[tristor]

I wonder if you've benchmarked your Go implementation against the Percona Lab's FDW for Clickhouse? https://github.com/Percona-Lab/clickhousedb_fdw

[arunk-s]

Hi, Author here.

I have to admit, I haven't done any benchmarking against the existing FDWs for Clickhouse. But I actually wrote the Go FDW 2 years ago(around Dec 2018). ;)

There weren't any Clickhouse FDW available at that time and I probably would've tried them as well.

Now I just got around to write the blog post and convincing the team to release the code.

Though I have a suspicion that the percona FDW might win in the benchmarks as they won't have to pay the penalties when crossing the Go land to C land.[1]

1: https://www.cockroachlabs.com/blog/the-cost-and-complexity-o...

[tristor]

Thanks for the reply. I really appreciate your post and the additional FDW. If you ever decide to give it a go, a benchmark would be interesting, if for no other reason than to see how much penalty cgo actually incurs vs possible efficiency gains in your implementation strategy.

[arunk-s]

Yes for sure. Definitely a benchmark would be a very solid approach to see the efficiency differences.

[AhtiK]

There's now also a more recent CH FDW written in C by the team at adjust, https://github.com/adjust/clickhouse_fdw

[tristor]

Good find, looks like a more up to date fork of the Percona FDW.

[AhtiK]

Yeah, the original Percona FDW hasn't been active for a while, only had a few commits after the first release and was missing quite a few features. Also doesn't support anything more recent than Postgres 11.

So the "adjust" CH FDW is the only usable choice today, thankfully quite actively maintained, primarily by the Adjust devs. I'd presume it's used for their production load.