RegEx - character class containing brackets, how to escape correctly
 

Hey there,
somewhat dim question I guess, but:

I'm trying to figure out how to grep for a character class like [a-d\[\]\*]. Which, as I see it, should match any of abcd[]*, but doesn't in grep.
Also trying to match a class containing both single and double quotes.
No particular reason, just trying to get a better handle on grep and REs.

Now, I came across a statement suggesting that there is no actual escaping in greps BREs, but rather in the shell, however this post suggests there is both simultaneously, so I tried double escaping like '[a\\'b]' - which doesn't work either.
So, I am very confused as to how the escaping works and why it doesn't in these particular instances.


Also, something that deeply weirds me out: when I forget to quote a character class grep always matches capital C's and nothing else. Why on earth?

//okay, now I just noticed something that really freaks me out: when I try to match '[a\-b]' grep matches a, b and every digit. Why?

Best wishes

There is escaping both in BREs and in the shell; but BRE escaping does not use the same rules as shell escaping. In particular, BRE escaping inside of character classes is a bit... funny.

GNU Grep manual: Character Classes and Bracket Expressions:
Therefore, to match any of abcd[]*, you want [][abcd*]. To protect this from shell expansion, you should then wrap in quotes: Hard to say without an example.

Backslash doesn't escape within a character class, so it's matching a, and everything between \ and b. The exact characters this consists of depends on your locale sort order. In your default locale, digits are sorted between \ and b.

Thanks a lot!
Sorry I wasn't thorough enough - I even was on that page but at a glance assumed it would not help me.