Fork me on GitHub
#xtdb
<
2023-11-07
>
Proctor23:11:52

I was looking through the docs on XTDB, and there does not seem to be an option to enable the AcceptEncoding header to get compressed results when using the new-api-client with the URL, is that correct?

Proctor23:11:37

We noticed that longer key names cause the result to come back slower than doing aliases like s0 , s1, etc. in the :find and :where to get the return values

Proctor23:11:46

doing that extra work for re-naming keys in the where seems that doing extra work there vs just a (pull e [*]) implies that it is data content transfer that makes it take twice as long

Proctor23:11:10

we were seeing about 150ms to 200ms return for the longer key names (around 80 characters) vs being reduced to ~50ms for sXX key names for about 20 items on the document

Proctor23:11:25

for 50 document results

Proctor23:11:18

(the 80 characters come about from adding namespaces to the keys and for uniqueness and differentiation between related values)

refset09:11:18

Hey @US03ZP2F5 thanks for sharing your analysis and feedback - unfortunately that looks to be the case currently. clj-http exposes the options needed but we're not making use of them currently: https://github.com/xtdb/xtdb/blob/f65c4a398584ec03830509baa2085f3a880a1694/modules/http-client/src/xtdb/remote_api_client.clj#L61-L66 PRs are always welcome (should you be interested to propose a fix 🙂) but I've raised an issue for the time being: https://github.com/xtdb/xtdb/issues/2911

djtango09:11:42

oh just to be precise, we saw the slow behaviour when returning maps with long key names and switched to returning tuples instead

👍 1
djtango09:11:05

but same difference I'm sure