Move r-{desolve,quadprog,pracma,subplex} from maths.scm to cran.scm

Move r-{desolve,quadprog,pracma,subplex} from maths.scm to cran.scm

[zimoun]

Dear,

The packages r-desolve, r-quadprog, r-subplex and r-pracma are defined
in `maths.scm`. They come from the CRAN archive. For consistency, they
should be defined in the file `cran.scm`. Do you agree to move them?

All the best,
simon

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[zimoun]

Hi,

On Thu, 31 Oct 2019 at 16:37, zimoun <zimon.toutoune@gmail.com> wrote:

> The packages r-desolve, r-quadprog, r-subplex and r-pracma are defined
> in `maths.scm`. They come from the CRAN archive. For consistency, they
> should be defined in the file `cran.scm`. Do you agree to move them?

What happens to the Copyright?

No issue with the Ricardo's one. Full of years. :-)

Roel and Leo have Copyright 2016 in the file maths.scm.
But Copytright 2017,2018 and 2018 in the file cran.scm.

Does a transfer of copyright between the two files apply?


Thank in advance.

All the best,
simon

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[Ricardo Wurmus]

zimoun <zimon.toutoune@gmail.com> writes:

> On Thu, 31 Oct 2019 at 16:37, zimoun <zimon.toutoune@gmail.com> wrote:
>
>> The packages r-desolve, r-quadprog, r-subplex and r-pracma are defined
>> in `maths.scm`. They come from the CRAN archive. For consistency, they
>> should be defined in the file `cran.scm`. Do you agree to move them?
>
> What happens to the Copyright?
>
> No issue with the Ricardo's one. Full of years. :-)
>
> Roel and Leo have Copyright 2016 in the file maths.scm.
> But Copytright 2017,2018 and 2018 in the file cran.scm.
>
> Does a transfer of copyright between the two files apply?

Yes, we should update the copyright lines.  I usually rely on the
information provided by the git log.

-- 
Ricardo

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[Ricardo Wurmus]

Hi simon,

> The packages r-desolve, r-quadprog, r-subplex and r-pracma are defined
> in `maths.scm`. They come from the CRAN archive. For consistency, they
> should be defined in the file `cran.scm`. Do you agree to move them?

I think moving them to cran.scm is fine.  In my opinion both of these
places are similarly poor choices: maths.scm because we drag in the R
build system and dependencies, and cran.scm because it’s huge and will
only get bigger.

On the plus side, R code is rarely as “involved” as Python code, so I
don’t expect cran.scm to become as tangled up with dependencies from all
over the place as python-xyz.scm, so letting cran.scm grow is probably
just fine.

Since nobody else seems to hold any strong opinions on this issue, I’d
say you’re welcome to move these package definitions to cran.scm.
Please also update the Copyright comments at the top of the file where
needed (use “git blame” and “git log” to simplify this task) and remove
whatever spurious module imports there may be in maths.scm after
removing the R package definitions.

Thank you!

--
Ricardo

PS: I also think that CRAN things in bioinformatics.scm should be moved
to cran.scm, and even some or all of the R stuff in statistics.scm.
(Same applies to Bioconductor packages, which should end up in
bioconductor.scm where possible.)

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[zimoun]

Hi Ricardo,

Thank you for your inputs.

On Fri, 22 Nov 2019 at 15:47, Ricardo Wurmus <rekado@elephly.net> wrote:

> > The packages r-desolve, r-quadprog, r-subplex and r-pracma are defined
> > in `maths.scm`. They come from the CRAN archive. For consistency, they
> > should be defined in the file `cran.scm`. Do you agree to move them?
>
> I think moving them to cran.scm is fine.  In my opinion both of these
> places are similarly poor choices: maths.scm because we drag in the R
> build system and dependencies, and cran.scm because it’s huge and will
> only get bigger.
>
> On the plus side, R code is rarely as “involved” as Python code, so I
> don’t expect cran.scm to become as tangled up with dependencies from all
> over the place as python-xyz.scm, so letting cran.scm grow is probably
> just fine.

Do you agree that we should document this rationale somewhere?
AFAIK, nothing is written down about the lang-xyz.scm or the R land
(cran vs bioconductor vs bioinformatics vs math vs statistiscs).
I will read our materials and come back with a proposal: manual or
cookbook or plain text in repo or etc.
What do you think?


> Since nobody else seems to hold any strong opinions on this issue, I’d
> say you’re welcome to move these package definitions to cran.scm.
> Please also update the Copyright comments at the top of the file where
> needed (use “git blame” and “git log” to simplify this task) and remove
> whatever spurious module imports there may be in maths.scm after
> removing the R package definitions.

