Open mikemccand opened 2 years ago
OK, I pushed a rough first cut at this! It produces this file each night. Maybe we can build on this to make alarms/anomaly detection/pre-release checks of some kind.
There are still some TODOs:
# TODO: also include all known annotations!
# TODO: when errorBars is true, the variance(s) is/are extra columns in the data, so the headers
# look incorrect now
# TODO: get all other benches into this -- geo, sparse, github PRs, etc.
Ooh, awesome - so that file is labeled "for_all_time.json" - should we publish a dated file? Oh never mind I see that it has all the previous days in it. SO It looks like "schema" is:
{
MetricName -> [
Title,
[
[ Date, header1, ... headerN],
[ "YYYY-MM-DD HH:MM:DD", data1, ..., dataN]
]
], ...
}
I noticed a few squdgy things though:
Suggestions:
- could we add some header column explaining the additional value(s)? Are they variances?
Yeah +1 I'll try to add that (they are variances, I think!).
2. Can we use JSON numbers?
Ahh good catch -- I'll try to fix that too.
One of the awesome suggestions that came out of the ApacheCon NA 2022 talk ("Learning from 11+ years of Apache Lucene benchmarks") was to export at least the values for all nightly charts to a simple format (JSON), each night.
We could then build on that to more automatically spot anomalies, regressions, set up alarms, etc.