Avoid q2 per-row global code remap

[snissn]

Summary

• replace q2 sorted grouped-distinct per-row global code materialization with per-part local-code-to-global-rank maps
• translate local dictionary codes to global ranks at iterator boundaries
• keep grouped output dictionary available while avoiding distinct global dictionary/global code arrays

Production-path caveat

A 1M JSONBench full-data q2 one_shot_end_to_end / no_aggregate_metadata run against this branch showed the production q2 cell currently uses dense grouped-count-distinct (dense_group_count_distinct_used=true, sort_layout=time_us), not the sorted grouped-distinct path changed here.

Latest artifact:

• /tmp/jsonbench_q2_lazy_1m_pUFYuebb/report.json
• total 85.240ms, setup 65.209ms, run 19.986ms, post-prepare 40.849ms
• dense q2 split: group rank 0.061ms, distinct rank 34.566ms, local rank 6.221ms
• rows scanned/matched/reduced 1,000,000 / 954,611 / 954,611
• result hash b63b8e1013c918fcda7c299ef64beb20c4988854cf0671a83803b6951d795860

So this PR is a valid sorted grouped-distinct algorithmic experiment, but it should not be presented as the production JSONBench full-data q2 improvement. The production q2 follow-up now targets dense grouped-count-distinct distinct-rank/local-rank setup; see #3324.

Claim boundary

This is no-aggregate one-shot typed-column q2 setup work only. It is not metadata acceleration evidence and is not a ClickHouse superiority claim. The PR stays draft until a production or harness q2 cell that actually exercises the sorted grouped-distinct path has current 1M evidence.

Validation

• git diff --check
• GOWORK=off go test ./TreeDB/collections -run 'TestColumnPhysicalTypedColumnOneShotCacheQ2NoMetadata3123|TestTypedColumnQ2.*1950|TestTypedColumnQ2DenseGroupCountDistinctRankMapCapacity3158' -count=1
• GOWORK=off go test ./TreeDB/collections -count=1

Evidence still required

• 1M q2 cell that exercises the sorted grouped-distinct path: setup, q2 post-prepare split, run, total
• preserve rows scanned/matched/reduced and result hash
• include q1/q3 no-regression samples if shared setup behavior is affected

Refs #3324. Refs #3070.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 40fe45ec-8676-41e8-af8c-ddcbf0445c01

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/q2-post-prepare-lazy-3324

---

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.}

[snissn]

Closing as superseded by the refreshed sorted-path q2 draft #3413 and the current dense grouped-count-distinct production lane in #3324. This PR remains useful historical/reference evidence, but it is not merge-ready production q2 one-shot/no-metadata evidence.