I submitted patches [1] that respect these advices. I hope so. :-)

[1] https://debbugs.gnu.org/cgi/bugreport.cgi?bug=38184


> PS: I also think that CRAN things in bioinformatics.scm should be moved
> to cran.scm, and even some or all of the R stuff in statistics.scm.
> (Same applies to Bioconductor packages, which should end up in
> bioconductor.scm where possible.)

I will do.
I will dig into the big move of Haskell or Lisp to see how it is
better to split the commits.


Thanks.
simon

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[zimoun]

Hi Ricarco


On Fri, 22 Nov 2019 at 15:47, Ricardo Wurmus <rekado@elephly.net> wrote:

> PS: I also think that CRAN things in bioinformatics.scm should be moved
> to cran.scm, and even some or all of the R stuff in statistics.scm.
> (Same applies to Bioconductor packages, which should end up in
> bioconductor.scm where possible.)

I am working on that.

Let consider some quick stats with "grep".

1.
Currently, the file bioconductor.scm contains 4 packages with
'cran-uri' and git blames you. :-)

Well, the 4 commits are:

: a207bca2ad gnu: r-codedepends: Move from cran to bioconductor.
: 3a0babacdc gnu: Add r-htscluster.
: 7ed869f796 gnu: Add r-nbpseq.
: 80eb01c776 gnu: r-gkmsvm: Move to (gnu packages bioconductor).

with some comments  as

--8<---------------cut here---------------start------------->8---
;; This is a CRAN package, but it uncharacteristically depends on a
;; Bioconductor package
--8<---------------cut here---------------end--------------->8---

Do you remember a special motivation?


2.
The second point is that the file statistics.scm contains 227 packages
and 206 are cran-uri. And the file cran.scm already contains 602
packages, so it will be almost as python-xyz.scm is. Why not. Then,
this file statistics.scm will be almost empty and I propose instead to
populate the files maths.scm, python-xyz.scm.

What do you think?


3.
It is a big move. Do you prefer
 - a commit per package, so more or less 300 commits?
 - or a commit bioconductor.scm->cran.scm, another
statistics.scm->cran.scm, bioinformatics.scm->bioconductor.scm and
bioconductor->bioinformatics.scm, etc.?


Thank you for your insights.

All the best,
simon

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[Ricardo Wurmus]

zimoun <zimon.toutoune@gmail.com> writes:

> 1.
> Currently, the file bioconductor.scm contains 4 packages with
> 'cran-uri' and git blames you. :-)
>
> Well, the 4 commits are:
>
> : a207bca2ad gnu: r-codedepends: Move from cran to bioconductor.
> : 3a0babacdc gnu: Add r-htscluster.
> : 7ed869f796 gnu: Add r-nbpseq.
> : 80eb01c776 gnu: r-gkmsvm: Move to (gnu packages bioconductor).
>
> with some comments  as
>
> --8<---------------cut here---------------start------------->8---
> ;; This is a CRAN package, but it uncharacteristically depends on a
> ;; Bioconductor package
> --8<---------------cut here---------------end--------------->8---
>
> Do you remember a special motivation?

Yes, the comments say why.  It’s very unusual for a CRAN package to
depend on Bioconductor packages.  Usually, it is the other way around.
I just didn’t want (gnu packages cran) to point to (gnu packages
bioconductor).  A more correct way to deal with this would be put these
outliers in a separate module, or even to ignore this all together until
it becomes a problem.

> 2.
> The second point is that the file statistics.scm contains 227 packages
> and 206 are cran-uri. And the file cran.scm already contains 602
> packages, so it will be almost as python-xyz.scm is. Why not. Then,
> this file statistics.scm will be almost empty and I propose instead to
> populate the files maths.scm, python-xyz.scm.
>
> What do you think?

statistics.scm was the only thing we had for R in the beginning.  When
it grew and it became apparent that more and more CRAN packages would be
added, I felt that the module became less suitable.

I don’t understand your comment about “maths.scm” and “python-xyz.scm”.
I think it’s good to move CRAN and Bioconductor packages out of
statistics.scm whenever possible.  We can leave the actual statistics
programs there as well as potentially unmovable CRAN packages.

> 3.
> It is a big move. Do you prefer
>  - a commit per package, so more or less 300 commits?
>  - or a commit bioconductor.scm->cran.scm, another
> statistics.scm->cran.scm, bioinformatics.scm->bioconductor.scm and
> bioconductor->bioinformatics.scm, etc.?

I prefer one commit per moved package.  I guess this is not what you
hoped for, but it makes for a more fine-grained log — and it boosts your
commit stats ;)

--
Ricardo

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[zimoun]

Hi,

On Tue, 10 Dec 2019 at 09:41, Ricardo Wurmus <rekado@elephly.net> wrote:

> > --8<---------------cut here---------------start------------->8---
> > ;; This is a CRAN package, but it uncharacteristically depends on a
> > ;; Bioconductor package
> > --8<---------------cut here---------------end--------------->8---
> >
> > Do you remember a special motivation?
>
> Yes, the comments say why.  It’s very unusual for a CRAN package to
> depend on Bioconductor packages.  Usually, it is the other way around.
> I just didn’t want (gnu packages cran) to point to (gnu packages
> bioconductor).  A more correct way to deal with this would be put these
> outliers in a separate module, or even to ignore this all together until
> it becomes a problem.

