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!