DOC-7015 Stop HSCAN/HVALS examples depending on hash field order

[andy-stark-redis]

DOC-7015

What

Redis doesn't guarantee hash field order under hashtable encoding. Forced hash-max-listpack-entries to 0 and confirmed (5 trials each, both encodings) six still-fragile assertions across the five clients this ticket named:

• redis-py: cmds_generic scan4's HSCAN NOVALUES list, cmds_hash's HVALS
• node-redis: cmds_generic scan4's HSCAN entries and NOVALUES list, cmds_hash's HVALS
• predis: cmds_hash's HVALS
• ruby: cmds_hash's HVALS

Fixed each by sorting (field-name lists and HVALS values) or pairing HSCAN's structured entries into an object (node-redis's scan4), matching the established ioredis/Go pattern from DOC-6968. lettuce-reactive needed no changes — its scan4 and HVALS were already sorted; DOC-6968's list of five affected clients had that one stale.

Folded in from DOC-7014

dt_hash.rb and dt_stream.rb documented Ruby Hash#inspect output as "key"=>"value" (no spaces), but Ruby 4.0 prints "key" => "value". Fixed all 27 lines, verified against a live run.

⚠️ This branch predates DOC-7014's merge, so these exact lines will conflict with DOC-7014's >>> prefix change when both land — trivial to resolve (apply both edits), but flagging so it's not a surprise.

Two unrelated bugs found, not fixed here

1. cmds_generic's scan2 step is inherently flaky. It asserts the last of five sequential SCAN calls returns exactly 18 matches. That split isn't guaranteed — reproduced 0,0,0,0,19 and 1,0,0,1,17 splits across repeated fresh Redis 8.8 instances, never landing on DOC-6968's originally observed 0,0,0,1,18. This blocks every full-script run of cmds_generic for redis-py, lettuce-reactive and predis (an uncaught exception halts the script before scan4 ever runs) — node-redis is accidentally immune because its equivalent check uses console.assert, which doesn't throw. Verifying this ticket's actual scan4/HVALS targets required isolated snippets that skip scan1-3 entirely.
2. Maven's surefire runs stale compiled classes. work/lettuce-reactive/target/test-classes/ still had a HashExample.class from an unrelated, earlier build; run_maven_java's cleanup only removes stale .java sources, not compiled classes, so surefire ran it alongside the current test and produced a spurious, unrelated failure. Removing the stale .class file fixed it for this session — the harness doesn't clean target/ between runs, so it will recur.

Both are captured as Gaps trailers on the commit for follow-up tickets.

Verification

• Each fix isolated and run 5 times under forced hashtable encoding (hash-max-listpack-entries=0) against a fresh Redis 8.8 container — all pass.
• Re-verified under default (listpack) encoding — all pass.
• Ran the full modified files through build/example-test-harness/run.sh under both encodings (cmds_hash for all 4 fixed clients, hash_tutorial/stream_tutorial for ruby) — all PASS.
• Local Redis restored to its original 7.2.7 state afterward.

🤖 Generated with Claude Code

---

Note

Low Risk
Changes are limited to local example scripts and expected-output comments; no application or library runtime behavior is modified.

Overview
Makes Redis client documentation examples deterministic where Redis does not guarantee hash field order, and updates Ruby sample output for Ruby 4.0 Hash#inspect formatting.

HSCAN / HVALS (DOC-7015): In cmds_generic scan4, node-redis now builds an object from HSCAN entries instead of asserting entry order, and both node-redis and redis-py sort field-name lists for NOVALUES-style checks. HVALS steps in node-redis, redis-py, predis, and ruby sort returned values (with brief comments) before print/assert, matching the existing ioredis/Go approach.

Ruby inspect (folded from DOC-7014): Comment-only fixes in dt_hash.rb and dt_stream.rb change documented inspect strings from "key"=>"value" to "key" => "value" so examples match Ruby 4.0 output; runtime assertions are unchanged.

^{Reviewed by Cursor Bugbot for commit 7012a20. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

DOC-7015

[github-actions]

🧠 Redis Memory

Found 7 related items from repository history (2 new this commit):

• Commit — DOC-6968 Stop the new hash examples depending on field order

  DOC-6968 Stop the new hash examples depending on field order

• Commit — DOC-6887 Fix hash field expiration examples after harness verification

  DOC-6887 Fix hash field expiration examples after harness verification

• PR — DOC-6968 Complete cmds_generic, and make C examples able to fail

  DOC-6968 — the four SCAN steps for seven clients

• Commit — DOC-6968 Add the four SCAN steps to seven cmds_generic client examples

  DOC-6968 Add the four SCAN steps to seven cmds_generic client examples

• PR — DOC-7014 Standardize dt_*.rb output comments to # >>> value

  DOC-7014

• Commit — DOC-7015 Stop HSCAN/HVALS examples depending on hash field order 🆕

  DOC-7015 Stop HSCAN/HVALS examples depending on hash field order

• PR — DOC-7017 Assert cumulative SCAN totals instead of pinning per-call counts in cmds_generic 🆕

  Summary

Memory updated at 7012a20

[dwdougherty]

LGTM.