Cluster code micro-optimization

[klaussfreire]

Description of change

A bunch of low-hanging micro-optimizations.

Mostly common subexpression extraction to local variables when beneficial. Barely noticeable benchmark impact but technically an improvement without readability impact IMO.

One of the most impactful changes concentrates around key extraction and node determination.
It optimizes the most common path for key extraction which is that of single-key commands,
precomputing key position during command parsing to have a simple fast path in get_keys.

This last one does produce a large impact in cluster benchmarks, of around 20%.

Some optimizations also apply to regular (non-cluster) clients, but the main concern of this optimization effort was around cluster ops, so they were not benchmarked.

Benchmark results

This uses a benchmark similar to those in basic_operations, but with a cluster client.
Two pipeline patterns were used:

• hset+x: hset followed by hexpire, each hset + hexpire call counts as a single "request"
• setex: to do both at once
  Both patterns include value payloads of around 50 bytes each.

The first one stresses multiple command types with various return types, while the second one stresses a streamlined single-command pipeline with limited response data.

The benchmark was run in AWS against a 3-node valkey EC cluster (EC versions of Redis OSS does not support setex), using hiredis 3.3.1. The setup builds a mostly CPU-bound case with a very fast but still realistic network between client and cluster.

Large pipelines show as dnf because they get deadlocked, something that will be covered on another PR.

Numbers are given in requests per second (RPS) and relative uplift.

Pipeline size	master hset+x	master setex	micro_opt hset+x	micro_opt setex	uplift hset+x	uplift setex
100	14285	23292	16973	26705	18.8%	14.6%
10000	17598	31909	22116	39784	25.7%	24.7%
50000	15785	30328	19386	36752	22.8%	21.2%
150000	dnf	dnf	dnf	dnf
300000	dnf	dnf	dnf	dnf

Pull Request check-list

Please make sure to review and check all of these items:

• Do tests and lints pass with this change?
• Do the CI tests pass with this change (enable it first in your forked repo and wait for the github action build to finish)?
• Is the new or changed code fully tested?
• Is a documentation update included (if this change modifies existing APIs, or introduces new ones)?
• Is there an example added to the examples folder (if applicable)?

NOTE: these things are not required to open a PR and can be done
afterwards / while the PR is open.

---

Note

Medium Risk
Changes sit on cluster slot/key routing and command key parsing; behavior should match prior logic but mistakes could misroute commands or break edge cases (subcommands, negative key positions).

Overview
Cluster routing and pipelines get the biggest wins: CommandsParser now tags single-key commands with _single_key_pos at init and takes a fast path in get_keys, plus _is_keyed_command so cluster pipelines can cache keyed CommandPolicies for simple commands. _determine_nodes caches resolved policy callbacks per command name (capped at 5000 entries), and policy lambdas take (self, command, …) so one unified policy_cb(self, …) call replaces per-policy branching. Pipeline execution reuses local refs, per-batch policy_cache, and the keyed-command cache when inferring policies.

Smaller hot-path tweaks include a no-options fast path in Redis.parse_response, early str handling in Encoder.encode, bool_ok without always decoding to str, bounded split in policy resolvers, a local reader in async hiredis read_response, and benchmark StringJoiningConnection joining list commands before sendall.

^{Reviewed by Cursor Bugbot for commit b37fdbb. Bugbot is set up for automated code reviews on this repo. Configure here.}

[jit-ci]

Hi, I’m Jit, a friendly security platform designed to help developers build secure applications from day zero with an MVS (Minimal viable security) mindset.

In case there are security findings, they will be communicated to you as a comment inside the PR.

Hope you’ll enjoy using Jit.

Questions? Comments? Want to learn more? Get in touch with us.

[petyaslavova]

Hi @klaussfreire, thank you for your contribution! I'll take a look at it shortly.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

There are 2 total unresolved issues (including 1 from previous review).

^{Reviewed by Cursor Bugbot for commit f3bff9e. Configure here.}

[klaussfreire]

Hi @klaussfreire, thank you for your contribution! I'll take a look at it shortly.

Hi. Is there anything I can do to help move this forward?