Bohu

@JackTan25 Why we disable the UDF in the `COPY INTO ...`? The UDF is useful for transform data here. The code: https://github.com/datafuselabs/databend/pull/13848/files#diff-294914ae3bf92e84309c54d27dfaed6d5c51e696d673a4fdc1e725d18d28fe84R338-R342

@yufan022 Yes, we should. Several considerations remain: - Handling forward failures. - Determining the layer for forwarding: Should it occur in the HTTP handler or at a lower level?

Here's a condensed version of the protocol: 1. Client sends a request to query-1 (warehouse-id: w1). 2. Client sends the next request to query-2 (warehouse-id: w2), including a header with...

What's the performance? It's better to add to the PR summary.

Conflicting files src/query/service/src/interpreters/interpreter_copy_into_location.rs

Suggestion: This PR appears not ready for review as it lacks logic or unit tests. I'll set it to draft status :)

## Benchmark Round-1: Many rows ### Generate Data ``` INSERT INTO log SELECT (FLOOR(RAND() * 1000000) % 10000)::INT, JSON_OBJECT( 'user_ip', CONCAT_WS( '.', FLOOR(RAND() * 255)::STRING, FLOOR(RAND() * 255)::STRING, FLOOR(RAND() *...

## Benchmark Round-2: A few rows(152 rows) ### Generate Data ``` INSERT INTO log SELECT (FLOOR(RAND() * 1000000) % 10000)::INT, JSON_OBJECT( 'user_ip', CONCAT_WS( '.', FLOOR(RAND() * 255)::STRING, FLOOR(RAND() * 255)::STRING,...

@Xuanwo may be able to assist with this issue.

The another weird is: tsv: bytes scanned: 15.18GB csv: bytes scanned: 75.55GB Does this also related to the io prefetching?