Added correct indexing for negative steps and check for 0-index

[wolfv]

This fixes stuff like
a = {1,2,3,4,5}
a[[3;;1;;-1]] == {3,2,1}

which previously resulted in {2}.

This should fix #196

[sn6uv]

Thanks for fixing this. I'm finding the python_seq function a little bit confusing. Do you think you could clean it up and add a short docstring explaining its purpose?

[wolfv]

Do you think there is a way to fix all of these at once?

Ie. Table[i,{i,1,-10,-1}] -> {1,0,-1...-10} but evaluates to {} currently
Also Range[i+1, i+10] -> {i+1, i+2, i+3...} but doesn't evaluate at all.
Range[0,-10,-1] doesnt work either.

[wolfv]

def python_seq(start, stop, step=+1):
    # step can be None, which means positive stepsize
    if not step: step = +1

    # What this function does is convert a 1 based index to
    # zero based index. Additionally, the start of a list
    # is indicated by None in python.
    if step > 0:
        if start > 0:
            start -= 1
        if stop is not None and stop < 0:
            stop += 1
            if stop == 0:
                stop = None
        return start, stop
    else:
        # in the case of a negative step, 
        # admissible values are start < stop
        # so that start get's decremented until stop is 
        # reached. 
        # However, note that in python a = [1, 2, 3, 4][3:0:-1] == [4, 3, 2]
        # and not, as one might think [4, 3, 2, 1] (which is the result of [3:None:-1])
        # Therefore, stop has -2 subtracted, and None is returned when stop == -1
        if stop and stop > 0:
            stop -= 2

        elif stop and stop < 0:
            stop -= 1

        if start and start > 0:
            # zero base index
            start -= 1

        if not start:
            # no start index = include all (0)
            start = None

        if stop == -1:
            # stop equals end, which equals -1
            stop = None
        elif stop == None:
            # No stop equals len(list)
            # This is actually weird in Mathematica. 
            # E.g. a = {1,2,3,4}; a[[4;; ;; -1]] == {4}
            # but all other combinations evaluate to an error
            # a[[3 ;;  ;; -1]] = cannot take positions -4 through -1
            # in Mathics it all evaluates to an empty list currently.
            stop = -1
        return start, stop

[sn6uv]

What's the status of this PR? That code in the comments looks much better but there are still some unhandled cases I think e.g. python_seq(-2, -3, 1) returns (-2, -2).

[wolfv]

Your example is the actually correct:

But you're right that this piece of code should get some proper testing and specification.

[sn6uv]

I'm planning to release Mathics 0.9 in the next couple of weeks. Would love to get these changes merged in if possible. @wolfv are you planning to finish this?

[sn6uv]

closed in favour of #325