[CHORE] Explain block_on function in common-runtime

[colin-ho]

Addresses: #3435

[colin-ho]

@andrewgazelka Lmk if it makes sense

[andrewgazelka]

it makes sense. so this function is kinda jank then? cause the sync function should probably be async in that situation? since trend seems we are moving to asyncland except on top level might wanna document that this is kinda a workaround to use this then and maybe at some point we can deprecate it?

[codspeed-hq]

CodSpeed Performance Report

Merging #3442 will degrade performances by 29.99%

_{Comparing colin/explain-runtime-block-on (42e43d6) with main (2db1233)}

Summary

❌ 2 regressions
✅ 15 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	colin/explain-runtime-block-on	Change
❌	test_iter_rows_first_row[100 Small Files]	236.1 ms	337.2 ms	-29.99%
❌	test_show[100 Small Files]	23.2 ms	31.3 ms	-25.68%

[colin-ho]

it makes sense. so this function is kinda jank then? cause the sync function should probably be async in that situation? since trend seems we are moving to asyncland except on top level might wanna document that this is kinda a workaround to use this then and maybe at some point we can deprecate it?

Just for more context, the main issue is calling expression evaluation trait methods which are sync.

#[typetag::serde(tag = "type")]
pub trait ScalarUDF: Send + Sync + std::fmt::Debug {
    fn as_any(&self) -> &dyn Any;
    fn name(&self) -> &'static str;
    fn evaluate(&self, inputs: &[Series]) -> DaftResult<Series>;
    fn to_field(&self, inputs: &[ExprRef], schema: &Schema) -> DaftResult<Field>;
}

The url-download expression implements ScalarUDF, and in the evaluate method it needs to call some async code, i.e. to download from the url. But the evaluate trait method is sync, and it is called from the executor, which is in a tokio runtime context.

So we need a way run the async url download code, thus we came up with the block_on method.

Of course, ideally we just want to somehow do a url_download.await from the executor.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 77.35%. Comparing base (2db1233) to head (42e43d6).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3442   +/-   ##
=======================================
  Coverage   77.35%   77.35%           
=======================================
  Files         684      684           
  Lines       83637    83639    +2     
=======================================
+ Hits        64695    64698    +3     
+ Misses      18942    18941    -1     

Files with missing lines	Coverage Δ
src/common/runtime/src/lib.rs	90.78% <ø> (ø)

... and 4 files with indirect coverage changes