script to export benchmark information as Line Protocol format

[logan-keede]

Which issue does this PR close?

• Closes Export benchmark information as line protocol #6107

Rationale for this change

a step towards #5504

What changes are included in this PR?

addition of python script that converts some **benchmark**.json to lineformat

Are these changes tested?

on some files in result.zip

[nix-shell:~/dev/datafusion/benchmarks]$ python3 lineformat.py ~/Downloads/results/alamb_sort-merge-accounting/sort.json 
benchmark,name=sort,version=28.0.0,datafusion_version=28.0.0,num_cpus=8 query="sort utf8",iteration=0,row_count=10838832,elapsed_ms=85626006 1691105678000000000
benchmark,name=sort,version=28.0.0,datafusion_version=28.0.0,num_cpus=8 query="sort utf8",iteration=1,row_count=10838832,elapsed_ms=68694468 1691105678000000000
benchmark,name=sort,version=28.0.0,datafusion_version=28.0.0,num_cpus=8 query="sort utf8",iteration=2,row_count=10838832,elapsed_ms=63392883 1691105678000000000
benchmark,name=sort,version=28.0.0,datafusion_version=28.0.0,num_cpus=8 query="sort utf8",iteration=3,row_count=10838832,elapsed_ms=66388367 1691105678000000000

Are there any user-facing changes?

No

[logan-keede]

@alamb Check this out.

do we have some documentation on json format? that would be helpful if we want to show more arguments in tag set.

[alamb]

@alamb Check this out.

do we have some documentation on json format? that would be helpful if we want to show more arguments in tag set.

I believe the JSON is a serialized version of this structure:

datafusion/benchmarks/src/util/run.rs

Lines 47 to 59 in 68306ac

	pub struct RunContext {
	/// Benchmark crate version
	pub benchmark_version: String,
	/// DataFusion crate version
	pub datafusion_version: String,
	/// Number of CPU cores
	pub num_cpus: usize,
	/// Start time
	#[serde(serialize_with = "serialize_start_time")]
	pub start_time: SystemTime,
	/// CLI arguments
	pub arguments: Vec<String>,
	}

This looks amazing @logan-keede -- I'll check it out later today

[alamb]

Thank you @logan-keede -- this is a great. I tried it on some of my own data and it works well ❤️

I am thinking through to the next steps (of trying to gather historical performance measurements, but I think I just need to try and gather that information and see what additional data is needed

[alamb]

Given this is just a conversion, I don't think there is any idea of baseline? As in maybe this comment needs to be adjusted

[logan-keede]

remnants of compare.py 😅

[alamb]

Would it be possible to add some comments here explaining what this script does along with the example you have in your PR description?

It would also be great to add some documentation to https://github.com/apache/datafusion/tree/main/benchmarks#comparing-performance-of-main-and-a-branch explaining how to use this script.

[logan-keede]

I have added some, check it out.

[alamb]

Perhaps we can rename this file from benchmarks/lineformat.py to benchmarks/lineprotocol.py as a way to have its name be a bit more self describing