zhangshenghang opened a new pull request, #13380:
URL: https://github.com/apache/seatunnel/pull/13380

   Yes, I am contributing to **Apache SeaTunnel**.
   
   ## What documentation issues were found
   
   While comparing the connector/transform documentation against the actual 
Java option definitions (`Options.key(...)`, `optionRule()` rules, enums, 
validators and runtime code) on the current `dev` branch, the following 
verified inaccuracies were found:
   
   1. **Clickhouse source** (`en`/`zh`): the outer-level `split.size` option 
(`ClickhouseSourceOptions.CLICKHOUSE_SPLIT_SIZE`, default `Integer.MAX_VALUE`, 
min `1`) was not documented at all; only the inner `split_size` of `table_list` 
entries was.
   2. **Redis source** (`en`/`zh`): docs claimed `single_field_name` is only 
required for primitive value types (`string`/`list`/`set`/`zset`). In fact 
`RedisSourceFactory` declares `.conditional(READ_KEY_ENABLED, true, 
SINGLE_FIELD_NAME)`, so the option rule requires it whenever `read_key_enabled 
= true`, regardless of `data_type` (values that parse as JSON objects simply do 
not use it at runtime). Updated the table row and three prose sections.
   3. **Elasticsearch sink** (`en`/`zh`): `auth_type` type cell said `string`, 
code is `enumType(AuthTypeEnum.class)`; the description also omitted the 
supported `api_key_encoded` value. Section header in the same file already said 
`[enum]`.
   4. **Redis source + sink** (`en`/`zh`): `mode`, `data_type`, `format`, 
`hash_key_parse_mode` are enum-typed options in code (`.enumType(...)`) but 
were listed as `string` in the option tables.
   5. **Pulsar source** (`en`): section header `### cursor.stop.mode [String]` 
contradicted the code (`.enumType(StopMode.class)`), the en option table and 
the zh doc. Changed to `[Enum]`.
   6. **RocketMQ source** (`en`/`zh`): the `tables_configs` item key list 
omitted `field.delimiter`, which `RocketMqSourceConfig.buildDeserialization` 
actually reads from each item config.
   7. **Maxcompute source** (`en`/`zh`): `project` was marked required (`yes`), 
but `MaxcomputeSourceFactory.optionRule()` puts it under `.optional(...)` (each 
`table_list` entry can override it). Only the sink factory requires it.
   8. **Maxcompute sink** (`en`/`zh`): removed the stale "Only work on 
multi-table mode at now" restriction on `save_mode_create_template` 
(`MaxcomputeSink.getSaveModeHandler()` applies it unconditionally), and 
documented that `overwrite` is deprecated and mapped to `data_save_mode = 
DROP_DATA` with a warning (as the code does).
   9. **IoTDB sink** (`en`/`zh`): removed an unsupported claim that the sink 
"picks an active node and falls over to the others on failure" — the code 
builds a single IoTDB `Session` and retries it; there is no node failover.
   10. **IoTDBv2 sink** (`en`/`zh`): `batch_size` default `1024` is declared 
but never applied when the option is omitted (`SinkConfig` only sets it 
`isPresent()`, flush guard is `getBatchSize() > 0`) — added the same "(see 
note)" caveat the v1 doc already has.
   11. **HugeGraph sink** (`en`/`zh`): the `mappings` table documented every 
supported key except `sourceTable` (`MappingConfig.sourceTable`), a functional 
multi-table binding key whose all-or-nothing misuse fails the job at config 
time.
   12. **Neo4j sink** (`en`/`zh`): `queryParamPosition` type cell said 
`Object`; code is `.mapType()` (`Map<String, String>`).
   13. **LLM transform** (`en`/`zh`): `model_provider` enumeration omitted 
`QIANFAN`, which is part of `ModelProvider` and accepted by 
`LLMTransformFactory.optionRule()`.
   14. **Replace transform** (`en`/`zh`): removed a wrong `[string]` type 
annotation from the `### common options` heading (sibling transform docs have 
no such annotation); fixed the empty `replace_fields` default cell 
(`noDefaultValue()` → `-`) and an untranslated `yes` in the zh table.
   15. **Zeta REST API v2 doc** (`zh`): the OpenAPI spec link used a broken 
`@site/docs/engines/zeta/openapi-v2.yaml` path pointing to a location that does 
not exist (the file only existed under `docs/en`). Added 
`docs/zh/engines/zeta/openapi-v2.yaml` (byte-identical copy of the en spec) and 
changed the link to a normal relative `openapi-v2.yaml`, matching the en page.
   
   ## What was changed / which areas were updated
   
   29 files: `docs/en` + `docs/zh` for connector source/sink docs (Clickhouse, 
Redis, Elasticsearch, Pulsar, RocketMQ, Maxcompute, IoTDB, IoTDBv2, HugeGraph, 
Neo4j), transform docs (llm, replace), and Zeta REST API v2 docs (+ new 
`openapi-v2.yaml` for zh). Every row-level change is backed by the cited code 
location; no formatting-only or cosmetic churn.
   
   ## Were English and Chinese docs both checked?
   
   Yes. Every fix was applied to both the `en` and `zh` versions where the same 
content exists (the Pulsar section-header issue existed only in `en` — the zh 
header already said `[Enum]`; the zh-only `@site` link issue has no `en` 
counterpart because the `en` link was already correct).
   
   ## How duplicate PRs from the last 7 days were checked
   
   Before implementing and again before opening this PR, I listed all PRs 
created in the last 7 days (`gh pr list`, author `zhangshenghang` plus 
repo-wide `docs in:title` searches) and inspected the changed-file lists of 
#13327, #13302, #13289, #13284, #13282, #13274. None of them covers the 
connectors/transforms fixed here: #13327 touches CDC/Http/Kafka/other 
transforms, #13302 touches JDBC-family docs and getting-started, #13289 touches 
file connectors, #13284 touches FtpFile/LocalFile/Paimon/HugeGraph (its 
HugeGraph change is only a `List<string>` → `List<String>` capitalization on 
different rows than the `sourceTable` row added here), #13282 touches Zeta 
deployment/tuning docs (its `rest-api-v2.md` edit does not touch the broken 
`@site` line fixed here), #13274 touches 
Doris/StarRocks/TDengine/Milvus/calcite/embedding/llm (its llm/embedding 
changes cover `api_key` semantics, not the `model_provider` enum list fixed 
here). Older open "Improve" PRs from August (#11704/#11713/#117
 67/#11787) are doc-improvement rewrites, not accuracy fixes verified against 
code; this PR's changes are independent of them.
   
   ## How the change was verified
   
   - Every documented claim was checked against the code before editing (option 
definitions, factory option rules, validators, enums, runtime paths — file/line 
evidence collected for each of the 15 issues).
   - Ran a repo-wide relative-link checker over `docs/en` and `docs/zh`: the 
only remaining "broken" hits are false positives from regex text inside 
backticks; the `@site` link is fixed and no new broken links were introduced.
   - Ran a table-structure checker over all edited markdown files: 0 malformed 
rows/column-count mismatches.
   - The new `docs/zh/engines/zeta/openapi-v2.yaml` is byte-identical to the 
`en` spec that CI already builds.
   - The full Docusaurus website build (CI job that clones 
`apache/seatunnel-website` and runs `npm run build`) was not run locally 
because it requires checking out and installing a separate repository; the 
local link and table checks above cover the changes' markdown validity.
   
   ## Scope
   
   - 29 documentation files (28 modified + 1 new), 261 insertions / 51 
deletions — satisfying the multi-file scope with real content changes only.
   


-- 
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]

Reply via email to