Docs: improve installation instructions, recommend Anaconda.

[lexual]

I believe installing pandas via Anaconda should definitely be the recommended installation method for most users.

It is cross platform, doesn't require a compiler, and will install all requirements for the user also (including NumPy).

[jreback]

nice

can u post a rendered version?

[lexual]

https://gist.github.com/lexual/525d30536505b7bfec6d

Can download and load this html??

I had the links wrong, fixed now.

[lexual]

Alternatively, this will just do minimal docs build:

python make.py --no-api --single install

[jreback]

easiest to render then save as a png

[lexual]

How do I do that, can see how in Chrome or Firefox?

[lexual]

[jreback]

perfect! as an aside, can you update the wiki on how to do that?

[jreback]

I would put miniconda instructions before Anaconda as in reality that's the most recommended route, no?

[lexual]

I'd tend to leave it as it is.

If you install conda via miniconda then install pandas via "conda install pandas", this will only install numpy & pandas, and not matplotlib, pytables, etc.

Basically I think installing Anaconda is the way to go for inexperienced users, for more power-users miniconda is the way to go. (BTW either way basically ends up doing the exact same thing. i.e. installing anaconda install conda, and running "conda install anaconda" after a miniconda install gets you to the same place).

If you install anaconda, most of the more optional dependencies (pytables, matplotlib, etc) will be there, and users will be less likely to get confused by those libraries being missing.

Also, where is the wiki. I installed some chrome extension to create that png.

[jreback]

go to the issues page
it's the 3rd or 4th icon down on the right

[jreback]

https://github.com/pydata/pandas/wiki

[lexual]

https://github.com/pydata/pandas/wiki/Posting-rendered-documentation-page-change-in-github-issues.

[lexual]

Do I need to do anything further to get this merged??

[jreback]

@jorisvandenbossche pls have a look

[jorisvandenbossche]

docs itself look OK

I was just wondering if the focus is a bit too much on Anaconda? I myself also use it, and I would also recommend this personally to others, but as 'pandas' we should maybe be a bit more neutral. As there are also other 'distributions' that you can use to get the full scipy/pydata stack (enthought canopy, pythonxy, ..).
Maybe we should do more something like on http://www.scipy.org/install.html: list some different scientific python distributions, and then we can still elaborate more on how to use it with Anaconda?

Another comment: I would try to structure the title a bit more (I mean in different levels of headings). Eg having the bigger headers 'python version support'/'installing pandas'/'dependencies' and then the current headers as subheadings of those?

[lexual]

Have made the heading structure changes requested.

I disagree about the need to be neutral.

If both me and you as power users personally use Anaconda and would personally recommend it to others, shouldn't we do the same for our users?

I'm giving a PyCon AU Pandas introductory talk later this week, and I fully intend to recommend installation through Anaconda. It's also part of the reason I'm keen to get this merged, so I can link to the development docs, and not my own notes.

I also think its important to guide new users down the simplest path, rather than confusing them by giving them 3-4 options.

Also: hadn't heard of pythonxy before today, but it appears to be Windows only. Anaconda is cross-platform.

Regardless, I think this patch is a big improvement on the previous installation docs, and there is nothing stopping users of those others options providing patches with other installation options listed.

[jorisvandenbossche]

@lexual For me it is OK to merge this as is, so you can refer to it in your talk. BTW, good luck with your talk!

On the 'neutrality' issue, I agree we should provide clear and easy instruction (and therefore, we (should) also recommenend to use distributions instead of installing it manually (as now with your PR)).
But you have also the other side: eg Enthought also has done a lot for the numpy/scipy community, and they maybe would appreciate it we list them (in fact, they are mentioned now, at the end of the dependencies section). @jreback @cpcloud What do you think?

On pythonxy, before Anaconda was there (and this is actually only recently), pythonxy was for me and my colleagues thé way to painless install the scipy stack on Windows. But I also don't use it anymore at the moment.

[jreback]

@lexual ok merged. Let's review again after your talk (for flow/content and such), see #7865

[lexual]

Is there any reason this link would have suddenly stopped working (404)?

http://pandas-docs.github.io/pandas-docs-travis/indexing.html

Also, the dev docs appear to be out of date too:

http://pandas.pydata.org/pandas-docs/dev/install.html

[jreback]

I restarted the builds. If travis fails for some reason the docs don't build (not sure why in this case)..will let u know

[jreback]

good to go now

[lexual]

Yep, travis links look to be working again.