Fixes #195 "large negative increment values"

[fabianp]

This should fix #195. Increment should reflect the largest increment with respect to the previous line, which is not what the previous code was doing. If run in a for loop, for some reason it was accumulating the increments, which was explaining why some people were seeing large negative increments (https://stackoverflow.com/questions/51077423/strange-negative-value-in-memory-profiler-while-reading-file-python)

I also disabled a test that was assuming that the increment of a function that deletes its temporaries need to be zero, which given the above was a wrong test.

@Juanlu001 : I would appreciate your OK if you have time

[astrojuanlu]

Hi @fabianp! I was trying to reproduce the behavior with this program:

$ cat test_file.py 
@profile
def get_file():
    filename = "/tmp/test.txt"
    f_out = "/tmp/out.txt"
    count = 0
    with open(filename, 'r') as f_in, open(f_out, 'w') as f_out:
        for line in f_in:
            f_out.write(line)

    print(str(count))


if __name__ == '__main__':
    get_file()

(that count variable does not make any sense as it's written, but I just transcribed what I saw in the image)

This is what I get with 0.52:

0
Filename: test_file.py

Line #    Mem usage    Increment   Line Contents
================================================
     1   31.941 MiB   31.941 MiB   @profile
     2                             def get_file():
     3   31.941 MiB    0.000 MiB       filename = "/tmp/test.txt"
     4   31.941 MiB    0.000 MiB       f_out = "/tmp/out.txt"
     5   31.941 MiB    0.000 MiB       count = 0
     6   31.941 MiB    0.000 MiB       with open(filename, 'r') as f_in, open(f_out, 'w') as f_out:
     7   31.941 MiB    0.000 MiB           for line in f_in:
     8   31.941 MiB    0.000 MiB               f_out.write(line)
     9                             
    10   31.941 MiB    0.000 MiB       print(str(count))

and this with your patch:

0
Filename: test_file.py

Line #    Mem usage    Increment   Line Contents
================================================
     1   31.914 MiB   31.914 MiB   @profile
     2                             def get_file():
     3   31.914 MiB    0.000 MiB       filename = "/tmp/test.txt"
     4   31.914 MiB    0.000 MiB       f_out = "/tmp/out.txt"
     5   31.914 MiB    0.000 MiB       count = 0
     6   31.914 MiB    0.000 MiB       with open(filename, 'r') as f_in, open(f_out, 'w') as f_out:
     7   32.016 MiB    0.000 MiB           for line in f_in:
     8   32.016 MiB    0.102 MiB               f_out.write(line)
     9                             
    10   32.016 MiB    0.000 MiB       print(str(count))

No one provided a script to reproduce in #195. Codewise this patch looks good to me, but I still would like to understand the problem more.

[fabianp]

@Juanlu001 would you be OK merging and closing #195 to see if it fixes the issue?

[astrojuanlu]

Ok, let's merge them!