improve aggregation performance

[githorse]

I've performed an informal grouping-and-aggregation benchmark across several popular grids (MUI, Material React Table, raw Tanstack, Adazzle, and AG Grid) in this CodeSandbox example.

Rough Loading Time (ms)¹

Rows	MUI	MRT	Tanstack	AG	Adazzle	Best Possible
1k	250	500	250	50	50	50
10k	1000	1000	500	250	50	50
50k	13000	1600	1200	800	150	150
100k	💣2	2500	2500	1200	300	300

Tanstack's performance is not bad, but it's roughly twice as slow as AG Grid and 10x slower than the fastest solutions. Ok, maybe the performance of the customized solution isn't really comparable to that of a general grid solution like Tanstack, but something close to whatever AG Grid is doing ought to be possible. (One simple solution might be to use Alasql).

Questions:

1. Is there anything I can do as a user of Tanstack to accelerate or optimize aggregations?
2. Are there any optimizations Tanstack can do to bring the performance of aggregations closer to AG Grid's (or, ideally, closer to my "best possible" algorithm)?

Footnotes

1. Most grids don't have a good way to instrument the exact loading time, so these are basically eyeballed and not very accurate. Obviously they'll be different on different machines. Nevertheless the overall trends are clear enough. ↩

2. Somewhere between 50k and 100k records the MUI grid blows up and kills the whole browser tab. ↩