Poor code quality

[dragetd]

The python code was written when I was starting to learn python. It has very poor quality. I should rewrite these simple scripts!

[n-st]

"A friend is someone who has seen your code and still loves you." :wink:

[n-st]

Alright, I had a go at modularising the code to make it easier to re-use:

• The chunker and compression benchmarks are now merged into a single script and run sequentially. (Reduces code duplication, but we might want a commandline switch to select which tests to run…)
• Chunker tests always run with compression=none (dedup is done on uncompressed blocks, so the compression algorithm doesn't influence the work of the chunker).
• Use Python's integrated tempdir creation and cleanup. This will avoid collisions between multiple parallel instances and ensure that the script respects $TMPDIR and cleans up the temporary directory even if something crashes.
• The input data set can (and must) be given as a commandline argument (no more hardcoded paths).
• The script prints usage instructions if the argument is invalid, -h or --help.
• Error messages from Borg are passed on to the user.
• The CSV output now has a header line, so the user doesn't have to inspect the code to understand what the various numbers mean.

Let me know if I've missed anything. :)

Complete overkill for a benchmark script, of course, but I needed a sense of achievement to unwind from upgrading legacy computer systems at work today. :sweat_smile:

[dragetd]

Uh, I love it!

I almost forgot about this repository. xD Yes, your code looks a lot nicer. Especially with the borg 1.1 or 1.2 release, I'd like to re-run the benchmark. There had been various updates to the hashing and IO code.

And maybe write down some conclusions that can be backed up by numbers. Like 'Do not go too crazy with chunking as RAM will explode' or 'LZMA, is only worth it on extremely* slow network links.

Anyhow… please feel free to replace the link on the borg community repository. Once I get around to play with it, I'll send a PR with updated benchmark data. :)