Home/ Questions/Q 7534243
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-30T05:57:39+00:00

This is not exactly a question, because in this moment I’m not able to

  • 0

This is not exactly a question, because in this moment I’m not able to arrange in my mind a precise question to ask. It is more a collection of loud voice thoughts that I want to share with the other members, in order to fix my doubts.
I’m trying to implement an autocompleter for JavaScript. DON’T waste your time saying to me that is impossible to have a full autocompleter for a not statically typed programming languate. I already know it.
What I want is a small autocompleter that if I define something like that:

var a = {
b: { first: "goofy",second: 5},
c: "minnie"};

if (a.b.second > 0)
   a.s = 0;
else
   a.t = "honey";

when I digit a. the autocompleter should show me b,c,s,t.
I spent the last month to learn how to parse things using the material provided by compiler course at stanford university http://www.stanford.edu/class/cs143/.
Now I’m a little confuse. Obviously if I had to create a compiler I would use a LALR grammar and tools like flex&bison. But autocompleter it’s a different stuff. It’s interactive. While you are typing you have not the complete code and the line that you are typing in a single momente is not a valid sentece in a programming language. If I’m typing a. a LALR parser will say that is not able to reduce the sentence to the starting production. That means that is not fully able to parse what I wrote till this moment. But in any case I have to parse the sentence in order to show what members are inside “a”.
If I have well understood LALR parsing is lazier than LL. LALR is trying to collect as much lexemes as possible before applying a reduction. So I have the impression that LALR is the right choice for a compiler but LL is better for an autocompleter (if the programming language you are going to parse has an LL grammar, obviously). Is it correct?
Sorry I know that my post is confusing but in this moment I have the same confusion in my mind and I hope someone can help me to clear up the mess I have in my brain. I’m looking for someone with some kind of experience in writing autocompleter that could give me some small and usefull hints.

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    I’d say that it will not matter if you’d use a LALR or a LL parser to achieve this. Regardless of the choice of generator, the key issue will be that the parser is robust enough to handle errors in a meaningful way.

    Since you will feed your parser incomplete input continuously, it must be able to recover from errors in a stable fashion without giving up. Typically an easy solution is that when you find an error you ignore tokens until you can reach an understandable state again. In JS I guess the best way would be to eat tokens until you get a ; or a }. From which you continue parsing again as if the error never happened. Bison actually handles this situation quite well, take a look at the error handling documentation.

    The autocompleter itself is basically a walk of the symbol table from any given point. As you correctly identifier, it’s impossible to do a complete autocompleter since the language isn’t statically typed, but you can at least know what has been statically declared at a given point.

    • 0