Here we change the format of the scenario results so that it now stores
the atomic actions data as a dict:
Before:
{
"duration": ...,
"error": ...,
"atomic_actions": [
{"action": "a1", "duration": 0.3},
{"action": "a2", "duration": 0.1},
{"action": "a1", "duration": 0.25}
],
...
}
After:
{
"duration": ...,
"error": ...,
"atomic_actions": {
"a1": 0.3,
"a1 (2)": 0.25,
"a2": 0.1
},
...
}
We also solve 2 connected problems here:
1. The statistics table that shows up in the CLI after benchmarks complete now
always has rows for all the atomic actions that were invoked in benchmark
scenarios. Before this patch, this was not the case since occasionally the
atomic actions data could not be written due to Exceptions.
2. We've removed lots of ugly code in benchmark/processing/plot.py, since now
there is no need to make passes through the atomic actions data to collect
their names.
This patch also:
* contains some removing of code duplicate between plot & CLI output generation;
* adds missing UTs for raw benchmark data processing;
* fixes a minor bug in histograms bins number calculation.
Change-Id: I17f563051bd1c2ec2fb47a385d4cc652895e1f9e
b55d6dfc63