Discussion:
scikits contribution?
(too old to reply)
Jonathan Stickel
2010-07-14 17:36:57 UTC
Permalink
Date: Thu, 8 Jul 2010 08:24:25 -0400
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
Some time ago, I offered to contribute (on the scipy-user list) some
http://mail.scipy.org/pipermail/scipy-user/2010-February/024351.html
http://mail.scipy.org/pipermail/scipy-user/2010-February/024408.html
So I am finally looking into scikits, and I am not sure how to proceed.
?My code consists of several functions in a single .py file. ?It seems
overkill to create a new scikit for just one file, but I do not see an
existing scikit that matches. ?'Optimization' would be the closest; in
core scipy I would put it in 'interpolate'.
So, what is the minimum that I need to do to create a scikit and upload
my code? ?Any suggestions for the name of the scikit (interpolate,
data_smoothing)?
The easiest to get started is to copy the setup structure from another scikit.
I think the template scikit in scikits svn is a bit out of date, the
last time I looked.
If you think your model could form the basis for enhancing the
smoother or noisy interpolation category in scipy, then a scikits
would be the best way, as we discussed.
If you want to add it to an existing scikits, then statsmodels would
be a possibility.
Although statsmodels is more oriented towards multivariate approaches,
I think a smoother category, together with some non-parametric
methods, e.g. the existing kernel regression, would be an appropriate
fit. There is a need for smoothers in gam, Generalized Additive
Models, but that one is not cleaned up yet.
And I think there will be more applications where it would be useful
to share the cross-validation code as far as possible.
Josef
Please know that I am just starting to learn python, being a convert
from matlab/octave. ?Although I have become fairly proficient using
numpy/scipy in ipython, I do not know much about python internals,
setuptools, etc.
OK, I created a scikit named "datasmooth" and included my current code.
It seems to install OK with "python setup install" and import
correctly. However, I am not able to commit to the svn repository. I
registered on the scikits wiki, but I guess there is something else I
need to do?

Thanks,
Jonathan

P.S.
Please cc me in your reply since I receive list emails in digest form.
j***@gmail.com
2010-07-14 17:57:49 UTC
Permalink
Post by Jonathan Stickel
Date: Thu, 8 Jul 2010 08:24:25 -0400
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
 Some time ago, I offered to contribute (on the scipy-user list) some
 http://mail.scipy.org/pipermail/scipy-user/2010-February/024351.html
 http://mail.scipy.org/pipermail/scipy-user/2010-February/024408.html
 So I am finally looking into scikits, and I am not sure how to proceed.
 ?My code consists of several functions in a single .py file. ?It seems
 overkill to create a new scikit for just one file, but I do not see an
 existing scikit that matches. ?'Optimization' would be the closest; in
 core scipy I would put it in 'interpolate'.
 So, what is the minimum that I need to do to create a scikit and upload
 my code? ?Any suggestions for the name of the scikit (interpolate,
 data_smoothing)?
The easiest to get started is to copy the setup structure from another scikit.
I think the template scikit in scikits svn is a bit out of date, the
last time I looked.
If you think your model could form the basis for enhancing the
smoother or noisy interpolation category in scipy, then a scikits
would be the best way, as we discussed.
If you want to add it to an existing scikits, then statsmodels would
be a possibility.
Although statsmodels is more oriented towards multivariate approaches,
I think a smoother category, together with some non-parametric
methods, e.g. the existing kernel regression, would be an appropriate
fit. There is a need for smoothers in gam, Generalized Additive
Models, but that one is not cleaned up yet.
And I think there will be more applications where it would be useful
to share the cross-validation code as far as possible.
Josef
 Please know that I am just starting to learn python, being a convert
 from matlab/octave. ?Although I have become fairly proficient using
 numpy/scipy in ipython, I do not know much about python internals,
 setuptools, etc.
OK, I created a scikit named "datasmooth" and included my current code.
 It seems to install OK with "python setup install" and import
correctly.  However, I am not able to commit to the svn repository.  I
registered on the scikits wiki, but I guess there is something else I
need to do?
Sharing the code would be much easier if you pick your favorite
decentralized revision control system, git, bazaar or mercurial, and
host it at the corresponding website.
It would also avoid any permission questions. I don't know who handles
setup and administration of http://projects.scipy.org/scikits.

