ENH: LineStats aggregation

[TTsangSC]

(Based on @Marxlp's #219 and the changes I proposed there.)

Code changes

• line_profiler/line_profiler.py[i]
  • LineStats:
    New API subclass of line_profiler/_line_profiler.pyx::LineStats which provides the following convenience methods:
    • __add__(), __iadd__(), from_stats_objects():
      Methods for aggregating the timing data between instances
    • to_file(), from_files(), print():
      IO methods
    • __eq__():
      Method for comparing instances
  • LineProfiler.get_stats(), .dump_stats(), .print_stats():
    Refactored to return/use the new LineStats class
  • main():
    Updated to allow for passing more than one .lprof files

Tests changes

• tests/test_cli.py::test_multiple_lprof_files():
  New test checking that python -m line_profiler handles multiple .lprof files correctly
• tests/test_line_profiler.py::test_load_stats_files():
  New test checking that LineStats.from_files() reads one or more .lprof files correctly, and handles .lprof (pickle) files generated by both new and old versions of line_profiler

[TTsangSC]

@Marxlp, please indicate if you would like to have co-authorship on this PR; and if so, also the appropriate email address so we can update the commit messages, and that GitHub can correctly attribute the contribution also to you. Cheers!

[codecov]

Codecov Report

❌ Patch coverage is 96.77419% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.13%. Comparing base (60e928f) to head (eef1d3b).
⚠️ Report is 12 commits behind head on main.

Files with missing lines	Patch %	Lines
line_profiler/line_profiler.py	96.66%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #380      +/-   ##
==========================================
+ Coverage   87.56%   88.13%   +0.56%     
==========================================
  Files          18       20       +2     
  Lines        1641     1921     +280     
  Branches      348      412      +64     
==========================================
+ Hits         1437     1693     +256     
- Misses        149      168      +19     
- Partials       55       60       +5     

Files with missing lines	Coverage Δ
line_profiler/__init__.py	100.00% <100.00%> (ø)
line_profiler/line_profiler.py	98.90% <96.66%> (-0.77%)	⬇️

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2295c1c...eef1d3b. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Theelx]

I think this may belong in a new area for 5.1.0 (not 5.0.1 because enhancements are minor version bumps according to SemVer)!

[Erotemic]

Yes, when that happens, bump the version in the 3 required places (in 6.x I want to consolidate this into a single place: the __init__.py file).

[TTsangSC]

For this PR I didn't feel comfortable presuming to bump the minor version – since I was under the assumption that there will be a 5.0.1 before that. But of course it makes better sense as far as semver goes, and if you already intended to do a minor bump I'm more than happy to oblige. The question now I guess will be whether we bump here and now, or stash this away in a dev/5.1.0 branch somewhere (which we don't currently have) until that happens.

Apropos, in principle the unification of the versions can already happen now without a major-version bump, since the days where kernprof could function independently of the line_profiler has been long gone. We can probably just define __version__ once somewhere and import it elsewhere for now, and just remove it where not explicitly needed in 6.0...

[Erotemic]

This looks good. Glad you took numeric stability into account.

[TTsangSC]

Cheers! Thanks for the review.