gmane.linux.gentoo.devel

sci-libs/scipy -> dev-python/scipy ?

Hello *,

Wouldn't it be nice to move scipy from sci-libs to dev-python? All similar 
and related packages live in dev-python: numeric, scientificpython, 
matplotlib... I know that moving packages is a major pain in the #$$, but 
the present situation seems illogical (I would never guess to search for 
this package in sci-libs if I didn't know it's there).

Andrey
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Donnie Berkholz | 8 Jul 05:34

Re: sci-libs/scipy -> dev-python/scipy ?

On 02:59 Tue 08 Jul     , Andrey Grozin wrote:
> Wouldn't it be nice to move scipy from sci-libs to dev-python? All 
> similar and related packages live in dev-python: numeric, 
> scientificpython, matplotlib... I know that moving packages is a major 
> pain in the #$$, but the present situation seems illogical (I would never 
> guess to search for this package in sci-libs if I didn't know it's 
> there).

Why would you be searching by looking in categories anyway? There are 
tons of dedicated search tools.

I actually object to having crap in dev-python, because things should be 
categorized functionally instead of by the language they're implemented 
in. 90% of the time you don't care about the language. But category 
moves are pretty much pointless, so I don't normally bring it up.

--

-- 
Thanks,
Donnie

Donnie Berkholz
Developer, Gentoo Linux
Blog: http://dberkholz.wordpress.com

headers

Joe Peterson | 8 Jul 05:51

Re: sci-libs/scipy -> dev-python/scipy ?

Donnie Berkholz wrote:
> I actually object to having crap in dev-python, because things should be 
> categorized functionally instead of by the language they're implemented 
> in. 90% of the time you don't care about the language. But category 
> moves are pretty much pointless, so I don't normally bring it up.

Do you mean it is pointless because categories are pointless, or because
it is not worth the trouble of doing the move?  I assume we inherited
the category idea from fbsd ports.

Since we have categories (and assuming we'll keep them), I think that
things should be categorized as correctly as possible (perhaps it's not
the highest priority, but better to have it right than not, eventually).

If it is better to have scipy in sci-libs rather dev-python, then
perhaps other dev-python packages (like you said) should be moved
elsewhere to be consistent.  A quick look at the fbsd ports shows that
scipy is in "science" and numpy is in "math" (for example), so that
agrees with your feeling that neither belong in dev-python.

					-Joe
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Donnie Berkholz | 8 Jul 06:02

Re: sci-libs/scipy -> dev-python/scipy ?

On 21:51 Mon 07 Jul     , Joe Peterson wrote:
> Donnie Berkholz wrote:
> > I actually object to having crap in dev-python, because things should be 
> > categorized functionally instead of by the language they're implemented 
> > in. 90% of the time you don't care about the language. But category 
> > moves are pretty much pointless, so I don't normally bring it up.
> 
> Do you mean it is pointless because categories are pointless, or because
> it is not worth the trouble of doing the move?  I assume we inherited
> the category idea from fbsd ports.

I meant moves were largely pointless, although categories are to a 
lesser extent. Tags would be a lot better, since nothing can be 
categorized perfectly into a single place.

> Since we have categories (and assuming we'll keep them), I think that
> things should be categorized as correctly as possible (perhaps it's not
> the highest priority, but better to have it right than not, eventually).

I don't think it's worth losing track of the CVS history just so we can 
have something in a different place that ultimately is hardly useful to 
anyone.

--

-- 
Thanks,
Donnie

Donnie Berkholz
Developer, Gentoo Linux
Blog: http://dberkholz.wordpress.com

(Continue reading)

headers

Joe Peterson | 8 Jul 06:16

Re: sci-libs/scipy -> dev-python/scipy ?

Donnie Berkholz wrote:
> I meant moves were largely pointless, although categories are to a 
> lesser extent. Tags would be a lot better, since nothing can be 
> categorized perfectly into a single place.

Yes, I can see the benefit of a tag paradigm.  I, myself, find it more
trouble than benefit to have the extra directory level.  I often do "cd
/usr/portage/*/foo" to get to the foo package, and it often gets a hit
in licenses or elsewhere that trips up this practice...

> I don't think it's worth losing track of the CVS history just so we can 
> have something in a different place that ultimately is hardly useful to 
> anyone.

Ah yes, CVS would present a problem here.  I suppose if/when the whole
tree is converted to svn, at that point moves would be more practical.
Too bad, though, that this has become a barrier to the ability to change
a category easily and without losing the history.

						-Joe
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Marijn Schouten (hkBst | 8 Jul 18:34

Re: sci-libs/scipy -> dev-python/scipy ?


Joe Peterson wrote:
> Donnie Berkholz wrote:
>> I meant moves were largely pointless, although categories are to a 
>> lesser extent. Tags would be a lot better, since nothing can be 
>> categorized perfectly into a single place.
> 
> Yes, I can see the benefit of a tag paradigm.  I, myself, find it more
> trouble than benefit to have the extra directory level.  I often do "cd
> /usr/portage/*/foo" to get to the foo package, and it often gets a hit
> in licenses or elsewhere that trips up this practice...
> 
>> I don't think it's worth losing track of the CVS history just so we can 
>> have something in a different place that ultimately is hardly useful to 
>> anyone.
> 
> Ah yes, CVS would present a problem here.  I suppose if/when the whole
> tree is converted to svn, at that point moves would be more practical.

I suppose you mean git. Since it tracks content and not files, moves are
trivial. Git actually finds your moves for you, after you've moved content
around; such as when doing a bump.

> Too bad, though, that this has become a barrier to the ability to change
> a category easily and without losing the history.
> 
> 						-Joe

Marijn

(Continue reading)

headers

Joe Peterson | 8 Jul 18:54

Re: sci-libs/scipy -> dev-python/scipy ?

Marijn Schouten (hkBst) wrote:
> I suppose you mean git. Since it tracks content and not files, moves are
> trivial. Git actually finds your moves for you, after you've moved content
> around; such as when doing a bump.

Even better!

	-Joe
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Ciaran McCreesh | 8 Jul 19:14

Re: sci-libs/scipy -> dev-python/scipy ?

On Tue, 08 Jul 2008 18:34:46 +0200
"Marijn Schouten (hkBst)" <hkBst <at> gentoo.org> wrote:
> I suppose you mean git. Since it tracks content and not files, moves
> are trivial. Git actually finds your moves for you, after you've
> moved content around; such as when doing a bump.

Ever tried git on an ebuild repository? Ebuilds are sufficiently
similar to each other that it quite often gets this horribly wrong. And
to make matters worse, there's no way of overriding it when it does.

--

-- 
Ciaran McCreesh

headers

Marijn Schouten (hkBst | 9 Jul 12:07

Re: sci-libs/scipy -> dev-python/scipy ?


Ciaran McCreesh wrote:
> On Tue, 08 Jul 2008 18:34:46 +0200
> "Marijn Schouten (hkBst)" <hkBst <at> gentoo.org> wrote:
>> I suppose you mean git. Since it tracks content and not files, moves
>> are trivial. Git actually finds your moves for you, after you've
>> moved content around; such as when doing a bump.
> 
> Ever tried git on an ebuild repository? Ebuilds are sufficiently
> similar to each other that it quite often gets this horribly wrong. And
> to make matters worse, there's no way of overriding it when it does.

Yes, we have a git overlay. I haven't noticed it getting it wrong yet.

Marijn

--
Marijn Schouten (hkBst), Gentoo Lisp project, Gentoo ML
<http://www.gentoo.org/proj/en/lisp/>, #gentoo-{lisp,ml} on FreeNode

headers

Donnie Berkholz | 9 Jul 18:41

Re: sci-libs/scipy -> dev-python/scipy ?

