The space character as a punctuator in JavaScript

In chapter 7.7 (Punctuators) of the ECMAScript spec ( http://www.ecma-international.org/publications/files/ECMA-ST/ECMA-262.pdf ) the grid of punctuators appears to have a gap in row 3 of the last column. This is in fact the space character punctuator, correct?

I understand that space characters may be inserted optionally between tokens in the JavaScript code (in order to improve readability), however, I was wondering where they are actually required...

In order to find this out, I searched for space characters in the minified version of the jQuery library. These are my results:

A space is required... (see Update below)

... between a keyword and an identifier:

function x(){}
var x;
return x;
typeof x;
new X();

... between two keywords:

return false;
if(x){}else if(y){}else{}

These are the two cases that I identified. Are there any other cases?

Note: Space characters inside string literals are not regarded as punctuator tokens (obviously).

Update: As it turns out, a space character is not required in those cases. For example a keyword token and a identifier token have to be seperated by something, but t开发者_StackOverflowhat something does not have to be a space character. It could be any input element which is not a token (WhiteSpace, LineTerminator or Comment).

Also... It seems that the space character is regarded as a WhiteSpace input element, and not a token at all, which would mean that it's not a punctuator.

---

Update (2021): The spec is much clearer now, and space is definitely not in the list of punctuators. Space is whitespace, which is covered in the White Space section.

---

Answer from 2010:

I don't think that gap is meant to be a space, no, I think it's just a gap (an unfortunate one). If they really meant to be listing a space, I expect they'd use "Whitespace" as they have elsewhere in the document. But whitespace as a punctuator doesn't really make sense.

I believe spaces (and other forms of whitespace) are delimiters. The spec sort of defines them by omission rather than explicitly. The space is required between function and x because otherwise you have the token functionx, which is not of course a keyword (though it could be a name token — e.g., a variable, property, or function name).

You need delimiters around some tokens (Identifiers and ReservedWords), because that's how we recognize where those tokens begin and end — an IdentifierName starts with an IdentifierStart followed by zero or more IdentifierParts, a class which doesn't include whitespace or any of the characters used for punctuators. Other tokens (Punctuators for instance) we can recognize without delimiters. I think that's about it, and so your two rules are pretty much just two examples of the same rule: IdentifierNames must be delimited (by whitespace, by punctuators, by beginning or end of file, ...).

Somewhat off-topic, but of course not all delimiters are equal. Line-breaking delimiters are sometimes treated specially by the grammar for the horror that is "semicolon insertion".

---

Whitespaces are not required in any of these cases. You just have to write a syntax that is understandable for the parser. In other words: the machine has to know whether you're using a keyword like function or new or just defining another variable like newFunction.

Each keyword has to be delimited somehow - whitespaces are the most sensible and readable, however they can be replaced:

return/**/false;
return(false);

---

This is just a guess, but I would say that spaces aren't actually required anywhere. They are used just as one of many alternatives to generate word boundaries between keywords. This means you could just as well replace them with other characters.

If what you want is to remove the unnecessary spaces from some code I would say that spaces (white-space to be more exact, tabs will work just as well) are mandatory only where there are no other means of separating keywords and/or variable identifiers. I.e. where by removing the white-space you no longer have the same keywords and identifiers in the resulting code.

---

What follows is not exactly relevant to your needs but you may find it interesting. You can write you examples so that they no longer have those spaces. I hope none of the examples are wrong.

x=function(){} instead of function x(){}
this.x=null; instead of var x;
return(x); instead of return x;
typeof(x); instead of typeof x;
y=X(); instead of y = new X();
return(false) instead of return false
if(x){}else{if(y){}else{}} instead of if(x){}else if(y){}else{}