Josef
Post by Jonathan Stickel
Thanks,
Jonathan
P.S.
Please cc me in your reply since I receive list emails in digest form.
_______________________________________________
SciPy-Dev mailing list
http://mail.scipy.org/mailman/listinfo/scipy-dev
Jonathan Stickel
2010-07-14 20:59:17 UTC
Permalink
forgot to CC
Josef
---------- Forwarded message ----------
Date: Wed, Jul 14, 2010 at 1:57 PM
Subject: Re: [SciPy-Dev] scikits contribution?
Post by Jonathan Stickel
Date: Thu, 8 Jul 2010 08:24:25 -0400
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
Some time ago, I offered to contribute (on the scipy-user list) some
http://mail.scipy.org/pipermail/scipy-user/2010-February/024351.html
http://mail.scipy.org/pipermail/scipy-user/2010-February/024408.html
So I am finally looking into scikits, and I am not sure how to proceed.
?My code consists of several functions in a single .py file. ?It seems
overkill to create a new scikit for just one file, but I do not see an
existing scikit that matches. ?'Optimization' would be the closest; in
core scipy I would put it in 'interpolate'.
So, what is the minimum that I need to do to create a scikit and upload
my code? ?Any suggestions for the name of the scikit (interpolate,
data_smoothing)?
The easiest to get started is to copy the setup structure from another scikit.
I think the template scikit in scikits svn is a bit out of date, the
last time I looked.
If you think your model could form the basis for enhancing the
smoother or noisy interpolation category in scipy, then a scikits
would be the best way, as we discussed.
If you want to add it to an existing scikits, then statsmodels would
be a possibility.
Although statsmodels is more oriented towards multivariate approaches,
I think a smoother category, together with some non-parametric
methods, e.g. the existing kernel regression, would be an appropriate
fit. There is a need for smoothers in gam, Generalized Additive
Models, but that one is not cleaned up yet.
And I think there will be more applications where it would be useful
to share the cross-validation code as far as possible.
Josef
Please know that I am just starting to learn python, being a convert
from matlab/octave. ?Although I have become fairly proficient using
numpy/scipy in ipython, I do not know much about python internals,
setuptools, etc.
OK, I created a scikit named "datasmooth" and included my current code.
It seems to install OK with "python setup install" and import
correctly. However, I am not able to commit to the svn repository. I
registered on the scikits wiki, but I guess there is something else I
need to do?
Sharing the code would be much easier if you pick your favorite
decentralized revision control system, git, bazaar or mercurial, and
host it at the corresponding website.
It would also avoid any permission questions. I don't know who handles
setup and administration of http://projects.scipy.org/scikits.
It seems strange to me that each scikit would host its own development
sources in separate locations! I'd prefer to use the existing SVN tree
for my small contribution, if possible.

I did register at http://projects.scipy.org/scikits with the username
"jjstickel". The wiki page indicates that this is all that is needed to
edit the wiki, but I do not see a "Edit this page" link. So it does
seem that someone needs to give me permission. I also need permission
for commit access to svn.scipy.org/svn/scikits/.

If someone wants to see some code before giving out permissions willy
nilly, I can do that of course! Just let me know.

Thanks,
Jonathan
Post by Jonathan Stickel
P.S.
Please cc me in your reply since I receive list emails in digest form.
Jarrod Millman
2010-07-14 21:50:09 UTC
Permalink
Post by Jonathan Stickel
I did register at http://projects.scipy.org/scikits with the username
"jjstickel".  The wiki page indicates that this is all that is needed to
edit the wiki, but I do not see a "Edit this page" link.  So it does
seem that someone needs to give me permission.  I also need permission
for commit access to svn.scipy.org/svn/scikits/.
Hey Jonathan,

Thanks for contributing a new scikit!

Unfortunately, it is a bit tedious to create accounts on the existing
subversion system and there is no way to separate developer accounts
between the various scikits. If you really want to use subversion, I
would create a sourceforge project. For example, that is where the
scikits.learn project is hosted:
http://sourceforge.net/projects/scikit-learn
Otherwise, I would recommend using github or one of the other dvcs
hosting sites.

In the near future (over the next month or so) it looks increasingly
likely that we will be moving numpy and possibly scipy to git/github,
so the existing scikits subversion repository will probably get moved
over and broken up as well.

