Skip to content

Latest commit

 

History

History
52 lines (38 loc) · 6.19 KB

EscapedChars.md

File metadata and controls

52 lines (38 loc) · 6.19 KB
Raw

List of Escaped Characters

This file contains a list of escaped characters under this proposal. Each character details the reason it is escaped.

SyntaxCharacter Proposal

This proposal is currently the primary proposal and escapes characters under the SyntaxCharacter class in the ES2015 specification. The characters included in the list are the following:

Character Why escape it?
^ So that new RegExp(RegExp.escape('^') + "a") will match "^a" rather than the ^ being treated as a negation or start of sentence control construct.
$ So that new RegExp("a" + RegExp.escape('$')) will match "a$" rather than the $ being treated as a end of sentence control construct.
\ So that new RegExp(RegExp.escape("\\")) won't throw a type error and instead match "\\", and more generally that \ won't be treated as an escape control construct.
. So that new RegExp(RegExp.escape(".")) won't be matched against single characters like "a" but instead against an actual dot ("."), and more generally that . won't be treated as an "any character" control construct.
* So that new RegExp(RegExp.escape("*")) won't throw a type error but instead match against an actual star ("*"), and more generally that * won't be treated as a "zero or more times" quantifier.
+ So that new RegExp(RegExp.escape("+")) won't throw a type error but instead match against an actual plus sign ("+"), and more generally that + won't be treated as a "one or more times" quantifier.
? So that new RegExp(RegExp.escape("?")) won't throw a type error but instead match against an actual question mark sign ("?"), and more generally that ? won't be treated as a "once or not at all" quantifier. Also that new RegExp("("+RegExp.escape("?=")+")") will match a literal question mark followed by an equals sign, instead of introducing a lookahead, and more generally that ? won't make groups become assertions or non-capturing.
( So that new RegExp(RegExp.escape("(")) won't throw a type error but instead match against an actual opening parenthesis ("("), and more generally that ( won't be treated as a "start of a capturing group" logical operator.
) So that new RegExp(RegExp.escape(")")) won't throw a type error but instead match against an actual closing parenthesis (")"), and more generally that ) won't be treated as a "end of a capturing group" logical operator.
[ So that new RegExp(RegExp.escape("[")) won't throw a type error but instead match against an actual opening bracket ("["), and more generally that [ won't be treated as a "start of a character class" construct.
] This construct is needed to allow escaping inside character classes. new RegExp("]") is perfectly valid but we want to allow new RegExp("["+RegExp.escape("]...")+"]") in which the ] needs to be taken literally (and not as the closing "end of character class" character.
{ So that new RegExp("a" + RegExp.escape("{1,2}")) will not match "aaa", and more generally that { is taken literally and not as a quantifier.
} So that new RegExp("a" + RegExp.escape("{1,2}")) will not match "aaa", and more generally that } is taken literally and not as a quantifier.
| So that | will be treated literally and new RegExp(Regxp.escape("a|b")) will produce a string that matches "a|b" instead of the | being treated as the alternative operator.

"Safe with extra escape set" Proposal.

This proposal additionally escapes - for context sensitive inside-character-class matching, hex numeric literals (0-9a-f) at the start of the string in order to avoid hitting matching groups and lookahead/lookbehind control characters.

Character Why escape it?
- This construct is needed to allow escaping inside character classes. new RegExp("-") is perfectly valid but we want to allow new RegExp("[a"+RegExp.escape("-")+"b]") in which the - needs to be taken literally (and not as a character range character).

And only at the start of strings:

Character Why escape it?
0-9 So that in new RegExp("(foo)\\1" + RegExp.escape(1)) the back reference will still treat the first group and not the 11th and the 1 will be taken literally - see this issue for more details.
0-9a-fA-F So that new RegExp("\\u004" + RegExp.escape("A"), "u") will not match the letter "J" (\u004A) but rather throw a type error (or that without the u flag, the sequence "u004A" would be matched, \u being an identity escape), or more generally that a leading hexadecimal character may not extend a preceding escape sequence - see this issue for more details.

Note that if we ever introduce named capturing groups to a subclass of the default RegExp those would also need to escape those characters.

Extended "Safe" Proposal

This proposal escapes a maximal set of characters and ensures compatibility with edge cases like passing the result to eval.

Character Why escape it?
/ So that eval("/"+RegExp.escape("/")+"/") will produce a valid regular expression. More generally so that regular expressions will be passable to eval if sent from elsewhere with /. Note that data indicates this is not a common use case.
WhiteSpace So that eval("/"+RegExp.escape("\r\n")+"/") will produce a valid regular expression. More generally so that regular expressions will be passable to eval if sent from elsewhere with /. Also improves readability of the escape output. See this issue for more details. Note that data indicates this is not a common use case.