On 12:07 Wed 09 Jul     , Marijn Schouten (hkBst) wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Ciaran McCreesh wrote:
> > On Tue, 08 Jul 2008 18:34:46 +0200
> > "Marijn Schouten (hkBst)" <hkBst <at> gentoo.org> wrote:
> >> I suppose you mean git. Since it tracks content and not files, moves
> >> are trivial. Git actually finds your moves for you, after you've
> >> moved content around; such as when doing a bump.
> > 
> > Ever tried git on an ebuild repository? Ebuilds are sufficiently
> > similar to each other that it quite often gets this horribly wrong. And
> > to make matters worse, there's no way of overriding it when it does.
> 
> Yes, we have a git overlay. I haven't noticed it getting it wrong yet.

I suspect the problematic case might be two pretty basic template 
ebuilds of different packages. This would most likely be a problem when 
one is added and another is removed in the same commit, although that's 
the wrong workflow. I'm not sure whether it would also be a problem with 
multiple commits. I suppose you'd see the problem when you were trying 
to track the passage of certain code blocks across multiple commits 
(pickaxe).

--

-- 
Thanks,
Donnie

Donnie Berkholz

(Continue reading)

headers

Robert Buchholz | 8 Jul 19:59

[OT] Re: sci-libs/scipy -> dev-python/scipy ?

On Tuesday 08 July 2008, Joe Peterson wrote:
> Donnie Berkholz wrote:
> > I meant moves were largely pointless, although categories are to a
> > lesser extent. Tags would be a lot better, since nothing can be
> > categorized perfectly into a single place.
>
> Yes, I can see the benefit of a tag paradigm.  I, myself, find it
> more trouble than benefit to have the extra directory level.  I often
> do "cd /usr/portage/*/foo" to get to the foo package, and it often
> gets a hit in licenses or elsewhere that trips up this practice...

You can avoid the issue with the license directory by appending a / at 
the end. Grobian showed me that a function is useful for this, I just 
do "ecd xorg-server"

$ grep -A 3 ecd ~/.bashrc
function ecd () {
      cd ~/devel/gentoo/gentoo-x86/*/$@/
}

You could even change the /*/ to /*-*/ to exclude packages or profiles.

Robert

headers

Fabian Groffen | 8 Jul 20:28

Re: [OT] ecd function

On 08-07-2008 19:59:05 +0200, Robert Buchholz wrote:
> You can avoid the issue with the license directory by appending a / at 
> the end. Grobian showed me that a function is useful for this, I just 
> do "ecd xorg-server"
> 
> $ grep -A 3 ecd ~/.bashrc
> function ecd () {
>       cd ~/devel/gentoo/gentoo-x86/*/$@/
> }

errr, just because my name is involved...

% alias ecd
cd $EPREFIX/usr/portage/*-*/!*

I have it as alias, not as function.  If you have a function you can
(and should) actually do more magic, like:

ecd() {
	[[ -z $1 ]] && return
	p=( $(echo /usr/portage/*/$1) )
	if [[ ${p[*]} == *"/*/"* ]] ; then
		echo "no such package: $1" > /dev/stderr
	elif [[ ${#p[@]} > 1 ]] ; then
		echo "multiple options: ${p[*]}"
	else
		# we can't handle spaces in the paths here anyway
		cd ${p[*]}
	fi
}

(Continue reading)

headers

Ryan Hill | 9 Jul 03:14

Re: sci-libs/scipy -> dev-python/scipy ?

On Mon, 7 Jul 2008 21:02:37 -0700
Donnie Berkholz <dberkholz <at> gentoo.org> wrote:

> I don't think it's worth losing track of the CVS history just so we
> can have something in a different place that ultimately is hardly
> useful to anyone.

Maybe it's time to test the feasibility of moving to SVN again?  What
were the blockers last time?

