Javascript string replace weirdness — $$$$ gets co

问题:

At work, I was encountering a problem where users of our application were receiving messages featuring an invalid unicode character (0xffff), which according to the standard, should never be mapped to a symbol.

As a quick work aound I did the following:

badStr.replace(/\uffff/g, " ");

Which works as expected, and lets the user continue using the application until we find a better solution.

However, while I was playing around with this, I randomly tried a string replacement of "$$$$" which somehow got collapsed "$$".

You can see for yourself. Try pasting the following lines in your browser url bar:

javascript: alert(String.fromCharCode(0xffff).replace(/\uffff/g, "@@@@"));

results in @@@@

but

javascript: alert(String.fromCharCode(0xffff).replace(/\uffff/g, "$$$$"));

results in $$

This actually seems to be a problem with any string replacement, with $$$$ as the string replacement.

Both:

javascript: alert(String.fromCharCode(0x1234).replace(/\u1234/g, "$$$$"));
javascript: alert("hella".replace("h", "$$$$")); 

result in the $$ collapse.

Any ideas as to why the string replacement behaves this way?

回答1:

That's because $ in the replace string has special meaning (group expansion). Have a look at this example:

alert("foo".replace(/(.*)/, "a$1b"));

That's why $$ is interpreted as $, for the case where you would need to actually replace something by $1 (literally, without group expansion):

alert("foo".replace(/(.*)/, "a$$1b"));

See also https://developer.mozilla.org/en/JavaScript/Reference/Global_Objects/String/replace#Specifying_a_string_as_a_parameter.

回答2:

The $ sign is a special character in the replacement argument to denote sub-matches from parentheses in the regex pattern ($1, $2, etc.). So to get a $ you have to "escape" it by typing $$. Which is what you did twice.

回答3:

The $ in a replace string is used to signal the use of the match groups $1, $2 etc, si to put a $ into the replace string you need to use two of them.