Conda installation hiccup

[Santosh Shanbhogue]

Hi Bryan,

FYI - 

On Win10, I just downloaded the latest version of Anaconda (64 bit, 3.5 -- the only one available) and then navigated to the root environment via the DOS prompt

I can successfully install cantera via : 

conda install -c cantera/label/dev cantera

but not:

conda install -c cantera cantera

for which I get the package not found error

----

In the documentation, there are instructions to install numpy 1.8, but the site which is linked does not carry:

numpy‑1.8.2+mkl‑cp34‑none‑win_amd64.whl

anymore. Is it a problem if a user just executes 

c:\python34\Scripts\pip.exe install numpy

which will install numpy 1.1?

Regards

Santosh

[Bryan W. Weber]

Santosh,

1) We haven't uploaded packages to the main channel yet, since we're still in beta

2) I think it would be worthwhile to recommend using Pip to install NumPy on all platforms now that the wheel format is relatively common: https://pypi.org/project/numpy/#files. I imagine you meant NumPy 1.11 (1.1 would be older than 1.8). However, users should not specify the full path to Pip, they should have Pip on their PATH variable. Care to make a PR?

Also, Pip installing NumPy should not be necessary for the conda packages since NumPy is listed as a dependency in the recipe.

Bryan

[Santosh Shanbhogue]

Hi Brian,

I should have mentioned that I was trying two separate ways of installing cantera. One method uses Anaconda. That problem is solved now since as you mentioned there are no packages in the main channel (and of course installing numpy is unnecessary)

The second method is using the instructions on http://cantera.github.io/docs/sphinx/html/install.html (for the stable version). Here, I install python first, then numpy and then download the msi. Yes, sorry, I meant version 1.11. My question was that why can't a user just execute 

pip install numpy

which will match the numpy version to the correct python version and OS ? Just curious.

Regards

Santosh 

[Ray Speth]

Santosh,

The filename numpy‑1.8.2+mkl‑cp34‑none‑win_amd64.whl is provided only as an example (hence the 'e.g.') and the documentation specifically suggests installing the latest 1.x release of Numpy available. In addition, in the dev-docs, the name mentioned is for numpy-1.11.

I would recommend against `pip install numpy` on Windows because it installs a copy that is built without an optimized BLAS/LAPACK implementation. In addition, the Windows binaries for Cantera are compiled using the MKL-linked versions, so compatibility there should be essentially guaranteed. Using the Numpy available through pip is also a bit of a dead-end on Windows, since there is no Scipy package available that way (not that Cantera requires it). I don't know whether there are any other potential issues using a different build of Numpy. For non-Windows installations, pip is just fine.

I assume that the removal of the Python 3.4 distributions of Anaconda is in anticipation of the release of Python 3.6.

Regards,

Ray

On Wednesday, December 21, 2016 at 10:01:01 AM UTC-5, Santosh Shanbhogue wrote:

[Santosh Shanbhogue]

Great, thanks

--
You received this message because you are subscribed to the Google Groups "Cantera Users' Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cantera-user...@googlegroups.com.
To post to this group, send email to canter...@googlegroups.com.
Visit this group at https://groups.google.com/group/cantera-users.
For more options, visit https://groups.google.com/d/optout.