On 12 March 2013 20:21, llanitedave <llanited...@veawb.coop> wrote: > On Tuesday, March 12, 2013 10:47:25 AM UTC-7, Maarten wrote: >> On Tuesday, March 12, 2013 6:11:10 PM UTC+1, Norah Jones wrote: >> >> > I want to create a random float array of size 100, with the values in the >> > array ranging from 0 to 5. I have tried random.sample(range(5),100) but >> > that does not work. How can i get what i want to achieve? >> >> Use numpy [SNIP] > > While numpy would work, I fail to see how encouraging the op to download and > install a separate library and learn a whole new set of tools would be > beneficial by default, without knowing the purpose of the need. This is like > recommending an RPG to fix a sticky door hinge.
This suggestion comes after others that show how to use the stdlib's random module. I don't think it's unreasonable to recommend numpy for this. If you want to create *arrays* of random numbers then why not use a library that provides an API specifically for that? You can test yourself to see that numpy is 10x faster for large arrays: Python 2.7 on Linux: $ python -m timeit -s 'import random' -- '[random.uniform(0, 5) for x in range(1000)]' 1000 loops, best of 3: 729 usec per loop $ python -m timeit -s 'import random' -- '[random.random() * 5 for x in range(1000)]' 1000 loops, best of 3: 296 usec per loop $ python -m timeit -s 'import numpy' -- 'numpy.random.uniform(0, 5, 1000)' 10000 loops, best of 3: 32.2 usec per loop I would use numpy for this mainly because if I'm creating arrays of random numbers I probably want to use them in ways that are easier with numpy arrays. There's also a chance the OP might benefit more generally from using numpy depending on what they're working on. Oscar -- http://mail.python.org/mailman/listinfo/python-list
