ClickHouse
Roadmap 2023
hi, when the next release will be published? It has been a long time after the v0.3.2-patch11 release😁
hi, when the next release will be published? It has been a long time after the v0.3.2-patch11 release😁
Happy new year @IcebergXTY. Still recovering from omicron but I'll update roadmap and release v0.4.0(just trying to follow semantic versioning) in a few days.
Happy new year @IcebergXTY. Still recovering from omicron but I'll update roadmap and release v0.4.0(just trying to follow semantic versioning) in a few days.
Happy new year @zhicwu. Wish you a speedy recovery~
Hi, sorry if this is the wrong place to ask but I tried on the CH slack and they sent me here.
The README.md says "clickhouse-tcp-client will be available in 0.5" - is it possible to get any sense of how soon this feature might be released? Looks like it's on the roadmap for this year, but I realise things can get complicated and so I was wondering if that's still plausible - if it might be quite soon this year or if it might end up getting delayed?
Hi @alexander-newby, sorry I haven't spent much time on this project lately, so the status of tcp client is same as last year - an incomplete AbstractSocketClient plus a half baked Native data format processor. Since I can only work on this in my spare time, the rough estimate would be a month or so, if I can concentrate on the implementation 😂 So technically, it's still possible this year.
To be honest, I'm not sure if it's worthy of doing this. Looking at numbers at here, the overhead of http protocol is not huge. I think the data format matters the most, especially for ingestion. Having said that, I expect the NIO-based socket client can be used in http-client too so that we can support zero-copy for both protocols.
Hi @zhicwu, thank you for your response!
I had done some cursory performance testing when we moved over from housepower/ClickHouse-Native-JDBC to this driver, and it looked like we were taking quite a big performance hit to our overall processing time, which I attributed to the change from TCP to HTTP.
I hadn't considered the data format might be a factor, so thank you for pointing that out. I will have a closer look at that performance page and see if we can get the HTTP driver going a bit faster.
On the other hand, if you do manage to get the tcp client implemented, we will be happy to use it, but I understand if it doesn't get done if you're only building this on a voluntary basis. Thanks again!