--

-- 
gcc-porting,                                      by design, by neglect
treecleaner,                              for a fact or just for effect
wxwidgets @ gentoo     EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662

headers

Donnie Berkholz | 9 Jul 05:49

Re: Re: sci-libs/scipy -> dev-python/scipy ?

On 19:14 Tue 08 Jul     , Ryan Hill wrote:
> On Mon, 7 Jul 2008 21:02:37 -0700
> Donnie Berkholz <dberkholz <at> gentoo.org> wrote:
> 
> > I don't think it's worth losing track of the CVS history just so we
> > can have something in a different place that ultimately is hardly
> > useful to anyone.
> 
> Maybe it's time to test the feasibility of moving to SVN again?  What
> were the blockers last time?

The blocker was that it wasn't distributed or offline, and there's not 
enough benefit to move to it when such better ones exist now.

--

-- 
Thanks,
Donnie

Donnie Berkholz
Developer, Gentoo Linux
Blog: http://dberkholz.wordpress.com

headers

Ryan Hill | 10 Jul 05:36

Re: sci-libs/scipy -> dev-python/scipy ?

On Tue, 8 Jul 2008 20:49:43 -0700
Donnie Berkholz <dberkholz <at> gentoo.org> wrote:

> On 19:14 Tue 08 Jul     , Ryan Hill wrote:
> > On Mon, 7 Jul 2008 21:02:37 -0700
> > Donnie Berkholz <dberkholz <at> gentoo.org> wrote:
> > 
> > > I don't think it's worth losing track of the CVS history just so
> > > we can have something in a different place that ultimately is
> > > hardly useful to anyone.
> > 
> > Maybe it's time to test the feasibility of moving to SVN again?
> > What were the blockers last time?
> 
> The blocker was that it wasn't distributed or offline, and there's
> not enough benefit to move to it when such better ones exist now.

My fear is that deciding on another VCS to use that is so different in
headspace than CVS will just degrade again into a holy war over who's
favourite to pick and nothing will ever get done.  If everyone agrees
git is awesome and actually benefits our common workflow and isn't just
the hip thing to be switching to these days, then I'm all for it. If
we're going to argue the pros and cons endlessly and still be using CVS
in 2010 then I'd really rather just do SVN now, which I admit would give
relatively minor benefits, but also have minor costs as far as learning
curve and flamemail-generation go.

Whatever we eventually switch to I'm behind 100%.  Just kill CVS
already.  It's eating the children.

(Continue reading)

headers

Alec Warner | 9 Jul 06:32

Re: Re: sci-libs/scipy -> dev-python/scipy ?

On Tue, Jul 8, 2008 at 6:14 PM, Ryan Hill <dirtyepic <at> gentoo.org> wrote:
> On Mon, 7 Jul 2008 21:02:37 -0700
> Donnie Berkholz <dberkholz <at> gentoo.org> wrote:
>
>> I don't think it's worth losing track of the CVS history just so we
>> can have something in a different place that ultimately is hardly
>> useful to anyone.
>
> Maybe it's time to test the feasibility of moving to SVN again?  What
> were the blockers last time?

There weren't any technological blockers other than it's size and slowness.

It was felt that we could get to git fairly quickly and git was teh awesome.

-Alec

>
>
> --
> gcc-porting,                                      by design, by neglect
> treecleaner,                              for a fact or just for effect
> wxwidgets @ gentoo     EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662
>
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Alec Warner | 8 Jul 06:14

Re: sci-libs/scipy -> dev-python/scipy ?

On Mon, Jul 7, 2008 at 8:51 PM, Joe Peterson <lavajoe <at> gentoo.org> wrote:
> Donnie Berkholz wrote:
>> I actually object to having crap in dev-python, because things should be
>> categorized functionally instead of by the language they're implemented
>> in. 90% of the time you don't care about the language. But category
>> moves are pretty much pointless, so I don't normally bring it up.
>
> Do you mean it is pointless because categories are pointless, or because
> it is not worth the trouble of doing the move?  I assume we inherited
> the category idea from fbsd ports.

