search

endojs / Jessie

Tiny subset of JavaScript for ocap-safe universal mobile code
Apache License 2.0
281 stars 16 forks source link

document automatic semicolon insertion (ASI) hazard? #49

Closed dckc closed 4 years ago

dckc commented 4 years ago

What exactly is the hazard / risk?

I filed https://github.com/tgrospic/rnode-client-js/issues/16 asking @tgrospic to avoid it, but when I did so, I struggled to come up with any argument other than authority; i.e. Agoric says so (and Crockford used to say so). @tgrospic is (quite reasonably, as far as I can tell) pushing back.

erights commented 4 years ago

Given the stated rules -- never start a line with [, (,, +, *, /, -, ,, .,` -- and given high quality IDE support, which is plausibly as good as he says, I can believe that the hazard goes away. But that's a lot of work and narrow set of assumptions for questionable benefit. I'd rather invest my IDE engineering on real problems, of which we have plenty.

erights commented 4 years ago

To answer your question, "What exactly is the hazard/risk?", it is that the programmer thought they wrote a program whose parse would give it one meaning but the machine runs it as a program with a different meaning based on a different parse. Without his restrictions and IDE support there are plenty of examples caused by ASI. Given his restrictions and IDE support, I don't know if there are any examples where ASI causes such ambiguity.

Here's a tricky non-ASI example. What does the following code do as a script? As a CJS module? As an EcmaScript module? As an EcmaScript module packaged up by your favorite packer, say rollup?

let x = 8;
let y = 9;
let z = y <!--x;
console.log(z);
dckc commented 4 years ago

OK. Thanks. I recall that it simplifies writing a Jessie parser, which is probably enough motivation for this scope.