Python Questions - July 25, 2015

BartC bc at freeuk.com
Sun Jul 26 16:58:25 CEST 2015


On 26/07/2015 15:22, Steven D'Aprano wrote:
> On Sun, 26 Jul 2015 10:49 pm, BartC wrote:
>
>> How do you actually install Numpy in Windows?
>
> Are you installing from source, or a pre-built binary?
>
> To install from source, you need a C or Fortran compiler, and a bunch of
> extra libraries (I think BLAS is one of them?).
>
>
>> I had a go a month or two ago and couldn't get anywhere.
>>
>> I realise that this is something that apparently no-one else on the
>> planet has no problem with except me, but nevertheless it would be
>> interesting to know exactly how it's done.
>
> No, it's not just you.
>
>> I've just at numpy.org, they direct me to scipy.org, which talks about
>> sources and binaries at sourceforge.
>
> Well, did you read what they said?

On which site, numpy, scipy or sourceforge? I tried this:

numpy.org; nothing on that page.

Click 'Getting Numpy' => Obtaining Numpy & SciFi libraries. The only 
references to Windows takes me to the sourceforge site (ignoring the 
link to do with building).

The sourceforge site (http://sourceforge.net/projects/numpy/files/) has 
nothing like that either. It just has 'Looking for the latest version? 
Download numpy-1.9.2.zip'. That's the obvious link.

Or I can ignore that and click on the 'NumPy' link.  That takes me to a 
long list of versions (1.9.2 to 1.0.4).

Nothing that looks like your paragraph 2, or the proper installers.

It's only after examining your link that I realise that each version 
number such as 1.9.2 actually opens up a new directory of files. And 
*now* I can see that some files are proper installers (which I've 
installed and it was a lot smaller and quicker than the Anacondo one).

>  Paragraph two:
>
> "For most users, especially on Windows and Mac, the easiest way to install
> the packages of the SciPy stack is to download one of these Python
> distributions, which includes all the key packages: ..."

But I've still found of the above quote.

You see the problem however? Loads of confusing links that all look very 
similar, and sometimes just go around in circles?

Why can't they just have that direct link on the numpy home page? Just 
for Windows users as presumably everyone else has any problem. And no 
one wants to mess around building things from scratch.

-- 
Bartc


More information about the Python-list mailing list