It is pointless because we should probably have tags; not categories.
It is akin to the Section[1] header in a debian control file.

[1] http://www.debian.org/doc/debian-policy/ch-archive.html#s-subsections

>
> Since we have categories (and assuming we'll keep them), I think that
> things should be categorized as correctly as possible (perhaps it's not
> the highest priority, but better to have it right than not, eventually).

Have you moved a package between categories before?

It is not simple to do correctly which is why no one likes to do them
unless they
are 'necessary'.

'epkgmove' invokes bad memories for many ;)

>

(Continue reading)

headers

Josh Saddler | 8 Jul 09:31

Re: Tags vs. categories (was: sci-libs/scipy -> dev-python/scip)

Alec Warner wrote:
> On Mon, Jul 7, 2008 at 8:51 PM, Joe Peterson <lavajoe <at> gentoo.org> wrote:
>> Donnie Berkholz wrote:
>>> I actually object to having crap in dev-python, because things should be
>>> categorized functionally instead of by the language they're implemented
>>> in. 90% of the time you don't care about the language. But category
>>> moves are pretty much pointless, so I don't normally bring it up.
>> Do you mean it is pointless because categories are pointless, or because
>> it is not worth the trouble of doing the move?  I assume we inherited
>> the category idea from fbsd ports.
> 
> It is pointless because we should probably have tags; not categories.
> It is akin to the Section[1] header in a debian control file.
> 
> [1] http://www.debian.org/doc/debian-policy/ch-archive.html#s-subsections

Tags instead of categories . . . Now here's a very interesting idea, 
indeed. Has there ever been a proposal like this for Gentoo?

I think we could improve on the Debian way of doing (sub)sections And I 
  think that a good system of tags would do better than most distros 
which have a fairly limited set of arbitrary categories (like desktop, 
system, utils; who knows what the heck those last two mean, anyway?) But 
blog-style multiple tags might be very, very nice, if we could agree on 
a set of tags to use, without trapping ourselves into some of the 
weirder categorization used by other distros, like Slackware's arcane 
alphabetic system.

Tags . . . I like the idea. I like it a lot. Thoughts? Exciting? Or is 
it an old issue, and I'm 5 years late to the party. :)

(Continue reading)

headers

Santiago M. Mola | 8 Jul 13:05

Re: Tags vs. categories (was: sci-libs/scipy -> dev-python/scip)

On Tue, Jul 8, 2008 at 9:31 AM, Josh Saddler <nightmorph <at> gentoo.org> wrote:
>
> Tags instead of categories . . . Now here's a very interesting idea, indeed.
> Has there ever been a proposal like this for Gentoo?
>

>
> Tags . . . I like the idea. I like it a lot. Thoughts? Exciting? Or is it an
> old issue, and I'm 5 years late to the party. :)
>

Exherbo will change its category system at some point. There has been
some discussion about it which may be interesting for people thinking
about possible replacements in Gentoo:
http://lists.exherbo.org/pipermail/exherbo-dev/2008-May/000149.html

There was similar discussions in @g-dev but I'm too lazy to start
serching the relevant threads.

Best regards,
-- 
Santiago M. Mola
Jabber ID: cooldwind <at> gmail.com
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Alec Warner | 9 Jul 06:28

Re: Tags vs. categories (was: sci-libs/scipy -> dev-python/scip)

