Why does echo "a']" | grep -E "\'" not match?

Question

Based on this question on SO, I have discovered that:

echo "a']" | grep -E "\a"

Matches OK.

echo "a']" | grep -E "\]"

Matches OK.

But:

echo "a']" | grep -E "\'"

Does not match. I cannot find any documentation to describe how exactly grep -E handles escaping a single quote. I am using GNU grep 2.16 on Ubuntu 14.04.

NB:

echo "a']" | grep -E "'"

Matches OK, but I am curious as to how grep -E is interpreting \'.

Update:

I've just tried this on Cygwin 2.6.1 and grep 2.27 and I can reproduce. Using grep --colour -Eo perhaps gives a clearer OK/Fail result?

Answer 1

It's not just GNU grep, with GNU sed and GNU awk: echo "a']" | sed "s/\'/foo/" -> a']foo and echo "a']" | awk "{sub(/\'/,\"foo\")}1" -> a']foo. When you run awk in POSIX mode it does not do that but instead exhibits the expected behavior: echo "a']" | awk --posix "{sub(/\'/,\"foo\")}1" -> afoo].

The issue is that there is a GNU-ism that \' means the same as $, see http://www.regular-expressions.info/gnu.html:

Additional GNU Extensions
....
The anchor \` (backtick) matches at the very start of the subject string,
while \' (single quote) matches at the very end. 

If anyone has a use-case where \' is different from $ I'd love to hear it.