refactoring saved df files

[feizheng10]

Save analysis dataframes in that loading function with verbose actually is not a bad idea! However, there are some reasons to refactoring the code a bit:

• Writing to any dir without asking user is not a good practice(My current implementation to write TopStats there is a tmp solution, it is not good and need to be fixed in proper way ).
• Implement it in tty will guarantee to save whatever user saw in cli. It is hard to imagine users want to save dfs in gui when loading(may be a bit confused when changing filter).
• It is easy to retrieve table title info in tty and put it into file name.

[coleramos425]

Tip for formatting test:

python3 -m pip install black
cd omniperf/
black --diff --check . # check all files for any formatting discrepancies
black . # re-format files which don't comply

[feizheng10]

black --diff --check .

Thanks! As long as we black with the same format style:)

[feizheng10]

black --diff --check .

Thanks! As long as we black with the same format style:)

Just check some discussion back a while ago... I would suggest to use yapf instead of black. The reason is: a few ROCm components have being evaluated python formatting solutions and we are all on the same page that yapf default formatting is much better than black, even though it is not perfect.

More details you might follow:

yapf installation: sudo apt install yapf

yapf version >= 0.32.0

For your local file formatting: yapf --style pep8 -ip your_file

CI verification: find . -iname '*.py' | grep -v 'build/' | xargs -n 1 -P 1 -I{} -t sh -c 'yapf -ip --style pep8 {}

Disable/enable syntax:

# yapf: disable
...

# yapf: enable 

[coleramos425]

I would suggest to use yapf instead of black.

Thanks for the suggestion. Omniperf uses black format to remain consistent with the rest of the Omni* toolset (i.e. Omnitrace). I'll need to evaluate performance differences