[melbourne-pug] Install ‘conda’ using Debian package manager

[Scott Wales]

Anaconda the package manager is a cross platform version of apt-get or yum. It can come with either a large set of packages pre-installed (the anaconda distribution) or a small set (miniconda). `conda` is the tool that the Anaconda package manager uses to install and manage packages, it is not really a stand alone thing to be installed with pip.


The benefits of Anaconda is that it includes a full dependency solver, so the environment is always consistent when you update packages, it can handle binary dependencies that pip on its own has trouble with (e.g. for numpy), and it makes it simple to share a consistent environment between different computers.


Anaconda has its own flavour of venvs, works pretty similar to Python's but binary dependencies are also kept separate.

The packages in your yaml file are most likely available on PyPI, although please do be mindful that the instructors will most likely be unable to spend time helping you with installation issues during the tutorial itself.

As a side note I don't believe that PyPI does any vetting of what gets uploaded, I'm not sure there's any real difference security wise between downloading and running a python script with pip and downloading and running a shell script.

Cheers, Scott

________________________________
From: melbourne-pug <melbourne-pug-bounces+scott.wales=unimelb.edu.au at python.org> on behalf of Ben Finney <ben+python at benfinney.id.au>
Sent: Wednesday, 2 August 2017 2:55 PM
To: melbourne-pug at python.org
Subject: Re: [melbourne-pug] Install ‘conda’ using Debian package manager

Sergio Valqui <svalqui at unimelb.edu.au> writes:

> Anaconda does the version control for all those packages, for a given
> version of Anaconda there is a given version of the packages

Most Python packages do this by declaring the packages (and versions)
they depend on, in the Distutils metadata.

Why doesn't Anaconda declare that, so I can install Anaconda by telling
Pip to bring in all its dependencies in a unified way with other Python
packages?

> it also manage the environment; so is not as easy as simple installing
> the packages.

Thank you. Is this more than a Python ‘venv’ environment? What would I
need to do to have a ‘venv’ environment set up so the Anaconda
assumptions will work?

> This is difficult to achieve as anaconda manages the package versions,
> and environment; also the packages are quite diverse too manage them
> individually

I have a list of dependencies (a YAML file) for the ‘conda’ tool. Are
they just PyPI packages that I can also install with Pip?

--
 \          “Generally speaking, the errors in religion are dangerous; |
  `\    those in philosophy only ridiculous.” —David Hume, _A Treatise |
_o__)                                           of Human Nature_, 1739 |
Ben Finney

_______________________________________________
melbourne-pug mailing list
melbourne-pug at python.org
https://mail.python.org/mailman/listinfo/melbourne-pug

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/melbourne-pug/attachments/20170802/46f6be35/attachment-0001.html>