The official scikits registry is here: http://scikits.appspot.com/
Any package named "scikits.xyz" on the Python Packaging Index will be
included on the scikits.appspot.com website automatically. So there
is no advantage to having commit access to the subversion site and
there are several disadvantages such as it is slow, there isn't an
obvious way to get account access, and it may disappear soon (assuming
we move to github).

Best,
Jarrod
Jonathan Stickel
2010-07-14 22:01:07 UTC
Permalink
Date: Wed, 14 Jul 2010 14:50:09 -0700
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
I did register athttp://projects.scipy.org/scikits with the username
"jjstickel". ?The wiki page indicates that this is all that is needed to
edit the wiki, but I do not see a "Edit this page" link. ?So it does
seem that someone needs to give me permission. ?I also need permission
for commit access to svn.scipy.org/svn/scikits/.
Hey Jonathan,
Thanks for contributing a new scikit!
Unfortunately, it is a bit tedious to create accounts on the existing
subversion system and there is no way to separate developer accounts
between the various scikits. If you really want to use subversion, I
would create a sourceforge project. For example, that is where the
http://sourceforge.net/projects/scikit-learn
Otherwise, I would recommend using github or one of the other dvcs
hosting sites.
In the near future (over the next month or so) it looks increasingly
likely that we will be moving numpy and possibly scipy to git/github,
so the existing scikits subversion repository will probably get moved
over and broken up as well.
The official scikits registry is here:http://scikits.appspot.com/
Any package named "scikits.xyz" on the Python Packaging Index will be
included on the scikits.appspot.com website automatically. So there
is no advantage to having commit access to the subversion site and
there are several disadvantages such as it is slow, there isn't an
obvious way to get account access, and it may disappear soon (assuming
we move to github).
OK, I'll check out github, or, if the learning curve is high, I may fall
back to sourceforge since I have more experience with that interface.

Can you elaborate on "Any package named "scikits.xyz" on the Python
Packaging Index will be included on the scikits.appspot.com website
automatically"? Where is the Python Packaging Index and how do I get a
scikit to show up there once I make the sources available?

I have to say this has become a lot more involved than I initially
expected. Simpler procedures and better "howto" documentation would be
helpful for sideline contributors like myself. Maybe this will improve
after the move to github?

Thanks,
Jonathan
Skipper Seabold
2010-07-14 22:06:19 UTC
Permalink
Post by Jonathan Stickel
Date: Wed, 14 Jul 2010 14:50:09 -0700
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
 I did register athttp://projects.scipy.org/scikits  with the username
 "jjstickel". ?The wiki page indicates that this is all that is needed to
 edit the wiki, but I do not see a "Edit this page" link. ?So it does
 seem that someone needs to give me permission. ?I also need permission
 for commit access to svn.scipy.org/svn/scikits/.
Hey Jonathan,
Thanks for contributing a new scikit!
Unfortunately, it is a bit tedious to create accounts on the existing
subversion system and there is no way to separate developer accounts
between the various scikits.  If you really want to use subversion, I
would create a sourceforge project.  For example, that is where the
http://sourceforge.net/projects/scikit-learn
Otherwise, I would recommend using github or one of the other dvcs
hosting sites.
In the near future (over the next month or so) it looks increasingly
likely that we will be moving numpy and possibly scipy to git/github,
so the existing scikits subversion repository will probably get moved
over and broken up as well.
The official scikits registry is here:http://scikits.appspot.com/
Any package named "scikits.xyz" on the Python Packaging Index will be
included on the scikits.appspot.com website automatically.  So there
is no advantage to having commit access to the subversion site and
there are several disadvantages such as it is slow, there isn't an
obvious way to get account access, and it may disappear soon (assuming
we move to github).
OK, I'll check out github, or, if the learning curve is high, I may fall
back to sourceforge since I have more experience with that interface.
Can you elaborate on "Any package named "scikits.xyz" on the Python
Packaging Index will be included on the scikits.appspot.com website
automatically"?  Where is the Python Packaging Index and how do I get a
http://pypi.python.org/pypi
Post by Jonathan Stickel
scikit to show up there once I make the sources available?
See below.
Post by Jonathan Stickel
I have to say this has become a lot more involved than I initially
expected.  Simpler procedures and better "howto" documentation would be
helpful for sideline contributors like myself.  Maybe this will improve
after the move to github?
I think most all of your questions are answered here.

http://www.scipy.org/scipy/scikits/wiki/ScikitsForDevelopers

Skipper
Skipper Seabold
2010-07-14 22:08:28 UTC
Permalink
Post by Skipper Seabold
Post by Jonathan Stickel
Date: Wed, 14 Jul 2010 14:50:09 -0700
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
 I did register athttp://projects.scipy.org/scikits  with the username
 "jjstickel". ?The wiki page indicates that this is all that is needed to
 edit the wiki, but I do not see a "Edit this page" link. ?So it does
 seem that someone needs to give me permission. ?I also need permission
 for commit access to svn.scipy.org/svn/scikits/.
Hey Jonathan,
Thanks for contributing a new scikit!
Unfortunately, it is a bit tedious to create accounts on the existing
subversion system and there is no way to separate developer accounts
between the various scikits.  If you really want to use subversion, I
would create a sourceforge project.  For example, that is where the
http://sourceforge.net/projects/scikit-learn
Otherwise, I would recommend using github or one of the other dvcs
hosting sites.
In the near future (over the next month or so) it looks increasingly
likely that we will be moving numpy and possibly scipy to git/github,
so the existing scikits subversion repository will probably get moved
over and broken up as well.
The official scikits registry is here:http://scikits.appspot.com/
Any package named "scikits.xyz" on the Python Packaging Index will be
included on the scikits.appspot.com website automatically.  So there
is no advantage to having commit access to the subversion site and
there are several disadvantages such as it is slow, there isn't an
obvious way to get account access, and it may disappear soon (assuming
we move to github).
OK, I'll check out github, or, if the learning curve is high, I may fall
back to sourceforge since I have more experience with that interface.
Can you elaborate on "Any package named "scikits.xyz" on the Python
Packaging Index will be included on the scikits.appspot.com website
automatically"?  Where is the Python Packaging Index and how do I get a
http://pypi.python.org/pypi
Post by Jonathan Stickel
scikit to show up there once I make the sources available?
See below.
Post by Jonathan Stickel
I have to say this has become a lot more involved than I initially
expected.  Simpler procedures and better "howto" documentation would be
helpful for sideline contributors like myself.  Maybe this will improve
after the move to github?
I think most all of your questions are answered here.
http://www.scipy.org/scipy/scikits/wiki/ScikitsForDevelopers
Up to date link: http://projects.scipy.org/scikits/wiki/ScikitsForDevelopers

Skipper
Jarrod Millman
2010-07-14 22:14:32 UTC
Permalink
Post by Skipper Seabold
Up to date link: http://projects.scipy.org/scikits/wiki/ScikitsForDevelopers
Unfortunately, that link is out dated. It says the all use a common
svn repo, which is no longer true and is no longer recommended. It
also doesn't say anything about PyPI or the scikits.appspot.com site.

Jarrod
Jarrod Millman
2010-07-14 22:15:57 UTC
Permalink
Unfortunately, that link is out dated.  It says the all use a common
svn repo, which is no longer true and is no longer recommended.  It
also doesn't say anything about PyPI or the scikits.appspot.com site.
python setup.py register sdist bdist_egg upload -s
Will register to PyPI, the python package index and upload a gpg
signed (-s) source (sdist) and binary (bdist) distribution of your
project.
j***@gmail.com
2010-07-14 22:16:28 UTC
Permalink
Post by Skipper Seabold
Post by Jonathan Stickel
Date: Wed, 14 Jul 2010 14:50:09 -0700
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
 I did register athttp://projects.scipy.org/scikits  with the username
 "jjstickel". ?The wiki page indicates that this is all that is needed to
 edit the wiki, but I do not see a "Edit this page" link. ?So it does
 seem that someone needs to give me permission. ?I also need permission
 for commit access to svn.scipy.org/svn/scikits/.
Hey Jonathan,
Thanks for contributing a new scikit!
Unfortunately, it is a bit tedious to create accounts on the existing
subversion system and there is no way to separate developer accounts
between the various scikits.  If you really want to use subversion, I
would create a sourceforge project.  For example, that is where the
http://sourceforge.net/projects/scikit-learn
Otherwise, I would recommend using github or one of the other dvcs
hosting sites.
In the near future (over the next month or so) it looks increasingly
likely that we will be moving numpy and possibly scipy to git/github,
so the existing scikits subversion repository will probably get moved
over and broken up as well.
The official scikits registry is here:http://scikits.appspot.com/
Any package named "scikits.xyz" on the Python Packaging Index will be
included on the scikits.appspot.com website automatically.  So there
is no advantage to having commit access to the subversion site and
there are several disadvantages such as it is slow, there isn't an
obvious way to get account access, and it may disappear soon (assuming
we move to github).
OK, I'll check out github, or, if the learning curve is high, I may fall
back to sourceforge since I have more experience with that interface.
Can you elaborate on "Any package named "scikits.xyz" on the Python
Packaging Index will be included on the scikits.appspot.com website
automatically"?  Where is the Python Packaging Index and how do I get a
http://pypi.python.org/pypi
Post by Jonathan Stickel
scikit to show up there once I make the sources available?
See below.
Post by Jonathan Stickel
I have to say this has become a lot more involved than I initially
expected.  Simpler procedures and better "howto" documentation would be
helpful for sideline contributors like myself.  Maybe this will improve
after the move to github?
I think most all of your questions are answered here.
http://www.scipy.org/scipy/scikits/wiki/ScikitsForDevelopers
It's a bit outdated, since now hosting with a decentralized revision
control system is preferred.
Also nothing about sphinx docs, neither in the template (at least when
I checked and looked for a template last year)
Basic hosting instructions or links to instructions with alternatives
on github, launchpad, mercurial?
Instructions for creating account and uploading to pypi are also missing.

It took me quite a bit of time last year to create the scikits
infrastructure. So improvements to and updating of the scikits docs
would be very useful.

Josef
Post by Skipper Seabold
Skipper
_______________________________________________
SciPy-Dev mailing list
http://mail.scipy.org/mailman/listinfo/scipy-dev
Jarrod Millman
2010-07-14 22:11:55 UTC
Permalink
I have to say this has become a lot more involved than I initially expected.
 Simpler procedures and better "howto" documentation would be helpful for
sideline contributors like myself.  Maybe this will improve after the move
to github?
Absolutely. No one has really stepped up to own the scikits overall
project/structure per se. Some of the documentation is out-of-date
and it has changed some from what it was originally conceived.
Hopefully, over the next year someone will step forward and help with
this aspect of the project.

If you are considering moving to git/github, but haven't used it yet,
you may want to wait a week or so. We will be adding instructions on
how to use it for numpy, which may be helpful to you as well.

Best,
Jarrod
Jonathan Stickel
2010-07-15 19:15:16 UTC
Permalink
Date: Wed, 14 Jul 2010 15:15:57 -0700
Subject: Re: [SciPy-Dev] scikits contribution?
Content-Type: text/plain; charset=ISO-8859-1
Unfortunately, that link is out dated. ?It says the all use a common
svn repo, which is no longer true and is no longer recommended. ?It
also doesn't say anything about PyPI or the scikits.appspot.com site.
python setup.py register sdist bdist_egg upload -s
Will register to PyPI, the python package index and upload a gpg
signed (-s) source (sdist) and binary (bdist) distribution of your
project.
Thanks for all the responses. I found github quite easy to use and with
good site documentation. Here is the source repository:

http://github.com/jjstickel/scikit-datasmooth/

When I attempted to register and upload to PyPI (after creating a gpg
key), I got an error:


$ python setup.py register sdist bdist_egg upload -s
running register
running egg_info
running build_src
...
running upload
gpg --detach-sign -a dist/scikits.datasmooth-0.5.tar.gz

You need a passphrase to unlock the secret key for
user: "Jonathan Stickel <***@vcn.com>"
2048-bit RSA key, ID 30D37C66, created 2010-07-15

Traceback (most recent call last):
File "setup.py", line 53, in <module>
'Topic :: Scientific/Engineering'])
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/site-packages/numpy/distutils/core.py",
line 186, in setup
return old_setup(**new_attr)
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/distutils/core.py",
line 152, in setup
dist.run_commands()
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/distutils/dist.py",
line 987, in run_commands
self.run_command(cmd)
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/distutils/dist.py",
line 1007, in run_command
cmd_obj.run()
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/distutils/command/upload.py",
line 57, in run
self.upload_file(command, pyversion, filename)
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/distutils/command/upload.py",
line 133, in upload_file
value = value[1]
IndexError: tuple index out of range


It looks like the index was created OK, but no source or binary package
was uploaded. Any help on this?

