ValkeySearch 1.2.0

[adam-fowler]

ValkeySearch are about to release 1.2. They have completely rebuilt their API model files and the generated code is completely different. This PR attempts to cleanup some of the breaking changes. Some features have been removed eg FT.SEARCH LOCALONLY, FT.AGGREGATE WITHCURSOR, FT.AGGREGATE SCORER

I am also going to suggest a couple of changes to their API model files to make the generated functions cleaner. I'll link to that PR when I post it.

EDIT: While looking at the json model file I found multiple issues with it. I have opened a PR with edits in the valley-search repo, but I thought I would add a phase in the command builder to cleanup up the valkey json model files. The cleanup includes

• Removal of unnecessary blocks (each block generates a new structure)
  • Blocks that consist of a pure-token and one other parameter are collapsed down to just the second parameter with the token field set.
  • Blocks that consist of one parameter, are collapsed down to the parameter
  • Remove unnecessary counts from API where the count can be ascertained from the size of an array.
• Added ability to patch json files inside code generator. Currently you can replace arguments wholesale or just edit one field of the argument. I use this twice in the valkey-search json to produce cleaner code.

[adam-fowler]

@eric-musliner you may find this of interest

[github-actions]

✅ Pull request no significant performance differences ✅

Summary

New baseline 'pull_request' is WITHIN the 'main' baseline thresholds.

Full Benchmark Comparison

Comparing results between 'main' and 'pull_request'

Host '50a6a63ea391' with 4 'x86_64' processors with 15 GB memory, running:
#17~24.04.1-Ubuntu SMP Mon Dec  1 20:10:50 UTC 2025

ValkeyBenchmarks

Client: GET benchmark metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	51	55	56	58	58	58	58	6
pull_request	51	54	55	57	57	57	57	6
Δ	0	-1	-1	-1	-1	-1	-1	0
Improvement %	0	2	2	2	2	2	2	0

Client: GET benchmark | parallel 20 | 20 concurrent connections metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	52	56	58	60	61	63	63	25
pull_request	51	55	58	60	61	65	65	25
Δ	-1	-1	0	0	0	2	2	0
Improvement %	2	2	0	0	0	-3	-3	0

Connection: GET benchmark metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	4	4	4	4	4	4	4	8
pull_request	4	4	4	4	4	4	4	8
Δ	0	0	0	0	0	0	0	0
Improvement %	0	0	0	0	0	0	0	0

Connection: GET benchmark – NoOpTracer metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	4	7	8	9	9	9	9	7
pull_request	4	7	8	9	9	9	9	7
Δ	0	0	0	0	0	0	0	0
Improvement %	0	0	0	0	0	0	0	0

Connection: Pipeline array benchmark metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	14	18	25	29	29	29	29	7
pull_request	14	18	25	29	29	29	29	7
Δ	0	0	0	0	0	0	0	0
Improvement %	0	0	0	0	0	0	0	0

Connection: Pipeline benchmark metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	7	24	24	25	25	25	25	7
pull_request	7	24	24	25	25	25	25	7
Δ	0	0	0	0	0	0	0	0
Improvement %	0	0	0	0	0	0	0	0

HashSlot – {user}.whatever metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	0	0	0	0	0	0	0	18
pull_request	0	0	0	0	0	0	0	18
Δ	0	0	0	0	0	0	0	0
Improvement %	0	0	0	0	0	0	0	0

ValkeyCommandEncoder – Command with 7 words metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	0	0	0	0	0	0	0	830
pull_request	0	0	0	0	0	0	0	830
Δ	0	0	0	0	0	0	0	0
Improvement %	0	0	0	0	0	0	0	0

ValkeyCommandEncoder – Simple GET metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	0	0	0	0	0	0	0	2004
pull_request	0	0	0	0	0	0	0	1994
Δ	0	0	0	0	0	0	0	-10
Improvement %	0	0	0	0	0	0	0	-10

ValkeyCommandEncoder – Simple MGET 15 keys metrics

Malloc (total): results within specified thresholds, fold down for details.

Malloc (total) *	p0	p25	p50	p75	p90	p99	p100	Samples
main	0	0	0	0	0	0	0	380
pull_request	0	0	0	0	0	0	0	377
Δ	0	0	0	0	0	0	0	-3
Improvement %	0	0	0	0	0	0	0	-3

[adam-fowler]

This PR to the json files valkey-io/valkey-search#832 reduces the generated code by 400 lines

[eric-musliner]

@eric-musliner you may find this of interest

Thanks for the awareness! Glad I'm not making use of the ones they are cutting

[adam-fowler]

@eric-musliner I have been playing with some code to clean up badly setup json files. You can see the list of things it does in the PR description.

This completely changes the API for valkey-search but I think it is a lot cleaner. Could you have a look and give us some feedback.

[eric-musliner]

I really like these changes to the API for valkey-search 1.2.0. I found the previous generated code for the commands when I did it a few months ago very clunky. I didn't like that the blocks with single pure tokens were turned into their own types. You've resolved the main issues I had with it originally. I think this looks great and it's a lot more ergonomic now.

[nilanshu-sharma]

Did a first pass and added some comments

[nilanshu-sharma]

Did another pass, added some comments.