Iterating Cst & Performance

[cdietrich]

Hi, we need to do some validations on the cst.

thus we currently iterate the cst like

if (node.$cstNode !== undefined) {
        for (const cst of CstUtils.streamCst(node.$cstNode)) {
          if (isLeafCstNode(cst) && cst.grammarSource !== undefined) {
               ..... check the grammar source for place and validate (e.g. GrammarAst.isRuleCall && grammarSource.rule.ref.name = ....
               

or

    if (node.$cstNode !== undefined) {
      for (const cst of CstUtils.streamCst(node.$cstNode)) {
        if (cst.grammarSource !== undefined) {
          if (GrammarAST.isRuleCall(cst.grammarSource)) {
            const ruleCall = cst.grammarSource
            if ("JSON_VALUE" === ruleCall.rule.ref?.name) {

but this seems to eat a lot of performance.
do you have an idea how to speed up.

[msujew]

Unfortunately, I'm also not entirely sure how to improve on this. Storing the information in the AST is not an option in this case? I would assume that AST iteration would likely be faster.

[cdietrich]

another usecase might be more specific datatype rules e.g. the one we have for JSON where we would error if the JSON is not valid

[spoenemann]

Is the value used as a symbol name for reference resolution?

[cdietrich]

Both declarations and references

[spoenemann]

Ok I think that's a very special case for which using the CST is a valid solution.

For other cases, I propose that we add the option to use custom object types as ValueConverter results. With that, you can attach additional information to a value in your custom ValueConverter and then use that information in validation rules.

[cdietrich]

yeah. this is why i wonder if there is a way to speed up cst traversal