`--with-small-ja-dic`

`--with-small-ja-dic`

[Stefan Monnier]

Is this option still relevant?
And if is, why is it that it is a compile-time choice rather than
a run-time choice?


        Stefan

Re: `--with-small-ja-dic`

[Po Lu]

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

> Is this option still relevant?

Yes.  It's been with us for not more than a year, and was introduced at
the request of a user.

> And if is, why is it that it is a compile-time choice rather than
> a run-time choice?

It affects the compile-time generation of certain files in leim, AFAIK.

Re: `--with-small-ja-dic`

[Stefan Monnier]

>> And if is, why is it that it is a compile-time choice rather than
>> a run-time choice?
> It affects the compile-time generation of certain files in leim, AFAIK.

Couldn't we generate both versions of the file and let users choose at
run-time?


        Stefan

Re: `--with-small-ja-dic`

[Po Lu]

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

> Couldn't we generate both versions of the file and let users choose at
> run-time?

That would defeat the purpose of the option, which was meant to enable
installing Emacs without that large file, IIRC.

Re: `--with-small-ja-dic`

[Ulrich Mueller]

>>>>> On Thu, 15 Feb 2024, Po Lu wrote:

>> Couldn't we generate both versions of the file and let users choose at
>> run-time?

> That would defeat the purpose of the option, which was meant to enable
> installing Emacs without that large file, IIRC.

I wonder if the option couldn't be replaced by another one like
--without-ja-dic? That would reduce both the installation footprint
and the compile time.

Rationale is that Japanese users may prefer the large file, but users
who don't speak Japanese don't need the file at all.

Re: `--with-small-ja-dic`

[Eli Zaretskii]

> From: Ulrich Mueller <ulm@gentoo.org>
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>,  emacs-devel@gnu.org
> Date: Thu, 15 Feb 2024 09:37:36 +0100
> 
> I wonder if the option couldn't be replaced by another one like
> --without-ja-dic? That would reduce both the installation footprint
> and the compile time.
> 
> Rationale is that Japanese users may prefer the large file, but users
> who don't speak Japanese don't need the file at all.

How is this different from any other dictionary we produce as part of
the build?  E.g., ZIRANMA is also large.

Users who are bothered by files in the distribution they don't need
can simply delete them.  Or the downstream distros could provide more
fine-grained installation options.  We here traditionally include in
the installation everything that some feature might need, and I don't
see any reason to change that policy, except (perhaps) when we find a
way to keep some of the stuff on ELPA.

Re: `--with-small-ja-dic`

[Ulrich Mueller]

>>>>> On Thu, 15 Feb 2024, Eli Zaretskii wrote:

>> I wonder if the option couldn't be replaced by another one like
>> --without-ja-dic? That would reduce both the installation footprint
>> and the compile time.
>> 
>> Rationale is that Japanese users may prefer the large file, but users
>> who don't speak Japanese don't need the file at all.

> How is this different from any other dictionary we produce as part of
> the build?  E.g., ZIRANMA is also large.

Other dictionaries don't need a progress report of their long build
time. :)

Re: `--with-small-ja-dic`

[Eli Zaretskii]

> From: Ulrich Mueller <ulm@gentoo.org>
> Cc: luangruo@yahoo.com,  monnier@iro.umontreal.ca,  emacs-devel@gnu.org
> Date: Thu, 15 Feb 2024 14:11:42 +0100
> 
> >>>>> On Thu, 15 Feb 2024, Eli Zaretskii wrote:
> 
> >> I wonder if the option couldn't be replaced by another one like
> >> --without-ja-dic? That would reduce both the installation footprint
> >> and the compile time.
> >> 
> >> Rationale is that Japanese users may prefer the large file, but users
> >> who don't speak Japanese don't need the file at all.
> 
> > How is this different from any other dictionary we produce as part of
> > the build?  E.g., ZIRANMA is also large.
> 
> Other dictionaries don't need a progress report of their long build
> time. :)

There are other files whose build or compilation is slow, for example,
loaddefs.el and ucs-normalize.el.  However, if you build Emacs with
"make -jN", this time is not wasted.

Re: `--with-small-ja-dic`

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Date: Wed, 14 Feb 2024 17:32:21 -0500
> 
> Is this option still relevant?

I think it is.  It isn't that old, either; see

  https://lists.gnu.org/archive/html/emacs-devel/2022-06/msg00122.html

which gave its birth (and the following discussion).

> And if is, why is it that it is a compile-time choice rather than
> a run-time choice?

Because it's simpler, and also avoids making the build longer and the
tarball larger.

Why are you bothered by this, may I ask?

Re: `--with-small-ja-dic`

[Stefan Monnier]

>> Is this option still relevant?
> I think it is.  It isn't that old, either; see
>
>   https://lists.gnu.org/archive/html/emacs-devel/2022-06/msg00122.html
>
> which gave its birth (and the following discussion).

Hmm... OK, so AFAICT the reason to have it as a compile-time option is
that we found no compelling reason to ever use the reduced dictionary,
but we were afraid to make the change without offering some way to get
the old behavior, and the compile-time option was the cheaper way to
get that.

Then I suggest we remove that option altogether.

>> And if is, why is it that it is a compile-time choice rather than
>> a run-time choice?
> Because it's simpler, and also avoids making the build longer and the
> tarball larger.
> Why are you bothered by this, may I ask?

It just seems very odd: when I read the (long) list of `configure`
options, this one really stands out as being "unrelated" to the build.

If the option were `--with-big-ja-dic` I could imagine that we
considered the generated file to be too large for a normal build, but
the other way around doesn't make sense given all the other stuff we
include that any specific user will never use.


        Stefan

Re: `--with-small-ja-dic`

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: emacs-devel@gnu.org
> Date: Thu, 15 Feb 2024 09:13:19 -0500
> 
> >> Is this option still relevant?
> > I think it is.  It isn't that old, either; see
> >
> >   https://lists.gnu.org/archive/html/emacs-devel/2022-06/msg00122.html
> >
> > which gave its birth (and the following discussion).
> 
> Hmm... OK, so AFAICT the reason to have it as a compile-time option is
> that we found no compelling reason to ever use the reduced dictionary,
> but we were afraid to make the change without offering some way to get
> the old behavior, and the compile-time option was the cheaper way to
> get that.
> 
> Then I suggest we remove that option altogether.

We probably will, at some point.  But not yet: it was only added in
Emacs 29.

> >> And if is, why is it that it is a compile-time choice rather than
> >> a run-time choice?
> > Because it's simpler, and also avoids making the build longer and the
> > tarball larger.
> > Why are you bothered by this, may I ask?
> 
> It just seems very odd: when I read the (long) list of `configure`
> options, this one really stands out as being "unrelated" to the build.
> 
> If the option were `--with-big-ja-dic` I could imagine that we
> considered the generated file to be too large for a normal build, but
> the other way around doesn't make sense given all the other stuff we
> include that any specific user will never use.

I hope it does make sense now that you have read the above thread.