DOC-6968 Fix two node-redis step names, and add the missing incr step for nine clients

[andy-stark-redis]

DOC-6968 — two node-redis step names that never matched their sets

Three lines changed across two files. The effect is larger than the diff: two docs pages stop dumping an entire example file where a short snippet belongs.

File	Was	Now
local_examples/tmp/datatypes/hashes/dt-hash.js	STEP_START hIncrBy_hGet_hMget	STEP_START incrby_get_mget
local_examples/tmp/datatypes/sets/dt-set.js	STEP_START sinster	STEP_START sinter
local_examples/tmp/datatypes/sets/dt-set.js	// >>> {'bike:1'}	// >>> ['bike:1']

What was actually broken

Both files always contained correct, working code for these steps. Only the name was wrong, so clients-example could not find the step and fell back to embedding the whole file:

• develop/data-types/hashes — the Node.js tab for incrby_get_mget rendered 120 lines instead of 14.
• develop/data-types/sets — the Node.js tab for sinter rendered 114 lines instead of 6.

Imports, connection setup and every unrelated example in the file, in a tab a reader opened to see one counter pattern. Nothing failed and no test broke — the tab was present and populated the entire time, which is exactly why this survived.

build/components/example.py:146 lowercases step names, so the other camelCase markers in these files (hmGet, hExpire, hpExpire) land on the canonical names by accident. These two didn't: one has the wrong name structure rather than the wrong case, and the other is a plain typo — sinster is not a command, a method, or a word.

The comment the rename exposed

Making the sinter pane visible surfaced a second defect in the six lines it revealed. The expected-output comment read {'bike:1'} — a Python set literal — in a JavaScript example whose own assertion two lines below reads assert.deepEqual(res7, [ 'bike:1' ]). Observed output is [ 'bike:1' ]. It had evidently been copied from the redis-py tab, and the whole-file fallback had buried it in 114 lines where nobody would read it against the assertion.

Worth noting as a pattern: this is the second time in this run of work that an expected value inherited from a sibling client's tab proved wrong for the client it was published under (the first was hpttl in #3808, inherited as 59994 where Ruby returns 59999).

Verification

Check	Result
Step names now match the reference client	✅ node-redis named_steps identical to Python's for both sets
Examples still run	✅ run.sh hash_tutorial node-redis and run.sh sets_tutorial node-redis both PASS against Redis 8.8.0
Output comments match reality	✅ hash step's 1, 2, 3, 1, 1, 3, ['1','1'] confirmed against the run; sinter corrected to the observed ['bike:1']
Panes now slice	✅ 14 lines and 6 lines respectively, checked in built HTML
Nothing else changed	✅ site-wide legacy-fallback panes 111 → 109, the two removed being exactly these, none added

Only comments and a marker name changed, so no runtime behaviour is affected — but both examples were run anyway rather than assumed.

Context

This is the "cause A" item from DOC-6968, which catalogues 74 panes site-wide rendering whole files this way. These two were the only ones fixable by a rename — the other ~64 are files that genuinely lack the example code and need examples written. The shortcode fix that makes the whole class impossible is tracked in the same ticket and has no prerequisites.

Doing these two before that fix is deliberate: afterwards the symptom becomes a quietly-omitted tab rather than an obvious whole-file dump, so a typo like sinster would get harder to spot, not easier.

🤖 Generated with Claude Code

---

Note

Low Risk
Changes are limited to local example sources, doc step markers, and the example test harness; no production services, auth, or data paths are affected.

Overview
Fixes DOC-6968 doc slicing and expands example coverage for incr and C (hiredis).

Node.js datatype tutorials: Renames mismatched STEP_START markers in dt-hash.js (incrby_get_mget) and dt-set.js (sinter), and corrects the sinter expected-output comment from a Python-style set to ['bike:1'], so those tabs render short snippets instead of whole files.

cmds_string / INCR: Adds a shared incr step (SET → INCR → GET) to nine clients (Python, Node, ioredis, Predis, Rust sync/async, Lettuce async/reactive, hiredis), with client-accurate >>> comments (e.g. redis-py True vs OK) and cleanup keys updated for mykey.

Example test harness: Registers hiredis as a portable runner (run_hiredis, system-library discovery, Linux-safe link order), sets portable to hiredis in clients.tsv, and on full sweeps emits explicit SKIP rows for clients excluded from the current mode so missing coverage is visible. testing.md documents hiredis limits and the new behavior.

^{Reviewed by Cursor Bugbot for commit cc44a07. Bugbot is set up for automated code reviews on this repo. Configure here.}

---

Second commit — the missing incr step for nine clients

Adds an incr step (SET mykey 10 → INCR → GET) to the redis-py, node-redis, ioredis, predis, rust-sync, rust-async, lettuce-async, lettuce-reactive and hiredis files in cmds_string. Nine of the twelve clients on content/commands/incr.md had no such step, so each of those tabs rendered its client's whole file — 12 to 51 lines — instead of three commands.

commands/incr.md now renders all 12 client tabs correctly sliced (verified in built HTML: 13 panes including redis-cli, none on the legacy path).