Thanks,
Jonathan
Matthieu Brucher
2010-07-15 20:43:36 UTC
Permalink
Post by Jonathan Stickel
It looks like the index was created OK, but no source or binary package
was uploaded.  Any help on this?
It couldn't be uploaded: "You need a passphrase to unlock the secret
key for". You may have a passphrase with your ssh key :|

Matthieu
--
Information System Engineer, Ph.D.
Blog: http://matt.eifelle.com
LinkedIn: http://www.linkedin.com/in/matthieubrucher
Pauli Virtanen
2010-07-15 20:02:52 UTC
Permalink
Thu, 15 Jul 2010 13:15:16 -0600, Jonathan Stickel wrote:
[clip]
Post by Jonathan Stickel
$ python setup.py register sdist bdist_egg upload -s
[clip: distutils error]
Post by Jonathan Stickel
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/
python2.6/distutils/command/upload.py",
Post by Jonathan Stickel
line 133, in upload_file
IndexError: tuple index out of range
It looks like the index was created OK, but no source or binary package
was uploaded. Any help on this?
I think I remember seeing this too. This is some distutils bug, I think...

My "fix" was just to edit the "distutils/command/upload.py" and replace

value = value[1]

on line 133 by the obvious

if len(value) > 1:
value = value[1]
else:
value = ""

and hope for the best. Or maybe I ran "upload" on a separate line from
the "sdist" and "bdist".

This is maybe not how it should go.

You can also upload the tarballs on the PyPi website.
--
Pauli Virtanen
Jonathan Stickel
2010-07-15 21:47:50 UTC
Permalink
Date: Thu, 15 Jul 2010 20:02:52 +0000 (UTC)
Subject: Re: [SciPy-Dev] scikits contribution?
Jonathan Stickel wrote: [clip]
Post by Jonathan Stickel
$ python setup.py register sdist bdist_egg upload -s
[clip: distutils error]
Post by Jonathan Stickel
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/
python2.6/distutils/command/upload.py",
Post by Jonathan Stickel
line 133, in upload_file
IndexError: tuple index out of range
It looks like the index was created OK, but no source or binary package
was uploaded. Any help on this?
I think I remember seeing this too. This is some distutils bug, I think...
My "fix" was just to edit the "distutils/command/upload.py" and replace
value = value[1]
on line 133 by the obvious
value = value[1]
value = ""
and hope for the best. Or maybe I ran "upload" on a separate line from
the "sdist" and "bdist".
This is maybe not how it should go.
You can also upload the tarballs on the PyPi website.
Ah, this seems to be it. I uploaded the egg using the website.

OK, now that I have the sources on git and an egg on pypi, would some of
you like to try out the scikit and let me know what I did wrong?

Thanks,
Jonathan
j***@gmail.com
2010-07-16 05:02:18 UTC
Permalink
Date: Thu, 15 Jul 2010 20:02:52 +0000 (UTC)
 > Subject: Re: [SciPy-Dev] scikits contribution?
Jonathan Stickel wrote: [clip]
 $ python setup.py register sdist bdist_egg upload -s
[clip: distutils error]
 "/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/
python2.6/distutils/command/upload.py",
 line 133, in upload_file
 IndexError: tuple index out of range
 It looks like the index was created OK, but no source or binary package
 was uploaded.  Any help on this?
I think I remember seeing this too. This is some distutils bug, I think...
My "fix" was just to edit the "distutils/command/upload.py" and replace
      value = value[1]
on line 133 by the obvious
          value = value[1]
          value = ""
and hope for the best. Or maybe I ran "upload" on a separate line from
the "sdist" and "bdist".
This is maybe not how it should go.
You can also upload the tarballs on the PyPi website.
Ah, this seems to be it.  I uploaded the egg using the website.
It would be better to upload also the sdist,
Since scikits.datasmooth is pure python, an egg doesn't really have an
advantage and is specific to a python version, py2.6 only.
(I'm on py25)

Why do you have the code duplication in two modules and the
conditional import in __init__.py ?
try:
from regularsmooth import *
except ImportError as error:

What is the import error that might occur with regularsmooth ?

Also an example to quickly try out the scikit would be very useful.
(And tests and docs are not included yet.)

