Add flat mode to console renderer

[matthiasdiener]

This was inspired by a similar flat mode in vmprof. Sometimes it is interesting to focus on functions that get called from multiple locations in the call graph in order to determine which functions consume the most time overall.

Run with pyinstrument ... -p flat.

Here is how it looks like:

$ pyinstrument --load-prev 2023-05-03T14-25-05  # standard mode

  _     ._   __/__   _ _  _  _ _/_   Recorded: 14:25:05  Samples:  165
 /_//_/// /_\ / //_// / //_'/ //     Duration: 1.285     CPU time: 0.068
/   _/                      v4.4.0

Program: examples/c_sort.py

1.284 <module>  ../examples/c_sort.py:1
├─ 1.193 slow_key  ../examples/c_sort.py:27
│  └─ 1.193 sleep  None
│        [2 frames hidden]  ..
└─ 0.090 <module>  numpy/__init__.py:1
      [158 frames hidden]  numpy, pickle, .., _compat_pickle, __...

To view this report with different options, run:
    pyinstrument --load-prev 2023-05-03T14-25-05 [options]

$ pyinstrument --load-prev 2023-05-03T14-25-05 -p flat  # flat mode

  _     ._   __/__   _ _  _  _ _/_   Recorded: 14:25:05  Samples:  165
 /_//_/// /_\ / //_// / //_'/ //     Duration: 1.285     CPU time: 0.068
/   _/                      v4.4.0

Program: examples/c_sort.py

1.193s sleep  ../<built-in>:0
0.023s <module>  numpy/random/_pickle.py:1
0.018s <module>  numpy/core/__init__.py:1
0.018s <module>  numpy/core/overrides.py:1
0.018s <module>  numpy/__init__.py:1
0.013s <module>  numpy/lib/__init__.py:1
0.000s <module>  ../examples/c_sort.py:1
0.000s slow_key  ../examples/c_sort.py:27
0.000s <module>  numpy/core/multiarray.py:1
0.000s <module>  numpy/random/__init__.py:1

To view this report with different options, run:
    pyinstrument --load-prev 2023-05-03T14-25-05 [options]

[joerick]

Nice, thanks! I'll take a look soon. Just a quick note, 'exclusive time' is known as 'self time' in this project. I think it's already available as total_self_time on Frame. But, there may have been a bug in that property until recently, can't remember if it's merged or not!

[matthiasdiener]

Nice, thanks! I'll take a look soon. Just a quick note, 'exclusive time' is known as 'self time' in this project. I think it's already available as total_self_time on Frame. But, there may have been a bug in that property until recently, can't remember if it's merged or not!

Thanks! I tried this with total_self_time, but it always returned zero. Looking at the other PRs, I think this might be a fix: https://github.com/joerick/pyinstrument/pull/236/files#diff-34c9ac7ed6042dfd8c5340562bb37088e871be076ce870b1325da6e3f8f4d5cb

[matthiasdiener]

Nice, thanks! I'll take a look soon. Just a quick note, 'exclusive time' is known as 'self time' in this project. I think it's already available as total_self_time on Frame. But, there may have been a bug in that property until recently, can't remember if it's merged or not!

Thanks! I tried this with total_self_time, but it always returned zero. Looking at the other PRs, I think this might be a fix: https://github.com/joerick/pyinstrument/pull/236/files#diff-34c9ac7ed6042dfd8c5340562bb37088e871be076ce870b1325da6e3f8f4d5cb

I added that fix to this PR and modified the code to use total_self_time. Could you please approve the CI workflow?

[joerick]

Looks good @matthiasdiener, thank you! A few comments inline. The last commit you added isn't correct though, it's not possible to say anything about 'ncalls' in pyinstrument because it is a statistical profiler - they could be many calls that pyinstrument didn't record because it's only recording every {interval}s.

[matthiasdiener]

The last commit you added isn't correct though, it's not possible to say anything about 'ncalls' in pyinstrument because it is a statistical profiler - they could be many calls that pyinstrument didn't record because it's only recording every {interval}s.

You're right, of course. I reverted that commit in bcca59c.

[joerick]

leave this with me @matthiasdiener, I have some more thoughts on code structure that are probably easier to do than describe!

[joerick]

Would you mind checking this still works as you intended @matthiasdiener ? I think I'm happy with it.

[matthiasdiener]

Thanks @joerick! This still works as intended.

[matthiasdiener]

Thanks @joerick! This is still working as intended.

[joerick]

thank you @matthiasdiener for this! gonna get a release out really soon