On Tue, Jul 8, 2008 at 12:31 AM, Josh Saddler <nightmorph <at> gentoo.org> wrote:
> Alec Warner wrote:
>>
>> On Mon, Jul 7, 2008 at 8:51 PM, Joe Peterson <lavajoe <at> gentoo.org> wrote:
>>>
>>> Donnie Berkholz wrote:
>>>>
>>>> I actually object to having crap in dev-python, because things should be
>>>> categorized functionally instead of by the language they're implemented
>>>> in. 90% of the time you don't care about the language. But category
>>>> moves are pretty much pointless, so I don't normally bring it up.
>>>
>>> Do you mean it is pointless because categories are pointless, or because
>>> it is not worth the trouble of doing the move?  I assume we inherited
>>> the category idea from fbsd ports.
>>
>> It is pointless because we should probably have tags; not categories.
>> It is akin to the Section[1] header in a debian control file.
>>
>> [1] http://www.debian.org/doc/debian-policy/ch-archive.html#s-subsections
>
> Tags instead of categories . . . Now here's a very interesting idea, indeed.
> Has there ever been a proposal like this for Gentoo?

many times.

>
> I think we could improve on the Debian way of doing (sub)sections And I
>  think that a good system of tags would do better than most distros which
> have a fairly limited set of arbitrary categories (like desktop, system,

(Continue reading)

headers

Josh Saddler | 9 Jul 06:31

Re: Tags vs. categories

>> Tags . . . I like the idea. I like it a lot. Thoughts? Exciting? Or is it an
>> old issue, and I'm 5 years late to the party. :)
> 
> Probably more than 5...

Well, that's not very helpful. Got any links? My archives.g.o-fu has 
failed me.

headers

Diego 'Flameeyes' Pettenò | 8 Jul 11:59

Re: sci-libs/scipy -> dev-python/scipy ?

"Alec Warner" <antarus <at> gentoo.org> writes:

> 'epkgmove' invokes bad memories for many ;)

Thanks for ruining the few hours of sleep I'll get in the next days by
bringing that name up...

J/K

--

-- 
Diego "Flameeyes" Pettenò
http://blog.flameeyes.eu/

headers

Andrey Grozin | 8 Jul 06:03

Re: sci-libs/scipy -> dev-python/scipy ?

On Mon, 7 Jul 2008, Donnie Berkholz wrote:
> I actually object to having crap in dev-python, because things should be
> categorized functionally instead of by the language they're implemented
> in. 90% of the time you don't care about the language. But category
> moves are pretty much pointless, so I don't normally bring it up.
Then this particular case belongs to the other 10% 
It is not really important for a user if a library is written in C or 
fortran, because he can call it from his own programs written in any 
language. But python modules are only useful for somebody who is going to 
write his own python code and import them.

sci-libs/scipi and dev-python/scientificpython are two competing projects 
which have practically identical aims and descriptions. Do you think this 
is logical?

Andrey
--

-- 
gentoo-dev <at> lists.gentoo.org mailing list

headers

Donnie Berkholz | 8 Jul 06:35

Re: sci-libs/scipy -> dev-python/scipy ?

On 04:03 Tue 08 Jul     , Andrey Grozin wrote:
> On Mon, 7 Jul 2008, Donnie Berkholz wrote:
>> I actually object to having crap in dev-python, because things should be
>> categorized functionally instead of by the language they're implemented
>> in. 90% of the time you don't care about the language. But category
>> moves are pretty much pointless, so I don't normally bring it up.
> Then this particular case belongs to the other 10% 
> It is not really important for a user if a library is written in C or  
> fortran, because he can call it from his own programs written in any  
> language. But python modules are only useful for somebody who is going to 
> write his own python code and import them.

Right, sure. Here's the way I see it: chances are that someone realizes 
they want to write a scientific application, and then look to see what 
libraries are available to write it with, instead of deciding they want 
to write some sort of library for an unknown purpose using Python and 
then say, "Hey, I think I'll do something scientific."

> sci-libs/scipi and dev-python/scientificpython are two competing projects 
> which have practically identical aims and descriptions. Do you think this 
> is logical?

Logical? Not particularly. Does it matter, when considering the costs of 
moving packages in CVS and how search tools work? Not particularly.

--

-- 
Thanks,
Donnie

Donnie Berkholz

(Continue reading)