zhangshenghang opened a new pull request, #12615:
URL: https://github.com/apache/seatunnel/pull/12615
## Purpose of this pull request
This PR fixes a batch of documentation inaccuracies found by systematically
comparing connector docs with the actual `Option` definitions in the source
code. Every fix below was verified against the corresponding Java source (file
paths noted inline). Both English (`docs/en`) and Chinese (`docs/zh`) docs were
checked and fixed together.
## Issues found & changes
### CDC connectors
- **SqlServer-CDC** (en + zh): the options table documents `connect.timeout`
(Duration, `30s`), but no such option exists — the code defines
`connect.timeout.ms` (`Long`, default `30000`) in
`connector-cdc-base/JdbcSourceOptions.java`. Also fixed `server-time-zone`
default: the doc said `UTC`, but the code defaults to `ZoneId.systemDefault()`.
- **PostgreSQL-CDC** (en + zh): three HOCON examples used `schema-names`,
which does not exist in the code (`PostgresIncrementalSourceOptions.java`
defines `schema-name`); the examples were silently ignored. Added the missing
`schema-name` row to the options table, fixed the `server-time-zone` default
(`UTC` → system default) and the `connect.timeout.ms` type (`Duration` →
`Long`).
- **MongoDB-CDC** (en + zh): the `format` option (`DEFAULT` /
`COMPATIBLE_DEBEZIUM_JSON`), which is part of the connector's option rule and
consumed in `MongodbIncrementalSource`, was missing entirely from the option
table. Added it.
### Hive
- **Hive source** (en + zh): the doc claimed support for text, CSV, parquet,
ORC, JSON and markdown, but `HiveTableUtils.parseFileFormat` only supports
text/parquet/orc (`"Hive connector only support [text parquet orc] table
now"`). Removed the CSV/JSON/markdown claims and the markdown-parsing section.
Also fixed a **reversed deprecation notice**: the doc marked `table_list` as
deprecated in favor of `tables_configs`, but in
`MultipleTableHiveSourceConfig.java` it is the `tables_configs` path that is
`@Deprecated` ("hive is structured, should use table_list"). Updated the option
rows, sections, and the multi-table example to use `table_list`.
`compress_codec` description corrected to text-only.
- **Hive sink** (en + zh): removed unsupported CSV/JSON file-format claims
(`HiveFormatUtils` supports only PARQUET/ORC/TEXTFILE).
### Other connectors
- **StarRocks sink** (en + zh): `batch_max_rows` and `batch_max_bytes` types
were swapped — code (`StarRocksSinkOptions.java`) declares `batch_max_rows` as
`int` and `batch_max_bytes` as `long`.
- **Paimon source** (en + zh): FAQ referenced a non-existent option
`paimon.catalog.type`; the real key is `catalog_type`
(`PaimonBaseOptions.java`).
- **Paimon sink** (en + zh): FAQ referenced a non-existent option
`paimon.auto-create-table`; table auto-creation is driven solely by
`schema_save_mode`.
- **Jdbc sink** (en + zh): `field_ide` is an enum
(`ORIGINAL`/`UPPERCASE`/`LOWERCASE`, `JdbcSinkOptions.FIELD_IDE`), not a
`String`.
### File connectors
- **FtpFile source/sink** (en + zh): the "Reading/Writing via SFTP" sections
described configuration that does not exist in the code —
`FtpConf.buildWithConfig` always builds `ftp://<host>:<port>` from
`host`/`port`, and `ftp_properties`, `fs.defaultFS`, and `known_hosts` support
do not exist anywhere in `connector-file` (SFTP host-key checking is
hard-disabled in `SFTPConnectionPool`). Replaced the sections with a short note
pointing users to the `SftpFile` connector. Also added `parquet`/`orc` to the
FtpFile supported-format list (they are supported by the shared read strategies
and used in FtpFile e2e tests), and fixed an example `schema` block that was
missing the required `fields { }` wrapper (the schema was silently ignored).
- **SftpFile source** (en + zh): added missing `parquet`/`orc` entries to
the supported-format list (the doc's own option table already claimed support).
- **HdfsFile sink**: moved `canal_json`/`debezium_json`/`maxwell_json` from
"compress codec" to "file format type" where they belong (en); added missing
`XML`/`binary` entries (zh).
### Transforms
- **DynamicCompile** (en + zh): fixed ` ```hacon ` code-fence marker typo →
` ```hocon ` (8 occurrences), which also fixes broken syntax highlighting.
## Files updated
28 files:
`docs/{en,zh}/connectors/source/{SqlServer-CDC,PostgreSQL-CDC,MongoDB-CDC,Hive,Paimon,FtpFile,SftpFile}.md`,
`docs/{en,zh}/connectors/sink/{Hive,StarRocks,Paimon,Jdbc,FtpFile,HdfsFile}.md`,
`docs/{en,zh}/transforms/dynamic-compile.md`.
## Duplicate check
Searched all PRs created in the last 7 days (`gh search prs --repo
apache/seatunnel --created ">=2026-09-26"`), specifically:
- #12609 (transforms + configuration docs), #12602 (formats +
Clickhouse/Doris/SelectDB-Cloud/Sls/Typesense/Kafka docs), #12570 (connector
option docs), #12514 (CosFile table format), #12611 (RAG embedding examples),
#12614 (CI docs build) — none of them touch the files/fixed in this PR. The
overlapping areas (transforms option tables except dynamic-compile, CDCD
formats docs, the five sinks listed above) were deliberately avoided.
## Verification
- Every documented option name/default/type was checked against the `Option`
definitions in the corresponding connector module (paths in the list above).
- Re-ran a relative-link checker over `docs/`: all modified links resolve;
no new broken links (4 pre-existing false positives inside Scala code blocks in
`dynamic-compile.md` are unchanged).
- Markdown table integrity check on all 28 modified files: no malformed rows.
- No Java code was changed, so no build/test run was needed; docs are
rendered by the separate seatunnel-website build, which is not part of this
repository's CI.
--
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]
