make numpy speed improvements more prominent

[philippkeller]

Following up on ladyadas comment I tried to make numpy's speed improvement more prominent.

That's how I approach that. Of course this could be solved in other ways, some alternatives I considered:

• output a warning on stderr when numpy can't be imported (at top of rgb.py), but this is annoying for anyone who wants to run without numpy
• explicitly require numpy, remove code in image() that handles when numpy is not there. I didn't dare to do this but might be the most straightforward way. But there might be situations for users where they don't want numpy as it's probably quite heavy to install.
• not adding numpy to requirements.txt, but I think it's best to lead users to the usually best way to roll

[ladyada]

hi - please dont do it this way - have the gif loader example import numpy, that will force it to be installed :)

[philippkeller]

hi - please dont do it this way - have the gif loader example import numpy, that will force it to be installed :)

I'm confused. So you mean add import numpy in the first line of the image() method? And then also remove the code which handles the case when numpy is not available?

[ladyada]

no to the gif loader example, the python script

[philippkeller]

oh, I was totally on another track :-) the thread was referring the gif example. But I didn't look at the example yet (rotating gif is not my use case).

Now I understand what you mean. I'll close this PR and I will later open another one with:

• the unused import in the gif example
• a new example which resembles my use case better

Also I think the examples could be put more prominently in readme, maybe I'll add a suggestion in the same PR

[ladyada]

awesome, sounds good :)