Allow line terminators and line comments inside character classes?

[rauschma]

Large character classes would profit from line terminators and line comments being removed:

[
  A-Z a-z 0-9 _
  .  # needed for (e.g.): Array.prototype.push
  \[ \]  # needed if the name is a symbol
]+

[slevithan]

Agreed!

To add context for other readers, note that there are multiple ways that the "free spacing and line comments" flag (aka extended mode) applies to character classes in various regex flavors:

• Perl, PCRE, .NET, Python, Ruby, and Oniguruma do not apply extended mode to character classes at all.
• Perl and PCRE, when using flag xx instead of x, ignore space and tab characters (only) within character classes.
• Java supports full free spacing and line comments within character classes when using extended mode.

In other words, @rauschma and I are both arguing that ECMAScript's flag x should follow Java's behavior. IMO Java's behavior is the most useful, consistent, and intuitive.

For even more context, one of the reasons most flavors do not follow Java's behavior is because they lack the concept of empty character classes. In most flavors, an unescaped leading ] within a character class is taken as a literal ] rather than ending the class. But ECMAScript already allows empty classes like /[]/, so it's not problem for /[ ]/x to mean the same thing. Of course, allowing line comments within a character class would also mean that /[#]/x should not be a valid/complete regex.

Why Perl and PCRE chose for flag xx to only treat space and tab as insignificant inside of character classes is unknown to me, but that does not seem desirable to follow in ES. Even if ES does not give # special meaning within character classes with flag x, it would be nice to at least treat any whitespace matched by \s as insignificant, both inside and outside of character classes.