STRIP-JAVASCRIPT-COMMENTS
A small (464 bytes) & fast(ish) implementation of comment-removal from javascript source code.
There are MANY such implementation on the interwebz. All (that I'm aware of) fall into 2 camps:
- Directly or indirectly use a full javascript parser, often esprima together with the escodegen project. Another example is the Babel ecosytem which uses the babylon parser under the hood.
- Small direct implementation using character processing (including regular expressions)
Parser-based Implementations
[RECOMMENDED] If your app can afford a parser-based implementation (code-size-wise, dev-effort-wise), this should definitely be your choice. There is no better alternative than native parsing in order to correctly detect and remove comments from javascript code. That's because javascript comments have a syntax that can make it indistinguishable from certain regular expressions (as per below).
A full parser will not get tripped up on "comments" inside strings or regular expressions.
Non-Parser-based Implementations
The second class of comment removing algorithms are those created "on-the-fly" by developers (like me!) who need something very lightweight and single-purpose and don't want (or can't afford) to include a full parser in their distro (usually because of code size constraints).
A simple implementation (using regex) can be only a few lines long (this is a good example and there are many many many more on the intertubes) but will usually fail in some cases. These failures may be acceptable depending on your use cases (if you have some control over the code you're likely to clean, this may not be an issue). But if you need to handle edge cases, these minimal implementations may not work since they may fail silently.
My Requirements
I implemented my own version because I could not include a large code base for what was a small requirement in our app but I wanted a more complete implementation than what was out there.
I also wanted a dead-simple function that did not use any options or settings (as many of the others do).
My implementation does a single thing: removes ALL comments from javascript code. That's it.
And the minified-gzipped implementation is only 464 bytes!!!
The closest to my implementation is this one, from lehni. That version is also available from the npm repo as uncomment.
My implementation handles edge cases that are ignored by many of the others. (please let me know if there was another implementation out there that I missed that was lightweight and also handled all the edge cases below (and maybe more? :-))
Please let me know if I missed some unhandled conditions.
Edge Cases ("pathological cases")
The following are cases that are difficult to handle for non-parser based comment removers.
Strings in quotation marks
Many pure regex will consider strings as special cases but ignore some issues:
- string quotes are themselves embedded as escaped characters in the string
- many (many) algos recognize a SINGLE backslash in a string but not multiple ones
- string quotes now include 'tick' marks (older algos that may not have been updated)
Regular Expressions
All these cases are kept in the pathological cases file
Patho-case #1: [OK-BECAUSE-DETECTABLE-AND-AVOIDABLE]
This occurs when an odd number of backslashes preceeds what looks to be a comment start in a regular expression.
var pc1a = /succeeds with any-reg-exp \/*/ ;
var pc1b = /succeeds with any-reg-exp \\\/*/ ;
var pc1c = /succeeds with any-reg-exp \\\\\/*/ ;
strip-comments WILL WORK CORRECTLY in these cases as it considers sequential backslash count in determining whether or not the slash-star is escaped or not.
Patho-case #2: [UNLIKELY but possible]
This occurs when a regex is immediately followed by an "operator" (i.e. multiplication * or division /). This is legal but does not make any sense: it's not clear why someone would want to multiply or divide a regular expression.
This is likely a user error, imo. Removing the second slash or the asterisk should clear up the case.
A syntax error WILL RESULT result from my comment stripping code:
var pc2a = /fails: trailing slash of regex lost->//g ;
var pc2b = /fails: trailing slash of regex lost->/*g ; // will steal everything until next */
Patho-case #3: [UNLIKELY but possible]
This occurs when a double-slash is used in a regular expression character matching expression, such as [//]. While this is legal, it's likely an error since there is no need to use the same character twice inside the [] operator.
This is likely an error, imo, and removing the second character will correct the situation.
A syntax error WILL RESULT result from my comment stripping code:
var pc3 = /fails: remainder of regex lost after ->[//] rest-of-re / ;
Patho-case #4: [BAD-AND-UNVERIFIABLE]
This will happen if the exact sequence of character matching in a regular expression is [/*] which is valid and reasonable. Unfortunately, because I'm not using a full parser, I have no way to know I'm in a regex and so I cannot detect that sequence accurately.
WORKAROUND is to change [/*] to equivalent [*/] or [\/*]
A syntax error WILL RESULT result from my comment stripping code:
var pc4 = /fails: remainder of regex lost after ->[/*] rest-of-re/ ; // code stripped until next */
Patho-case #5: [BAD-AND-UNVERIFIABLE]
This occurs when a valid comment immediately follows a regular expression, without an intervening space, which leads my code to STRIP THE WRONG slashes (first two, including end-of-regex slash, instead of following 2 chars).
WORKAROUND is to simply include a separator character (i.e. space, or semicolon) before the comment (which most developers would consider as good form anyway [ahemm... :-)])
A syntax error WILL RESULT result from my comment stripping code:
var pc5a = /fails: trailing slash of regex lost->/// valid EOL comment following a regex
var pc5b = /fails: trailing slash of regex lost->//* valid multiline comment following a regex */
Caveat
While my implementation does NOT cover the last 2 "patho-cases," (#4 & #5) it WILL GENERATE A SYNTAX ERROR in the resulting code. This error can be caught and the suggested workarounds used (if it wasn't an error in the first place).
For me, this compromise is adequate for the use cases of the app I'm working on. Your requirements may differ.
IMPORTANT OTHER CAVEAT It is possible to embed tick-strings inside other tick strings. While the ultimate number of ticks will (must) balance out (and the code is likely to work 99.999% of the time) IT IS POSSIBLE that comments can be inserted WITHIN the ${expressions} with those nested tick marks. This COULD affect the resulting code. Again, I consider this to be a serious edge case and am not concerned with it (I think it would be VERY poor form to use comments embedded within tick-based strings)