hutiefang76 commented on PR #12443:
URL: https://github.com/apache/seatunnel/pull/12443#issuecomment-5842978107

   Thanks for the careful follow-up. I pushed `8ff338252` with the en/zh REST 
notes and an additional reserved-character regression, and changed the title to 
`[Fix][Config]` to reflect the shared code.
   
   1. The guarded production `ConfigFactory.parseMap` sites in this PR are 
`ConfigBuilder.of(Map)` (REST/raw map), the `ConfigBuilder` adapter path, 
`ConfigShadeUtils.decryptConfig`, and `ReadonlyConfig.toConfig`. I also checked 
the remaining production calls: `MetadataProviderManager` still parses provider 
properties and reparses unwrapped job/connector maps, and `AssertSink` parses 
its own rule map. In particular, metadata-enabled datasource resolution can 
still hit this class of failure. I have stated that limit in the PR 
description; I do not want to imply that every map-to-config caller is covered 
here. The metadata path needs a targeted follow-up with its own regression.
   2. Agreed on F3. Successfully parsed keys keep the existing parser behavior, 
including `->` nesting, removal of syntactic quotes, and trimming of 
surrounding unquoted whitespace. Only `BadPath` keys are quoted as literal 
segments. The PR description and both REST docs now say this explicitly.
   3. `ConfigBuilderTest#testConfigAdapterPreservesInvalidPathKeys` covers a 
nested regex key and an escaped quote/backslash key through the adapter. 
`ConfigBuilderTest#testInvalidPathKeysSurviveConfigShadeRoundTrip` checks 
nested keys after `ConfigShadeUtils.decryptConfig`, including `^t_nova_.*$` 
(`^`, `.`, `*`, `$`), `a:b` (`:`), `a"b\c`, and `${FOO}`. `RestUtilTest` covers 
single/batch JSON conversion, and 
`RestApiSubmitJobConfigShadeDecryptTest#testSubmitJobWithRegexFieldInJson` 
exercises the nested regex key through real HTTP submission and decrypt.
   4. The English and Chinese `docs/*/engines/zeta/rest-api-v2.md` notes now 
describe literal reserved characters and the continuing `->` separator behavior.
   
   The focused `ConfigBuilderTest` run passed (9/9) with JDK 17. A fresh 
full-repository verify stopped while resolving an unrelated DB2 connector 
dependency; I have recorded that limitation in the PR description. I'll keep 
checking the new CI run rather than treating the earlier failed run as green.
   


-- 
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