emacs --script=-Q" to get what you're asking for.
>
> Apparently it's somewhat conventional (perl, ruby) to use "-x" for this
> purpose. A nice write-up at
>
> https://github.com/smikes/node/blob/minus-x-switch/doc/Minus-X-Switch-Proposal.md
>
> Amazingly, "-x" is available as an Emacs command-line switch...

I think that sounds really attractive, so I've added this to Emacs 29.
The new -x switch is like --script, but will kill Emacs when it reaches
the end of the script, and return the value of the final form as the
return value, so you can say use 0 at the end to signal success to the
caller.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#25671: Feature request: emacs -Q --script as a single binary

[Michael Heerdegen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> > Amazingly, "-x" is available as an Emacs command-line switch...
>
> I think that sounds really attractive, so I've added this to Emacs 29.
> The new -x switch is like --script, but will kill Emacs when it reaches
> the end of the script, and return the value of the final form as the
> return value, so you can say use 0 at the end to signal success to the
> caller.

Thanks - but it doesn't work correctly for me.  Contrarily to --script,
my init file is being loaded with -x.  Does that work correctly for you,
Lars?

TIA,

Michael.

bug#20682: bug#25671: Feature request: emacs -Q --script as a single binary

[Lars Ingebrigtsen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> Thanks - but it doesn't work correctly for me.  Contrarily to --script,
> my init file is being loaded with -x.  Does that work correctly for you,
> Lars?

Nope.  Should be fixed now.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#20682: bug#25671: Feature request: emacs -Q --script as a single binary

[Michael Heerdegen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Nope.  Should be fixed now.

Better - thanks.

Michael.