Consider expanding parsing API

[tjvr]

As discussed in #305, we might like to have an alternative to finish() (which is undocumented), which throws an error if there are no results.

How do I tell nearley that there will be no more input, so that it can wrap up its parsing?

---

Or, perhaps this could be part of lightweight parsing API? For users who don't want streaming, instead of writing:

const parser = new Parser(myGrammar)
try {
  parser.feed(input)
} catch (err) {
  // handle SyntaxError
}
assert(parser.results.length <= 1); // if you're sure your grammar's not ambiguous
if (parser.results.length === 0) {
  // handle "Unexpected end of input"
}
var result = parser.results[0]

It'd be neat if they could write:

try {
  var result = grammar.parse(input)
} catch (err) {
 // handle any SyntaxError, including EOF
}

...perhaps with a corresponding allParses() which returns all results, rather than the only result.

---

This is quite possibly a v3 thing, since there are some other API changes I think would be good (for example, I'd like to have an "ordered choice / no ambiguity" mode).

[matthew-dean]

As to this:

try {
  var result = grammar.parse(input)
} catch (err) {
 // handle any SyntaxError, including EOF
}

It's unfortunate that parsing has to be wrapped in a try/catch at all, since that will de-optimize JavaScript JIT compiling. Ideally, the parser would never throw errors, but would return a callback (/ Promise?) with an Error object. I mean, unless the concern is that errors can be caused that aren't caused by Nearley, this will reduce JIT VM efficiency AFAIK.

[tjvr]

since that will de-optimize JavaScript JIT compiling

That certainly used to be true for V8, but it's no longer the case for newer JITs I believe.

Regardless, we still haven't done any real planning toward a v3, so I wouldn't be concerned 🙂

_{Sent with GitHawk}

[bd82]

That certainly used to be true for V8, but it's no longer the case for newer JITs I believe.

@tjvr is quite correct, See: https://www.nearform.com/blog/node-js-is-getting-a-new-v8-with-turbofan/

[matthew-dean]

@bd82 Well bless my grits. I guess this is an issue no longer.

[bd82]

@matthew-dean Yeah it allowed me to remove some optimizations to mitigate performance loss due to try/catch usage and simplify the code in one of my projects 😄