[doc](4.1.4) Update Chinese 4.x docs for the 4.1.4 release - #4123
Merged
Conversation
Review every commit between `4.1.3` and `4.1.4-rc04` (431 commits, 340 of which touch production code) and bring the Chinese 4.x documentation in line with the release. Chinese only for now; the English pages and `dev` will be synced after review. Corrections to statements that were already wrong: - `require_partition_filter` claimed 4.1.2 support on the 4.1 line; it actually lands in 4.1.4. - Multimodal `EMBED` claimed 4.1.5; it lands in 4.1.4. - The Paimon catalog page claimed a 1.0.0 dependency; 4.1.3 shipped 1.3.1 and 4.1.4 ships 1.4.2. - Several "since 4.0.8" notes only reached the 4.1 line in 4.1.4 (`fe_meta_auth_token`, `autobucket_min_buckets`, group commit BE forwarding, Stream Load compute group, Routine Load `FirstErrorMsg`, `enable_file_cache_write_index_file_only`). New or changed user-facing surface now documented: - `ADMIN COMPACT TABLET`, removal of `iceberg_meta()`, removal of the `/api/.../upload` endpoints, `SHOW COMPUTE GROUPS` in non-cloud mode. - `parse_to_variant()` / `try_parse_to_variant()` (new pages). - Iceberg: VARIANT read/write, V3 default values and the `DEFAULT()` expression, nested-column DDL, `ALTER TABLE ... SET`, database property matrix, `$position_deletes`, OSS Tables REST catalog, mixed Parquet/ORC data files. - File Scanner V2, BE-side split refinement and external scan task reuse. - Streaming jobs: OceanBase source, MySQL schema-change sync, `LagBytes`, JDBC URL normalization, `snapshot_split_size` default. - Group commit WAL limits and follower forwarding; Stream Load privilege scope. - FLOAT/DOUBLE to string now uses the shortest round-trip form. - ANN indexes on merge-on-write Unique Key tables. - New FE and BE configuration entries, plus an "upgrading to 4.1.4" section covering removals, default changes and deployment changes. Add a session-variable reference page covering only the variables 4.1.4 touches, marked as work in progress. Paimon write support and the unified TLS framework are deliberately left undocumented for now. Note for reviewers: the three new pages are not yet wired into `versioned_sidebars/version-4.x-sidebars.json`, because that file is shared with the English site and the English pages do not exist yet. They will be added together with the English translation. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GRr7pw6ymrf99x3DUr2zxg
morningman
temporarily deployed
to
Production
September 8, 2026 03:25 — with
GitHub Actions
Inactive
Variant V2 is not a released capability yet, so remove the FE config `enable_variant_v2`, the ColumnVariantV2 section on the VARIANT type page, and the Iceberg / Paimon VARIANT column support that is gated on it. `parse_to_variant()` / `try_parse_to_variant()` stay: they are registered unconditionally in `BuiltinScalarFunctions` and their BE implementation builds a legacy `ColumnVariant`, so they do not depend on `enable_variant_v2`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GRr7pw6ymrf99x3DUr2zxg
morningman
temporarily deployed
to
Production
September 8, 2026 05:54 — with
GitHub Actions
Inactive
Port the Chinese 4.x changes to the English 4.x docs and to both dev trees, and wire the new pages into the sidebars now that their English counterparts exist. The dev docs describe `master`, which has diverged from `branch-4.1`, so they are not a copy. Verified against master and adjusted: - `enable_external_scan_task_reuse`, `file_split_size_on_fe`, `file_split_size_on_be` and `external_meta_cache_max_weight` do not exist on master, so they are omitted from dev. - `enable_expr_zonemap_filter` defaults to `true` on master (`false` in 4.1.4); `meta_service_rpc_rate_limit_enabled` defaults to `true` (`false` in 4.1.4); `max_scanners_concurrency` is still `4`, so the 4 -> 8 note is 4.1-only. - `paimon.version` is 1.3.1 on master and 1.4.2 on branch-4.1, and the `paimon-scanner` -> `paimon-connector` module rename is 4.1-only. - Iceberg database properties beyond HMS, `ALTER TABLE ... SET` on Iceberg tables, and several error strings differ on master; the dev text follows master. - `SHOW TABLETS` without `ORDER BY` but with `LIMIT` deliberately keeps scan order on master, unlike 4.1.4. Corrections to the Chinese 4.x pages found while porting: - `default.replication_num` / `default.replication_allocation` are not mutually exclusive; modifying one removes the conflicting legacy property, with a rolling-upgrade caveat. - The `enable_nereids_distribute_planner` refresh is gated on `variable_version < 400`, so it only applies to a 3.x -> 4.1.4 upgrade. - `enable_forward_group_commit_stream_load_to_follower` is mutable and `default_get_version_from_ms_timeout_second` is not; both were stated the wrong way round. - `SHOW COMPUTE GROUPS`.`BackendNum` counts every backend in the group, not only the live ones. - The floating-point output example contradicted the new shortest round-trip rule. - The MaxCompute `test_connection` row had been inserted mid-table, splitting it, and named the property `mc.enable_namespace_schema` instead of `mc.enable.namespace.schema`. Also port `schema-change-mysql.md` from dev down to 4.x: MySQL schema change sync ships in 4.1.4 and the 4.x overview already points at it. `parse_to_variant` / `try_parse_to_variant` reuse the existing dev pages rather than the ones written for 4.x, with their version line corrected from 4.2 to 4.1.4 in every tree. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GRr7pw6ymrf99x3DUr2zxg
morningman
temporarily deployed
to
Production
September 9, 2026 04:27 — with
GitHub Actions
Inactive
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this PR does
Reviewed every commit between
4.1.3and4.1.4-rc04(431 commits, 340 of which touch production code) and brought the Chinese 4.x documentation in line with the release.Chinese only for now, on purpose — the English pages and
devwill be synced after this is reviewed.Corrections to statements that were already wrong
require_partition_filtersupported since 4.1.2EMBEDsupported since 4.1.5fe_meta_auth_token,autobucket_min_buckets, group commit BE forwarding, Stream Load compute group, Routine LoadFirstErrorMsg,enable_file_cache_write_index_file_onlyiceberg_meta()TVF documented as availableposition_deletessystem table "not supported yet"New / changed user-facing surface now documented
SQL and interfaces
ADMIN COMPACT TABLET <id> WHERE TYPE=...iceberg_meta(), the/api/.../uploadendpoints andhttp_load_submitter_max_worker_threadsSHOW CLUSTERS/SHOW COMPUTE GROUPSnow work in non-cloud mode, with different columns and relaxed privilegesparse_to_variant()/try_parse_to_variant()Lakehouse
initial-default/write-defaultand theDEFAULT()expression, nested-column DDL,ALTER TABLE ... SET, database property matrix,$position_deletes, Alibaba OSS Tables REST catalog, mixed Parquet/ORC data files,MERGE INTOduplicate-match check__HIVE_DEFAULT_PARTITION__now reads as a real NULLtest_connection, and the timeout/retry properties now taking effectfile_split_size_on_fe/file_split_size_on_be) and external scan task reuseLoad
ADD/DROP COLUMNsync andschema_change_enabled,Lag→LagBytes, JDBC URL normalization (TINYINT(1)andYEARmapping changes),snapshot_split_sizedefault 8096 → 40960group_commit_max_wal_num_per_table, follower forwardingLOAD_PRIVis now sufficientOther
--add-opens=java.base/java.nio=org.apache.arrow.memory.core,ALL-UNNAMEDchange)Deliberately left out
INSERT/INSERT OVERWRITE/ row-levelUPDATE/DELETE/MERGE INTO, the associated DDL andpaimon_jni_writer_memory_pool_limit_bytes) — not an officially released capability yet. The Paimon 1.4.2 dependency bump is kept.enable_variant_v2, the ColumnVariantV2 section on the VARIANT type page, and the Iceberg / Paimon VARIANT column support gated on it.parse_to_variant()/try_parse_to_variant()are kept: they are registered unconditionally and their BE implementation builds a legacyColumnVariant, so they do not depend onenable_variant_v2.English and dev are now synced
The third commit ports everything to
versioned_docs/version-4.x/(English 4.x),docs/(English dev) andi18n/zh-CN/.../current/(Chinese dev), and wires the new pages intoversioned_sidebars/version-4.x-sidebars.jsonandsidebars.ts.The dev trees are not a copy —
masterhas diverged frombranch-4.1, and every factual claim was re-checked against it:enable_external_scan_task_reuse,file_split_size_on_fe,file_split_size_on_be,external_meta_cache_max_weightenable_expr_zonemap_filterfalsetruetruemeta_service_rpc_rate_limit_enabledfalsetruetruemax_scanners_concurrency84paimon.versionpaimon-scanner→paimon-connectorALTER TABLE ... SETon IcebergSHOW TABLETSwithLIMITand noORDER BYSeveral Iceberg error strings also differ on master (
Cannot specify invisible column ...,DEFAULT requires a target column reference); the dev text follows master.Corrections found while porting
Reviewing each claim against the source turned up mistakes in the Chinese pages, now fixed in all four trees:
default.replication_num/default.replication_allocationare not mutually exclusive — modifying one removes the conflicting legacy property, with a rolling-upgrade caveat.enable_nereids_distribute_plannerrefresh is gated onvariable_version < 400, so it only applies to a 3.x → 4.1.4 upgrade, not to every upgrade.enable_forward_group_commit_stream_load_to_followeris mutable anddefault_get_version_from_ms_timeout_secondis not — both were stated the wrong way round.SHOW COMPUTE GROUPS.BackendNumcounts every backend in the group, not only the live ones (getAllClusterBackends(false)).cast('12345678' as float)→1.234568e+07) contradicted the new shortest-round-trip rule it sat under.test_connectionrow had been inserted mid-table, splitting it in two, and named the propertymc.enable_namespace_schemainstead ofmc.enable.namespace.schema.Also in this PR
schema-change-mysql.mdported from dev down to 4.x — MySQL schema change sync ships in 4.1.4 and the 4.x overview already points at it.parse_to_variant/try_parse_to_variantreuse the existing dev pages instead of the ones written for 4.x, with their version line corrected from 4.2 to 4.1.4 everywhere.Known gap
require_partition_filteris not documented in the dev trees at all — it was added to dev, removed from dev, then re-added to 4.x only. The property does exist on master, so dev has a real hole, but filling it means writing the feature up rather than porting this change.🤖 Generated with Claude Code
https://claude.ai/code/session_01GRr7pw6ymrf99x3DUr2zxg