Thanks,

Josef
OK, now that I have the sources on git and an egg on pypi, would some of
you like to try out the scikit and let me know what I did wrong?
Thanks,
Jonathan
_______________________________________________
SciPy-Dev mailing list
http://mail.scipy.org/mailman/listinfo/scipy-dev
Jonathan Stickel
2010-07-16 17:19:16 UTC
Permalink
Subject: Re: [SciPy-Dev] scikits contribution? To: SciPy Developers List
Content-Type: text/plain; charset=ISO-8859-1 On Thu, Jul 15, 2010 at
Date: Thu, 15 Jul 2010 20:02:52 +0000 (UTC)
?> Subject: Re: [SciPy-Dev] scikits contribution?
Jonathan Stickel wrote: [clip]
?$ python setup.py register sdist bdist_egg upload -s
[clip: distutils error]
?"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/
python2.6/distutils/command/upload.py",
?line 133, in upload_file
?IndexError: tuple index out of range
?It looks like the index was created OK, but no source or binary package
?was uploaded. ?Any help on this?
I think I remember seeing this too. This is some distutils bug, I think...
My "fix" was just to edit the "distutils/command/upload.py" and replace
? ? ? value = value[1]
on line 133 by the obvious
? ? ? ? ? value = value[1]
? ? ? ? ? value = ""
and hope for the best. Or maybe I ran "upload" on a separate line from
the "sdist" and "bdist".
This is maybe not how it should go.
You can also upload the tarballs on the PyPi website.
Ah, this seems to be it. ?I uploaded the egg using the website.
It would be better to upload also the sdist,
Since scikits.datasmooth is pure python, an egg doesn't really have an
advantage and is specific to a python version, py2.6 only.
(I'm on py25)
Why do you have the code duplication in two modules and the
conditional import in __init__.py ?
from regularsmooth import *
What is the import error that might occur with regularsmooth ?
Also an example to quickly try out the scikit would be very useful.
(And tests and docs are not included yet.)
Thanks for the feedback. I was confused what was the "sdist", but I
think I figured it out and uploaded it.

I will get to examples, tests, and docs later when I have time (this
will require more learning for me). The functions are well documented
and the primary use function "smooth_data" shows an example.

There are two primary implementations: "smooth_data" that includes
cross-validation, and "smooth_data_constr" which will take constraints
but does not include cross-validation. The later requires the module
"cvxopt". If this is not available, I wanted to allow a user to still
use the unconstrained smoother. Maybe there is a better way to do this
than my "try/except" hack. Suggestions are welcome.

Jonathan
Jonathan Stickel
2010-07-16 18:09:56 UTC
Permalink
Post by j***@gmail.com
It would be better to upload also the sdist,
Post by j***@gmail.com
Since scikits.datasmooth is pure python, an egg doesn't really have an
advantage and is specific to a python version, py2.6 only.
(I'm on py25)
Why do you have the code duplication in two modules and the
conditional import in __init__.py ?
from regularsmooth import *
What is the import error that might occur with regularsmooth ?
Also an example to quickly try out the scikit would be very useful.
(And tests and docs are not included yet.)
Thanks for the feedback. I was confused what was the "sdist", but I think I
figured it out and uploaded it.
I will get to examples, tests, and docs later when I have time (this will
require more learning for me). The functions are well documented and the
primary use function "smooth_data" shows an example.
There are two primary implementations: "smooth_data" that includes
cross-validation, and "smooth_data_constr" which will take constraints but
does not include cross-validation. The later requires the module "cvxopt".
If this is not available, I wanted to allow a user to still use the
unconstrained smoother. Maybe there is a better way to do this than my
"try/except" hack. Suggestions are welcome.
Try/except are fine for this, but it wasn't very informative. You
could also try/except import cvxopt, or add a note that the full
version requires cvxopt. Maybe it's noted somewhere in your code, I
haven't looked that carefully.
Kind of noted in the README. I am thinking of putting the try/except in
the script itself rather than in __init__.py, but I will get to it later.
Is it possible to replace cvxopt with scipy optimizers? But this is
not necessary for a scikit.
I needed a quadratic programming (qp) implementation, and I couldn't
find any qp in scipy. I found the one in cvxopt to be quite good.

Jonathan

Continue reading on narkive:
Loading...