[andy-stark-redis]

Summary

DOC-7267 ioredis follow-up (the ticket listed the other clients; this one turned out to be a different defect from redis_py).

The ioredis data was scraped from RedisCommander.d.ts by something that ran text together. Each entry is now matched to the overload it came from and rebuilt from that declaration:

• Signatures had stray commas and empty leading parameters (bitcount(, key: RedisKey, ...)). 175 are regenerated verbatim from the declaration. Text that already matched stays byte for byte.
• params arrays were often one garbled item (...args typed [key). 264 are rebuilt with one item per tuple element, as HIMPORT PREPARE already does. Existing descriptions are kept.
• Returns were Any on 416 of 429 entries. Every return is now the declared Result<T, Context>. No non-Any return changed.
• 16 entries glued several declarations together: one command's overload, a JSDoc comment, then other commands' overloads, in one string. Collapsed entries are split into their real overloads, so the file goes from 429 to 458 entries. The 16 pieces that belonged to other commands are dropped.

The merged data/command-api-mapping.json is regenerated and equals the per-command data. Its only semantic changes are the 154 ioredis lists. No other client changed. ioredis isn't in the API Methods tab's client list, so nothing renders differently.

How to review

• Reference version is ioredis 5.11.1, the latest 5.x. The harness pins ^5.4.0, and 430 of the 458 existing pieces match a 5.11.1 overload exactly, against 333 for 6.0.0. The six himport entries are the exception: himport only exists in 6.0.0, where they were authored, and they were already correct, so they are untouched.
• Every piece matches a declared overload by name and normalized parameters (quote style and redundant parentheses ignored). Two ambiguous ones (PFCOUNT, PFMERGE, where the old text had a rest parameter before callback, which isn't valid TypeScript) are matched by parameter names and rest-ness.
• A tsc --strict compile of typeof redis.<name> succeeds on the 5.11.1 Redis type for 153 of the 154 distinct method names. The exception is himport.
• HEXPIRE now lists ten entries and HPEXPIRE ten. Those were each already one entry that contained 8 or 9 real overloads (the NX, XX, GT and LT variants), so splitting preserves information. Whether a command should list that many is a separate question.

Not fixed

• Overload selection. The data lists a subset of each command's declared overloads (429 entries against 2,891 declared). I kept that selection and didn't try to define a better one.
• The 16 dropped pieces (hget, hmget, hrandfield, lmove, migrate, script, sdiffstore, sintercard, subscribe, sunionstore, xpending, zcount, incr, incrby, decr, decrby) were not added to those commands' own lists. They were stray fragments, not a reviewed selection.
• ioredis 6.x differs from 5.x on about 100 overload signatures (bitop and geoadd option names, for one). When the docs move to 6.x, rerun the match.

🤖 Generated with Claude Code

---

Note

Low Risk
These edits are documentation metadata only; there is no runtime or client-facing behavior change beyond regenerated JSON.

Overview
Refreshes the data/command-api-mapping JSON shards so ioredis entries mirror the actual TypeScript overloads instead of scraped noise.

Signatures drop stray commas/comments and split merged command strings into distinct overloads, while params now enumerate each positional/rest argument instead of a single ...args blob. Returns that were previously labeled Any now expose the declared Result<T, Context> shapes, and follow-on tweaks (e.g., mget/mset map vs tuple variants, HEXPIRE option breakdowns) align with current typings. data/command-api-mapping.json is regenerated from the per-command files so only the ioredis sections change; other clients remain untouched.

^{Reviewed by Cursor Bugbot for commit 2166b40. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

DOC-7267

[github-actions]

🧠 Redis Memory

Found 5 related items from repository history (5 new this commit):

• Commit — DOC-7267 Rebuild the ioredis entries from the declared overloads 🆕

  DOC-7267 Rebuild the ioredis entries from the declared overloads

• PR — DOC-7267 Fix malformed and non-callable redis_py entries 🆕

  Summary

• PR — DOC-7267 Make node_redis entries callable and typed, and add ctx to go-redis HOTKEYS 🆕

  Summary

• PR — DOC-7263 Fill missing client signatures in the command-API mapping 🆕

  Summary

• PR — DOC-7226 Replace deprecated node-redis quit()/disconnect() and fix harness-exposed examples 🆕

  First slice of DOC-7226 (deprecated client API calls in example code): the node-redis items.

Memory updated at 2166b40

[dwdougherty]

I'm not seeing ioredis entries in the pulldown menu, but maybe I'm looking for the wrong thing. That said, I'll go ahead an approve to speed you along.

[andy-stark-redis]

I'm not seeing ioredis entries in the pulldown menu, but maybe I'm looking for the wrong thing. That said, I'll go ahead an approve to speed you along.

@dwdougherty Yes, Claude seems to have figured that getting the data in place first merited a separate PR, which does make this one seem a bit odd. You should see ioredis appear soon :-) Anyway, thanks for the review!