Why this is in the same PR, and why it comes first

The shortcode fix tracked in DOC-6968 omits a tab when the client lacks the requested step. On this page that would have cut 12 client tabs to 3, losing Python and Node.js — the two clients most readers arrive with — from a core string command page. Wrong-but-present beats absent for a reader, so the coverage lands before the fix rather than after it. With this in, the fix takes nothing from incr.md.

set() does not return OK everywhere

redis-py's set() returns a bool, so the Python tab's comment reads # >>> True where the Go and Jedis reference implementations read OK. Copying the reference value into the Python tab would have been wrong — the same failure mode as the two other defects on this branch, both from an expected value inherited off a sibling client's tab.

Verification

Client	Result
redis-py	✅ ran — True, 11, 11
node-redis	✅ ran — OK, 11, 11
ioredis	✅ ran — OK, 11, 11
predis	✅ ran — OK, 11, 11
rust-sync / rust-async	✅ ran, 1 test passed each; values confirmed by the in-step assert_eq! calls (cargo captures stdout)
lettuce-async / lettuce-reactive	✅ ran; values confirmed by the in-chain assertThat calls
hiredis (C)	✅ ran — OK, 11, 11 (compiled and executed after hiredis was installed locally; see the update below)

build/example-test-harness/run.sh cmds_string reports PASS for all nine, with the pre-existing jedis, go-redis and nredisstack implementations untouched and still passing. The remaining SKIP (no source) entries are clients with no file in this set at all.

The C gap is now closed. It was initially unverifiable here — clients.tsv gives hiredis no portable runner, and no hiredis headers were installed. With the headers now in place it compiles clean (gcc -Wall -Wextra; the only two warnings are pre-existing complaints about main's unused argc/argv) and runs against Redis 8.8.0, printing OK, 11, 11 — matching all three comments exactly, exiting 0, with no assertion-failure output and no keys left behind.

Note for anyone re-running this: hiredis remains untestable through run.sh, so this was a manual gcc + execute. The Gaps: trailer on commit 282798c still says the C implementation is unverified; that was true when written and is now stale — see the reflect-note comment on this PR.

Whole-site effect of both commits

Legacy-fallback panes across the entire built site: 111 → 100. The eleven removed are exactly the two renames plus these nine incr panes; none added.

Rendered panes were also checked by eye for Python, C and Java-Async: assertions and cleanup stripped, no REMOVE/HIDE/STEP_ scaffolding reaching the page, and the Lettuce chain still valid Java once its inner REMOVE blocks are removed.

[github-actions]

DOC-6968

[github-actions]

🧠 Redis Memory

Found 7 related items from repository history:

• PR — DOC-6909 Fix pre-existing broken links across the docs

  Part 1 of 4 of a stack that splits DOC-6909 Add link render hook to replace relref (investigation) #3732 into independently reviewable pieces, after @paoloredis noted the single PR was too much to review

• PR — DOC-6887 Add hash field expiration examples to hash data type page

  What

• PR — DOC-6951 Backfill aliases for 256 pages that moved without one

  Restores 256 dead URLs — item B1 of DOC-6951.

• PR — DOC-6954 Give the Client tools page its categories back

  Fixes DOC-6954. One deleted line.

• PR — DOC-6909 Migrate develop/clients/redis-py to portable Markdown links and callouts

  Part 4 of 4. Stack: DOC-6909 Fix pre-existing broken links across the docs #3795 → DOC-6909 Teach the feed pipeline and version archiver the /content/ link form #3796 → DOC-6909 Add link and blockquote render hooks to replace relref #3797 → DOC-6909 Migrate develop/clients/redis-py to portable Markdown links and callouts #3798 (this)

• Commit — DOC-6968 Fix two node-redis TCE step names that never matched their sets

  DOC-6968 Fix two node-redis TCE step names that never matched their sets

• Commit — DOC-6968 Add the missing incr step to nine cmds_string client examples

  DOC-6968 Add the missing incr step to nine cmds_string client examples

Memory updated at cc44a07

[andy-stark-redis]

Correction to commit 282798c's Gaps: trailer — the hiredis gap is closed.

That trailer says the C implementation is not execution-verified and should be run under --fidelity or compiled before its output is trusted. True when written; no longer true. hiredis headers were installed locally afterwards, and the example now compiles clean and runs against Redis 8.8.0, printing OK, 11, 11 — matching all three of its comments, exit 0, no assertion-failure output, no keys left behind.

Not amended into the commit because it was already pushed. /finalize should drop that Gaps: line from the durable squash message rather than carrying it forward, and replace it with the fact that matters:

hiredis cannot be verified through run.sh at all — clients.tsv gives it no portable runner, and the harness does not report it as a SKIP either, so it is simply absent from the results table. Anyone editing a C example needs to know that a green harness run says nothing about it, and that the verification route is a manual gcc -I/usr/local/include -L/usr/local/lib -lhiredis plus DYLD_LIBRARY_PATH=/usr/local/lib to execute.

That absence-not-SKIP behaviour is the durable lesson here, and it is worth considering against the skill's own gate ("every client PASS, or SKIP with a stated reason — a silent SKIP is a failure"), which hiredis silently evades.

[dwdougherty]

LGTM.

[cursor]

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

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 6a12a9a. Configure here.}

[andy-stark-redis]

Thanks @dwdougherty !