If I understand well, the policy is: the packages in the file cran.scm
cannot import '(gnu packages bioconductor).

In the file cran.scm, for example the packages r-shiny or r-sankeyd3
do not come from CRAN but directly from Github.
Maybe, we could group all the 6 non-CRAN packages and the 4 CRAN
packages depending on Bioconductor to a unique file.

What do you think?


> > 2.
> > The second point is that the file statistics.scm contains 227 packages
> > and 206 are cran-uri. And the file cran.scm already contains 602
> > packages, so it will be almost as python-xyz.scm is. Why not. Then,
> > this file statistics.scm will be almost empty and I propose instead to
> > populate the files maths.scm, python-xyz.scm.
> >
> > What do you think?
>
> statistics.scm was the only thing we had for R in the beginning.  When
> it grew and it became apparent that more and more CRAN packages would be
> added, I felt that the module became less suitable.

Sorry, my words was not good enough. I propose to remove the file
statistics.scm by reordering the packages.


> I don’t understand your comment about “maths.scm” and “python-xyz.scm”.
> I think it’s good to move CRAN and Bioconductor packages out of
> statistics.scm whenever possible.  We can leave the actual statistics
> programs there as well as potentially unmovable CRAN packages.

In the file statistics.scm, once move to cran.scm or bioconductor.scm,
there is few packages. For example 2 packages coming from PyPI and I
propose to move them to the file python-xyz.scm. The other ones could
be moved to the file maths.scm or the unamed (yet) file (see above).


> > 3.
> > It is a big move. Do you prefer
> >  - a commit per package, so more or less 300 commits?
> >  - or a commit bioconductor.scm->cran.scm, another
> > statistics.scm->cran.scm, bioinformatics.scm->bioconductor.scm and
> > bioconductor->bioinformatics.scm, etc.?
>
> I prefer one commit per moved package.  I guess this is not what you
> hoped for, but it makes for a more fine-grained log — and it boosts your
> commit stats ;)

Ahah! I was almost sure you answered that. :-)
Ok, I will.
Be prepared for the storm. ;-)


All the best,
simon

Re: Move r-{desolve, quadprog, pracma, subplex} from maths.scm to cran.scm

[Ricardo Wurmus]

zimoun <zimon.toutoune@gmail.com> writes:

> If I understand well, the policy is: the packages in the file cran.scm
> cannot import '(gnu packages bioconductor).

Yes.  It’s not official policy, but I think we should avoid mutually
recursive module imports when we have a choice.

> In the file cran.scm, for example the packages r-shiny or r-sankeyd3
> do not come from CRAN but directly from Github.
> Maybe, we could group all the 6 non-CRAN packages and the 4 CRAN
> packages depending on Bioconductor to a unique file.
>
> What do you think?

Yes, the should probably be moved.  It’s also very easy to miss them
when doing mass upgrades, hidden among all these CRAN packages.

>> > 2.
>> > The second point is that the file statistics.scm contains 227 packages
>> > and 206 are cran-uri. And the file cran.scm already contains 602
>> > packages, so it will be almost as python-xyz.scm is. Why not. Then,
>> > this file statistics.scm will be almost empty and I propose instead to
>> > populate the files maths.scm, python-xyz.scm.
>> >
>> > What do you think?
>>
>> statistics.scm was the only thing we had for R in the beginning.  When
>> it grew and it became apparent that more and more CRAN packages would be
>> added, I felt that the module became less suitable.
>
> Sorry, my words was not good enough. I propose to remove the file
> statistics.scm by reordering the packages.

I think it’s fine to keep statistics.scm even when it’s small.  R and
pspp *are* stats packages after all, so this module remains the perfect
location for them.

>> I don’t understand your comment about “maths.scm” and “python-xyz.scm”.
>> I think it’s good to move CRAN and Bioconductor packages out of
>> statistics.scm whenever possible.  We can leave the actual statistics
>> programs there as well as potentially unmovable CRAN packages.
>
> In the file statistics.scm, once move to cran.scm or bioconductor.scm,
> there is few packages. For example 2 packages coming from PyPI and I
> propose to move them to the file python-xyz.scm. The other ones could
> be moved to the file maths.scm or the unamed (yet) file (see above).

I’d like to avoid moving anything *to* python-xyz.scm, because it’s a
bit of a dump.  I think it’s fine to just focus on the R packages for
now and address the rest later — if at all.

--
Ricardo