Why are the syntax trees obtained by chevrotain/ecma5 and acorn different in level when executing the same code?

[alamhubb]

Is there a solution to the associativity problem in chevrotain/ecma5? If so, how is it done?

When I use chevrotain/ecma5 to execute var a = 1+2*3+4, the result is

const code = 'var a = 1+2*3+4'
const tokens = tokenize(code);
const parserInstance = new ECMAScript5Parser();
parserInstance.input = tokens;
parserInstance.orgText = code;
const res = parserInstance.Program();

When I use acorn in https://astexplorer.net/ I get the result

I also tried chevrotain/calculator and got the same result. Chevrotain's processing method is to group by token type, instead of classifying left and right nodes like acorn.

So, I don't understand how chevrotain solves the associativity problem. Can you tell me where I can find documentation on this? Or can you point me to it?

#387

I executed the chevrotain/calculator code and got the same result. It does not differentiate between left and right nodes, but uses grouping. How does this solve the TB problem?

[bd82]

Chevrotain produces a Concrete Syntax Tree which is implicitly defined by the grammar.
So of course the structure would be different from a different parser (e.g Acorn).
And could be different between different implementations of the same grammar in Chevrotain.

In regards to expressions parsing:

1. The grouped nodes are in the orders which they have been encountered, so the order information is there.
2. There is also location information included in these nodes, which is another way to re-construct the order.
3. Operator precedence is another thing which can affect the "level" / structure of the CST
   • In the Calculator example the precedence is "baked" into the grammar and is thus represented in the CST model.
   • In the ECMA5 example the precedence is not handled and the binary operators are parsed as a flat list.
   • For anything beyond a simple scenario I would recommend the flat list approach, see, note in SWIFT Parser docs

BTW for the simplest of scenarios, 2 operads, one operator.
You could use LABELS to distinguish left / right operators, these labels will change the name of the CST nodes, e.g:

• chevrotain/examples/grammars/xml/xml_parser.js

  Lines 61 to 69 in 4272462

  	$.OR([
  	{
  	ALT: () => {
  	$.CONSUME(t.CLOSE, { LABEL: "START_CLOSE" });
  	$.SUBRULE($.content);
  	$.CONSUME(t.SLASH_OPEN);
  	$.CONSUME2(t.Name, { LABEL: "END_NAME" });
  	$.CONSUME2(t.CLOSE, { LABEL: "END" });
  	},

[alamhubb]

Thank you for your answer.

I have another question. #2046

If you have time, I hope you can help answer it. I look forward to your answer. I am very sorry if it disturbs you.