Backslashes missing on output code

[almosnow]

I'm using uglify-js inside a node program. I've noticed that sometimes the code returned from an AST is not what I would expect. I don't know if this is a bug or the intended behavior.

How to reproduce:

Execute the following node program:

var uglify = require('uglify-js');
var js = JSON.stringify(['var a = "OK";']);
console.log('A:' + js);
var code = "JSON.parse('" + js + "');";
console.log('B:' + code);
code = uglify.parse(code).print_to_string();
console.log('C: ' + code);

Output:

A: ["var a = \"OK\";"]
B: JSON.parse('["var a = \"OK\";"]');
C: JSON.parse('["var a = "OK";"]');

As you can see, the backslashes are missing on C, this removes the 'escaping' of the quotes surrounding OK and creates an error later if one would try to evaluate the JSON.parse code.

Also, it behaves the same way when explicitly using an OutputStream object.

I'm using uglify-js v2.3.6 and node v0.10.12.

[rvanvelzen]

This is expected, because \" in a single quotes string is equal to ". See also section 7.8.4 for the specifics of handling escape sequences.

[almosnow]

Yeah, well, I don't think those should get escaped.

Check this out:

var uglify = require('uglify-js');
var code = '\'\\" NO WAY, JOSE \\"\'';
console.log('A:' + code);
code = uglify.parse(code).print_to_string();
console.log('B: ' + code);
code = '\'" NO WAY, JOSE "\'';
console.log('C:' + code);
code = uglify.parse(code).print_to_string();
console.log('D: ' + code);

As you will see A != C but B == D, and that is wrong. Sure both strings are actually 'equivalent' (and I can imagine this as one of the reasons why this could've been purposedly coded as a feature), but they are not the same; and you can see in my first example that this situation leads to code that breaks.

[mishoo]

var code = '\'\\" NO WAY, JOSE \\"\'';

After this line, your variable code will contain the following characters:

'\" NO WAY, JOSE \"'

Now you're asking UglifyJS to parse that as code. Copy/paste the above line in a file if you want and run UglifyJS on it, and you'll get the correct result:

'" NO WAY, JOSE "'

Because this thing starts with a quote ('), UglifyJS starts parsing a string. Inside the string, the sequence \" simply means a " character. No bug there.

[rvanvelzen]

Inside the string, the sequence " simply means a " character. No bug there.
This was exactly what I was just typing. Funny.

[mishoo]

@rvanvelzen great minds think alike. ;-)

[almosnow]

K, got it now, thanks both.

[evolutionxbox]

This is still an issue for Regex strings.

new Regex('\d+') will output new Regex('d+')
new RegExp('\d+') will output new RegExp('d+') Edited: 22/06/17 - typo

[rvanvelzen]

That is not an issue - '\d' === 'd'

[evolutionxbox]

That's not true.'\d' !== 'd'

'\d' will match any digit, whereas 'd' will match the character d.

See: https://regex101.com/r/vejuUa/1 and https://regex101.com/r/vejuUa/2

[rvanvelzen]

We're talking about strings now, not regexes. Just execute the snippet in your console and you'll realize.

[kzc]

$ node
> new RegExp('\d+')
/d+/
> new RegExp('\\d+')
/\d+/
> new Regex('\d+')
ReferenceError: Regex is not defined

[evolutionxbox]

Sorry @kzc that's a typo on my part. I did mean RegExp, not Regex.
And yes @rvanvelzen I do now realise...