Please make a release? (+due questions re: maintenance) #348
Comments
|
Hi Ivan, thanks for bringing this up.
I will spend some time on hyperopt in the next couple of weeks. I can set
up e.g. travis and fix the tests. At that point a new pypi version would
make sense. Could I ask you for code review on those changes?
You are welcome to do the conda recipe, keep the list posted.
For the code refactoring and/or cleanup that could be great, lets have a
chance to discuss on list or github issues before anyone does a lot of work
to make sure the whole process to merge goes smoothly. Where would you like
to start?
On Dec 27, 2017 4:31 AM, "Ivan Smirnov" <notifications@github.com> wrote:
Dear authors/maintainers,
Would you please make a tarball release so that
1. Users installing it via pip install don't run into networkx 2.0
incompatibility which is fixed on master
2. We can build a conda-forge hyperopt recipe (
conda-forge/staged-recipes#4710
<conda-forge/staged-recipes#4710>) so that the
latest hyperopt is available via conda install
On a side note, as I've tried running the full test suite on all platforms
while building the conda-forge package, more than a dozen different tests
were failing on master. It looks like there's a rather significant amount
of work to be done e.g. to catch up with the latest numpy, plus fixing the
flaky tests. Is anyone planing to take on that? I could probably find time
to help out, but it would then involve *major* cleanup and refactoring so
the codebase/tests are easier/more pleasant to work with from development
standpoint which I'm not sure the authors are okay with. Ideally, there
should also automated tests, i.e. tox/travis setup, which would also
simplify the process of conda releases and reduce downstream surprises.
// @jaberg <https://github.com/jaberg>? @dwf <https://github.com/dwf>?
Thanks a million!
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#348>, or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAKdDLDL8Ga50IDS7I2LsV0OJ6m7N1iZks5tEg5fgaJpZM4RNVJy>
.
|
|
@jaberg Hi James, thank you for your reply. I'd be glad to do a code review, of course, or discuss how to fix the current test failures and/or help with setting up automated testing. As for the refactoring/cleanup -- I have quite a few ideas on how to clean things up in a major way to hopefully simplify future maintenance. I could try to sum it up in a separate GH issue in the coming days, if you would. Thanks! |
|
Sounds good, I will look out for those GH issues. If you remember, post
links to those issues to this thread too.
…On Dec 27, 2017 4:01 PM, "Ivan Smirnov" ***@***.***> wrote:
@jaberg <https://github.com/jaberg> Hi James, thank you for your reply.
I'd be glad to do a code review, of course, or discuss how to fix the
current test failures and/or help with setting up automated testing.
As for the refactoring/cleanup -- I have quite a few ideas on how to clean
things up in a major way to hopefully simplify future maintenance. I could
try to sum it up in a separate GH issue in the coming days, if you would.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#348 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAKdDCUh3puzdC2xXCvtPCBrrrGPDU5oks5tErAvgaJpZM4RNVJy>
.
|
|
Original question was to do a tarball release. What do you mean exactly? What would it take to make this done? |
|
github allow to get a tarball of the dev version. One example:
https://github.com/hyperopt/hyperopt/archive/master.zip
…On Wed, Feb 7, 2018 at 4:42 PM James Bergstra ***@***.***> wrote:
Original question was to do a tarball release. What do you mean exactly?
What would it take to make this done?
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#348 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AALC-4Mp4PhYRLREb0uD9zJwB5tz1vrCks5tShjXgaJpZM4RNVJy>
.
|
|
Really glad to see this project coming back to life. It's impressive how quickly dependency packages like numpy are updating, but also a bummer to see how frequently those changes are breaking. The last PyPi release for this package was just over a year ago, and yet it doesn't work any longer, due to how rapidly our space is moving. I've found it enormously helpful to pin supported version ranges for auto_ml. If someone releases a breaking change to a package we rely on (seems to happen every other month or so, with Keras being the most recent culprit), we just specify in our setup.py that we don't support that version yet. It's pretty trivial from there to cut a new release and push it to PyPi, and work on support later when we've got more time for it (or let a contributor take it on who wants some cool functionality that's baked into the new Keras release!). Really glad to see a community building this project back up. Can't wait to put the new version to use. |
|
@aldanor we had a release a few weeks ago. let me know if this should become an issue again. |
Dear authors/maintainers,
Would you please make a tarball release so that
On a side note, as I've tried running the full test suite on all platforms while building the conda-forge package, more than a dozen different tests were failing on master. It looks like there's a rather significant amount of work to be done e.g. to catch up with the latest numpy, plus fixing the flaky tests. Is anyone planing to take on that? I could probably find time to help out, but it would then involve major cleanup and refactoring so the codebase/tests are easier/more pleasant to work with from development standpoint which I'm not sure the authors are okay with. Ideally, there should also automated tests, i.e. tox/travis setup, which would also simplify the process of conda releases and reduce downstream surprises.
// @jaberg? @dwf?
Thanks a million!
The text was updated successfully, but these errors were encountered: