zhangshenghang opened a new pull request, #13302:
URL: https://github.com/apache/seatunnel/pull/13302
Yes, I am contributing to Apache SeaTunnel.
## What documentation issues were found
While comparing the docs on `upstream/dev` against the actual connector
source code, I found the following real inaccuracies (each one verified against
the Java sources before editing):
1. **Kingbase source/sink (en + zh)**: the data type mapping table claimed
every type outside the short listed set is `Not supported yet`.
`KingbaseTypeConverter extends PostgresTypeConverter`
(`KingbaseTypeConverter.java:44`) and additionally converts many more types:
`TINYINT`, `MONEY`, `BLOB`, `CLOB`, MySQL-compatible types (`DATETIME`,
binary/text types), Oracle-compatible types (`NUMBER`, `VARCHAR2`, ...) and SQL
Server-compatible types (`DATETIME2`, `SMALLDATETIME`, `DATETIMEOFFSET`).
2. **PostgreSql sink (en + zh)**: the combined row `_INT2/_INT4 →
ARRAY<INT>` was wrong. Code maps `_int2` → `ARRAY<SMALLINT>`
(`PostgresTypeConverter.java:156-157`), only `_int4` → `ARRAY<INT>`. The source
doc already had the correct split rows. The table also omitted `TIMESTAMPTZ(s)
→ TIMESTAMP_TZ(s)` (`PostgresTypeConverter.java:284-292`, reconvert `488-500`),
so it fell under the catch-all "not supported yet" row, contradicting the
source doc.
3. **Mysql source (en + zh)**: `partition_column` was documented as "only
support numeric type", but `ChunkSplitter.isSupportSplitColumn`
(`ChunkSplitter.java:466-483`) also accepts `STRING` and `DATE` split columns;
the doc even contradicted itself in the "Parallel Reader" section ("Supported
split data type"). Also fixed the nonexistent type name `VARBINAR` →
`VARBINARY` in the type mapping table.
4. **Vertica source (en + zh)**: same `VARBINAR` typo; the sink doc and
`VerticaTypeMapper.java:161` spell it `VARBINARY`.
5. **Cloudberry sink (en + zh)**: the "timer flush" feature checkbox was
unchecked, but timer flush (`batch_interval_ms`, `JdbcSinkOptions.java:87-94`)
is a dialect-independent JDBC sink option and the PostgreSQL sink doc marks it
supported; the Cloudberry doc itself states it uses the same options as
PostgreSQL.
6. **PostgreSQL source (en + zh)**: the PostgreSQL-only option
`numeric_special_value_as_null` (`JdbcSourceOptions.java:189-194`, default
`false`) was missing from the options table.
7. **Edge Agent download (en + zh)**: the prerequisite said "Java 8 or 11",
but the project builds with `maven.compiler.release=11` (`pom.xml:81-90`), so a
Java 8 JRE cannot run it, and `edge-agent/quick-start.md` and
`getting-started/locally/deployment.md` both require Java 11 or 17.
8. **Zeta quick start (en + zh)**: the sample console log showed `output
rowType: new_name<STRING>, age<INT>` with values in that order, but the
example's `field_mapper = { age = age, name = new_name }` produces output
columns in mapping order (`FieldMapperTransform.java:104-122`;
`docs/en/transforms/field-mapper.md` documents this). Users comparing their
first job's output against this log would see a different order.
9. **zh-only alignment**: `zh/connectors/source/GoogleBigtable.md` used an
H3 `### Schema 映射` heading where the English doc uses H2;
`zh/connectors/sink/Aerospike.md` misspelled the project name as "Seatunnel
Zeta".
## What was changed
All of the above were corrected so the docs match the code. No config option
names were renamed, no defaults invented — every new/changed value was taken
from the cited source locations. Both the English and Chinese versions of each
affected page were updated in the same way.
## Which areas/files were updated
20 files:
`docs/{en,zh}/connectors/{source,sink}/{Kingbase,PostgreSql,Mysql,PostgreSQL,Vertica,Cloudberry,GoogleBigtable,Aerospike}.md`,
`docs/{en,zh}/edge-agent/download.md`,
`docs/{en,zh}/getting-started/locally/quick-start-seatunnel-engine.md`.
## Were English and Chinese docs both checked?
Yes. Every fix was applied to both language versions where both exist; the
two zh-only divergences (GoogleBigtable heading level, Aerospike spelling) were
also fixed to match the English docs.
## Duplicate PR check (last 7 days)
I checked all PRs opened against `apache/seatunnel` in the last 7 days (both
mine and other contributors'), including #13289, #13284, #13282, #13274,
#13258, #13234, #13292 and #13270. None of them touch the
Kingbase/PostgreSql-sink/Mysql-source/Vertica/Cloudberry/PostgreSQL-source/edge-agent/quick-start
issues fixed here; file connectors, Zeta engine docs,
Doris/StarRocks/TDengine/Milvus, Clickhouse/Elasticsearch/Paimon/Redis, the CDC
connectors and the llm/embedding/calcite/dynamic-compile/text-chunk transforms
they cover were deliberately excluded from my analysis. My older open PR #13193
rewrites the `partition_column` row of
`docs/en/connectors/source/PostgreSQL.md`; this PR deliberately does NOT touch
that row (it only adds the missing `numeric_special_value_as_null` row further
down the same table), so both PRs can merge cleanly.
## How the change was verified
- Every doc claim was checked against the Java sources (file + line cited
above) before editing.
- Re-ran a relative-link/anchor checker over all 1114 doc files: no broken
links or anchors were introduced (the new `#parallel-reader` / `#并行读取器` anchors
resolve to existing headings).
- Validated markdown table column consistency for all 20 edited files: 0
issues.
- `./mvnw spotless:apply` / full build were not run because this PR contains
only `.md` changes under `docs/`; no Java resources, code, or build files were
modified.
--
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]