I need help distinguish math operations and string concatenations with function calls in them

[phartenfeller]

I am trying to write a PL/SQL parser and first off I want to say that Chevrotain works great and I like to use it very much 👍🏽 .

Yesterday I stumbled into a problem where I want to distinguish string concatenations from math operations when both call a function in them. I guess it's a logical error by my code and no bug that's why I am writing this in the discussion section.

I reduced my code to the following where only the problematic part is in. This is ready to use in the Playground:
https://gist.github.com/phartenfeller/f6fadf907a9fd7b58f355d874178a685

This is the test input. Note that all these all are valid PL/SQL values, some are numbers some strings. So I want all these statements to be parsed correctly.

Problematic is the value rule.

Regarding which of the rules mathAdditionExpression and stringExpression is first, other parsing errors will occur. If mathAdditionExpression is first the string concat keyword || will fail because everything before gets parsed as a math expression. If stringExpression is the first math operations like + will fail because everything is a string.

I thought that backtracking could fix this but this seems to not be the case. I guess increasing lookahead wouldn't make sense because the first || or math operator could appear many tokens ahead.

So how can I effectively distinguish math operations and string concatenations to make all the statements parse? And is the root problem that both string and math expressions can be a function?

Thank you very much for your time 😃 .

[bd82]

So how can I effectively distinguish math operations and string concatenations to make all the statements parse? And is the root problem that both string and math expressions can be a function?

Remove the semantics from the syntactic analysis. Don't try to figure out if its a string concatenation or a number addition during parse time. Build a more general "expression" during parse time and figure out what it means in a post-parsing.
Said post-parsing steps may also perform validations to assert that correct operands types match the operators used.

[phartenfeller]

Thanks a lot, I will try that 👍🏽