Python 3 Support #902
Comments
👍 Pandas often has issues with Python 2 version like pandas-dev/pandas#12142, and the management of string in Python 2 it is really a pain. |
@VelizarVESSELINOV Agreed. I'd really LOVE to see a Py3 version 👍 |
Yes, please. |
Is there any plans to support python3? |
I'm not sure we'll be able get to it soon unfortunately. PRs are more than welcome though, so please feel free to continue the discussion or send a change if you feel like it. :) |
I think we need to actually think through a few things about how to provide this - one container per version, one version per kernel etc. and implications of those approaches. |
Ideally, the version should be tied to the notebook, so the user can choose what kernel to run when creating a new notebook. We should also find a way of making this visible inside the notebook somehow. I believe this is what Jupyter does too. |
This is all solved by jupyter docker stacks.
https://github.com/jupyter/docker-stacks. Just let us use those on Google
cloud, and everyone will be happy!
On Wed, Nov 23, 2016, 7:40 PM Yasser Elsayed ***@***.***> wrote:
Ideally, the version should be tied to the notebook, so the user can
choose what kernel to run when creating a new notebook. We should also find
a way of making this visible inside the notebook somehow. I believe this is
what Jupyter does too.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#902 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEaCfU2-XQBenHepdtZGuR5Alsn8XeLRks5rBNzogaJpZM4JHsR9>
.
--
Sent from phone; typos by Google autocorrect.
|
This likely requires the equivalent pydatalab support: googledatalab/pydatalab#21 |
* Adding a Python 3 Kernel to Datalab. Installs Python 3 and py3 versions of our supported packages, including the python3 jupyter kernel, and adds a dropdown to the notebook page to allow users to switch kernels. Fixes #902 * Fix merge error. * Changes based on review comments. * Fixing minimum toolbar width. * Fixing build error.
Love it! Thank you so much @chmeyers! |
Love it! Thanks so much guys :) 👍 |
Did some investigation with the first official build that has this. Docker image (compressed on GCR) is 790MB vs 566 MB for the previous version. Expectation is that initial startup time for VMs is increased by ~15s due to the extra download time. RAM usage didn't seem to change significantly, this and the previous version appeared to use within tens of MB of each other. Still have to go through all the Sample Notebooks and libraries and check what doesn't work in Python 3. That will be tracked by googledatalab/pydatalab#21 |
Hi, How can I start using the py3 version?
Thanks in advance! |
To use py3 currently, you'll need to grab a pre-release version of Datalab. You can do this by specifying an image-name during creation: |
I'm trying |
When I'm using kernel Python3, my expectation was that pip and python commands are by default using the kernel version of Python 3 and not the Python 2. Do you agree with my expectations? For migration of notebooks from 2to3 with commands inside it will be logical that basic command stays the same, and during the kernel switch, the aliases are directed to the current version. |
Both the py2 and py3 kernels run on the same underlying file system, so 'pip' and 'python' always point to either py2 or py3 without regard for which kernel is actually running at the moment. I'm trying to think of a good way to avoid that (without using entirely separate file systems for the two kernels) but can't think of one. |
New projects tend to not use a lot of Python 2 -- we've reached the
inflection point. But if you want to keep python 2 you could do what most
have done and go with pip3 for python 3. Or, if you feel like leaning into
python 3, you could have pip refer to python 3 and pip2 refer to python 2.
Or you could just drop Python 2 and have a completely different image. Are
there a lot of python 2 only data science packages out there now?
…On Mon, Aug 28, 2017 at 12:57 PM, chmeyers ***@***.***> wrote:
Both the py2 and py3 kernels run on the same underlying file system, so
'pip' and 'python' always point to either py2 or py3 without regard for
which kernel is actually running at the moment.
I'm trying to think of a good way to avoid that (without using entirely
separate file systems for the two kernels) but can't think of one.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#902 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEaCffEehhon9ey0x6vsPrnHuoc6yiwMks5scvGDgaJpZM4JHsR9>
.
--
Don
202-503-4132
donald.braman@gmail.com
|
@donaldbraman I agree there is no reason to start a new Python notebook using the old version 2. I hope default will be Py3. All the packages that I'm using are Py3 compatible: http://py3readiness.org |
@chmeyers If Py3 is the default will be good to have |
Generally agree that we should optimize for either Py2 or Py3 rather than both at the same time. |
@nikhilk Py3 is the future, suggestion to start with the future rather than the past Py2 |
Hi, is it possible currently to upgrade the version of Python used by Datalab? My default install uses 2.7.9, but would love to use 3.xx. Thoughts?
Thank you.
The text was updated successfully, but these errors were encountered: