zhangshenghang opened a new pull request, #13149: URL: https://github.com/apache/seatunnel/pull/13149
Yes, I am contributing to Apache SeaTunnel. ## What documentation issues were found All issues below were verified against the actual source code on the current `dev` branch before editing. Three areas audited (messaging connectors, transforms, NoSQL connectors - 40+ pages) were confirmed clean and are intentionally untouched. **JDBC connector (source + sink, en + zh):** 1. The `dialect` option told users to set `dialect = starrocks`, but "StarRocks" is not a dialect name - it is a `compatible_mode` value. Setting it leaves no matching dialect factory and job initialization fails (`JdbcDialectLoader` throws instead of falling back). Same for "Inceptor" (selected via `compatible_mode` on the HIVE dialect). 2. The dialect list tables listed `StarRocks` and `Inceptor` as dialect names; replaced with real dialect names verified from `dialectFactoryName()` (`OpenGauss`, `Presto`, `DuckDB`). 3. The claim "unsupported dialect falls back to GenericDialect" is only true when `dialect` is unset (URL path); an explicit unmatched value fails. Clarified both languages. 4. `use_select_count` / `skip_analyze` were documented as Oracle-only, but the YashanDB dialect honors them too (`YashanDbDialect`). 5. `split.string_split_mode` and `split.string-strategy` are enum options in code (`enumType(...)`), not plain strings. **Kafka connector:** 6. Sink intro claimed "by default we use 2pc for exactly-once", contradicting the option table itself: `semantics` defaults to `NON` (en + zh). 7. Both protobuf source examples lacked the required `schema` block. Without it, `format = protobuf` is silently ignored and falls back to the text deserializer (`KafkaSourceConfig`). Added working `schema` blocks mirroring the passing e2e config, plus an explanatory note (en + zh). 8. zh sink claimed `kafka_headers_fields` "configure carefully" with `format = NATIVE`; the code actually throws `OPERATION_NOT_SUPPORTED` at serializer construction. zh aligned with the (correct) English page. 9. `[formats](../formats)` links pointed to a directory with no index page; fixed to `../formats/overview.md` (Kafka + Pulsar source, en + zh). **File connectors:** 10. FTP source Key Features checklist missed `parquet`/`orc`, which the same page's own option table and examples use (en + zh). 11. GCS source `field_delimiter` default was documented as `\001` for both text and CSV; code defaults to `,` for CSV (`CsvReadStrategy`). 12. GCS source option table had 4 duplicated rows (`csv_use_header_line`, `file_filter_modified_start/end`, `sort_files_by_modification_time`) - removed the duplicates (en + zh). 13. COS source `null_format` was documented as text-only; CSV reading honors it too (en + zh). 14. zh sink CosFile/FtpFile `field_delimiter` said "text only" while en said "text and csv" (code applies it to CSV writes); zh aligned. **Zeta engine:** 15. `event-listener.md` (en + zh) documented non-existent event types `READER_OPEN`/`READER_CLOSE`/`WRITER_OPEN`/`WRITER_CLOSE` and a non-existent `WriterOpenEvent` class. Replaced with the real `EventType` values (`LIFECYCLE_*`, `READER_MESSAGE_DELAYED`) and the example's wrong imports/switch cases fixed so the sample code now matches the API. 16. `rest-api-job-lifecycle.md` (en + zh): the `/logs/:jobId` endpoint was described as returning "last N lines"; it returns a list of matching log files with `logLink` entries (`LogService`). Quoted the real duplicate-job-id error message instead of an invented one, and replaced the misleading "Savepoint path not found" troubleshooting row that contradicted the page's own statement that REST cannot target a savepoint path. 17. zh `local-mode-deployment.md` used the deprecated `-e local` flag (deprecated since 2.3.1); fixed to `-m local`, matching the en page and the same zh page's own second example. **Documentation index (sidebars.js):** 18. Ten existing user-facing docs were unreachable from the navigation: `transforms/encrypt`, `getting-started/recipes/jdbc-to-jdbc`, `introduction/concepts/incompatible-changes`, `introduction/concepts/metadata-spi`, `introduction/concepts/gravitino-type-mapping`, `engines/zeta/stain-trace`, `engines/zeta/stain-trace-quickstart`, `engines/zeta/realtime-observability` (all exist in en and zh). Registered them in the sidebar. The PoC docs under `developer/` were left unlisted intentionally (one is explicitly marked as such). **Chinese documentation missing large sections vs English (verified same-content sync):** 19. `zh/connectors/source/OssJindoFile.md` was missing ~30 option subsections and the whole Example section (172 vs 592 lines). 20. `zh/connectors/source/MyHours.md`, `Persistiq.md`, `SftpFile.md`, `SqlServer-CDC.md` were missing option sections / example sections / changelog coverage vs en. All synced sections keep code blocks byte-identical to the English docs (verified by scripted diff). ## What was changed 29 files: 15 English, 13 Chinese, 1 sidebar config. Net +1103 / -91 lines of real documentation content (option tables, examples, descriptions, links, index entries). ## Duplicate PR check (last 7 days) Checked `apache/seatunnel` PR list immediately before submitting: recent docs PRs are #13135, #13097, #13072, #13050 (my own earlier fixes), #13148 (RegexExtract rewrite), #13122 (FilterRowKind), #13118 (Split). The files touched by the open PRs (#13135/#13097/#13072/#13050) were explicitly excluded from this change, and none of the other PRs overlap with the areas fixed here. ## Verification - Scripted link integrity scan over all en/zh markdown (relative targets resolved against the repo, code fences excluded): no broken links introduced; the only remaining hits are regex artifacts inside code fences. - Code-fence balance check on all 29 modified files: all balanced. - `docs/sidebars.js` validated by requiring it in Node; all newly registered ids were checked to exist as files under both `docs/en` and `docs/zh`. - For every option/default/type claim, the defining Java source (Options classes, factories, dialect loader, deserializer, REST service, EventType enum) was read and quoted before editing. - A full docusaurus build locally was not run because the site build lives in the separate `apache/seatunnel-website` repository (CI syncs this repo into it); the GitHub `document` workflow will run that build on this PR. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
