Rosetta stone should use Python 3

[asmeurer]

The Rosetta stone should use Python 3. None of the current examples will run because they use Python 2-only syntax (print statements). There's also at least one example I saw that assumes / does integer division.

Also, the NumPy example for matrix multiplication can use @ instead of dot.

[asmeurer]

By the way, this is unrelated, but many of the NumPy examples are unNumPythonic, and tend to show NumPy in a worse light than they should. For example:

from numpy import array, pi
a = array([i for i in range(1, 7)])
b = array([(2*i*pi+1)/2 for i in range(1, 7)])
c = array([i for i in range(1, 7) \
    for j in range(1, 4)])

While these examples do work, the NumPy way is to use array operations. This is both faster and easier to read.

Also, NumPy defaults to 0-based indexing so I think the NumPy examples that are based on range should all use 0–6 instead of 1–7. Or if you want to remain neutral pick a range that doesn't start at 1 or 0, so you can see the difference in both.

from numpy import arange, broadcast_to, pi
a = arange(7) # Note, this is [0, 1, 2, 3, 4, 5, 6]
b = 2*arange(7)*pi+1)/2
c = broadcast_to(arange(1, 7), (3, 6)).T.reshape(6*3) # There's probably a simpler way to do this

[certik]

Good point about numpy, I should improve it.

Yes, I need to update them to Python 3, I wrote this about 10 years ago.

I am in the process of slowly migrating the information on this website to https://fortran-lang.org/.

[asmeurer]

Thanks. I was looking at your site because of some of the stuff you wrote on indexing here https://www.fortran90.org/src/faq.html#what-is-the-most-natural-way-to-handle-initial-and-final-points-of-intervals.

By the way, I was wondering, how does Fortran handle negative or "reverse" indexing? In Python, you can use negative numbers to index from the end of a collection, like a[-3] will give the 3rd to last element. Does Fortran have any special syntax to do this, or do you always have to do a(size(a) - n + 1)? And how does that work if a has a starting index other than 1?

[asmeurer]

There was an in-progress PR to do this here #31 (it doesn't look like it fixed the integer division issue).

There was some discussion there about __future__ imports, but I would just drop Python 2 completely. Python 2 is already dead at this point, and It isn't important for a Rosetta stone anyway.

[certik]

The 3rd to last element is a[-3] in Python and a(size(a)-2) in Fortran.

If the indexing does not start with 1, you can use ubound, something like a(ubound(a)-2).

I agree, Python 3 is enough.

[asmeurer]

I see. So there isn't any special syntax. I was wondering if any languages like Fortran use a special syntax to make reverse indexing easier, other than using negative integers (which I think is a bad idea). It looks like Julia has an end keyword as a shorthand for the size of the indexed axis, which has the same effect.