Add collapsed markdown table

[letmaik]

It's sometimes useful to look at the raw data.

[cimetrics]

letmaik/md-table@685 aka 20210402.14 vs main ewma over 11 builds from 637 to 669

Click to see table

build_id	build_number	Throughput (tx/s) ^	Latency (ms)	Peak working set size (GB)	Accuracy (%) ^	Error rate (%)	Memory fragmentation (%)	CPU usage (%)	New metric (U)
637	20210205.1	48261	10	46	0.68	0.05	0.02	0.28	189
638	20210205.2	50232	11	51	0.69	0.04	0.019	0.32	101
646	20210205.10	47148	10	55	0.72	0.05	0.02	0.31	129
647	20210205.11	49154	10	49	0.68	0.05	0.022	0.32	197
652	20210205.16	47145	9	52	0.69	0.06	0.02	0.3	100
653	20210205.17	50877	10	55	0.7	0.04	0.022	0.28	100
655	20210326.1	46562	10	51	0.69	0.06	0.021	0.31	155
658	20210326.4	50718	9	53	0.68	0.05	0.021	0.29	128
660	20210326.6	46897	9	48	0.73	0.05	0.022	0.28	133
664	20210401.3	49393	11	53	0.73	0.05	0.019	0.29	188
669	20210401.8	51450	9	54	0.72	0.06	0.022	0.29	191

[achamayou]

That’s awesome, thank you!

[letmaik]

Hm... the Actions job fails due to DNS problems connecting to Cosmos DB and the DevOps job doesn't even trigger now. Oh well...

[letmaik]

OK, so the build number is not really working yet, a bit hard to test locally, but I think it would be nice to have it since that's what's shown in the plots and what you see in CI.

[achamayou]

Agreed!

[letmaik]

pandas uses tabulate for markdown formatting and by default tabulate tries to interpret strings as numbers. This can only be disabled for all columns, at which point it also disables nice formatting for floats (reasonable number of decimals).
I opened an issue (astanin/python-tabulate#121) about that since it should be possible to disable parsing for specific columns.
Let's wait a bit and see if the maintainer is open to the idea. If not, then we may have to manually truncate floats in the dataframe before calling the markdown formatter.

EDIT: Hm, the formatting looks fine here, but locally I get a lot of extra decimals. I have the same pandas and tabulate versions, weird.

[letmaik]

Looking at the docstring, disable_numparse actually does support specifying individual columns, so all good.

EDIT: oh dear... still not working as it should... sorry about the spam

[letmaik]

OK, that was more fiddly than expected but it's all working now and ready to be merged.

[achamayou]

It looks great, thank you!