stevedlawrence commented on PR #901:
URL: https://github.com/apache/daffodil/pull/901#issuecomment-1382229205
I think I've potentially tracked down part of the issue. In
`elementParseAndUnspecifiedLengthGenerateCode`, we have this logic for complex
types:
```
context.elementChildren.foreach { child =>
cgState.addFieldDeclaration(context, child) // struct member for
child
cgState.addComputations(child) // offset, ERD computations
cgState.addPerChildStatements(child) // initERD, parseSelf,
unparseSelf for child
}
```
So this is iterating over each child of the complex type. But that only
includes actual element children, and so it skips over any other children (like
the sequence) that are part of the content of the complex type.
I'm not exactly sure how to fix this since I'm not too familiar with the
code generator, but it feels like in this complex case, you instead want do
something like `Runtime2CodeGenerator.generateCode(elementContentGram,
cgState)` to recurse into the content gram of the complex element. This should
recursively walk into the children to generate their individual parse/unparse
code instead of doing it all in this loop.
Note that if you do this, then I think you'll actually start to walk into
AlignmentFill grammars. Then you won't even need to check for
`isKnownToBeAligned` because Daffodil has already done it and optimized out any
AlignmentFill grammars where alignment is known. So if if your code generator
ever walks to this grammar then you just generate the right `parse_align` and
`unparse_align` string.
--
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]
