fix(prices): batch get_multiple — one request per 20 codes, not per code (api#7240)

[karlwaldman]

Closes #7240 (filed as api#7240).

The defect

get_multiple() looped over get(), issuing one HTTP request per commodity. The REST API accepts up to 20 codes in one request that counts once against quota.

So the method whose entire purpose is fetching several prices cost up to 20× more quota than writing the call by hand. Through it, the free plan was 50 code-reads a day; through the raw API it is 1,000.

That is a plausible contributor to a measured result: sdk-python converts at 2.2% while people using plain Python and ignoring our SDK convert at 4.2% (api#7214). Our SDK made the free plan twenty times harder to live inside than not using it.

Verified against production before writing the fix

 20 codes -> HTTP 200
 21 codes -> HTTP 400  "Too many commodity codes requested (max: 20, requested: 21)"

And one multi-code call writes one api_requests row — confirmed by making a marked 7-code call and finding a single row for it. Response shape: 1 code → flat data; 2+ → data.prices[].

Both clients

The async one was worse: asyncio.gather fanned out one request per code concurrently, which also trips the 60-per-60s rate limit on a long list. It now gathers chunks — 25 codes are 2 concurrent requests, not 25.

Failure contract preserved

The API rejects the whole request when any code in it is unknown, so a chunk failure does not say which code was at fault. That chunk — and only that chunk — is retried per code to rebuild the per-code failure list. A one-code chunk has nothing to narrow down and is recorded directly rather than refetched.

Live smoke against production

 10 codes  -> 1 request,  10 prices    (was 10 requests)
  1 code   -> 1 request,   1 price     BRENT_CRUDE_USD $93.60
 30 codes  -> 2 requests, 29 prices    (was 30 requests)
 bad code  -> 3 requests,  1 ok + 1 reported failure

Tests

Four new tests assert request count, not just results — this regression is invisible without that. And two existing tests asserted call_count == 2 for two codes, encoding the defect; they now assert 1 and carry a note explaining why they changed.

645 passed, 42 skipped, 0 failed
coverage 63.38% (gate 50%)

Repo-wide ruff/black findings are pre-existing and unchanged — prices.py and async_client.py are 0 before, 0 after.

Behaviour change worth noting in release notes

Existing users' quota consumption drops — in their favour, but it is a change.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JKAExynd9zoKwt6rYA66EA

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 2b0c95d9-9b87-49e4-85c5-b001d5e7a67e

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}