[SOLVED] shell scripting - Grep a character class containing '-'

Ramurd · 08-22-2014, 04:08 AM

Ok, I admit it; I don't really see the solution yet...

I have this script, among all the things it does is grep files for 'words';
The definition of a word is (as I thought in grep expression): "@[a-zA-Z0-9\.\-_]+@"

But it does not find a word like: '@foo.bar-test@'; which I thought it should detect.
The problem is that [[:alnum:][:punct:]] matches too much (eg: @foo=bar@)

I tried to use grep in various ways: grep -o -e, grep -o -E, grep -o -- etc
Also tried double-escaping the dash (it seems the only caharacter that is making issues is the dash, so focussing on that one now).

So far I've had the following commands to test it out:

Code:

echo "@foo.bar-match@" | grep "@[a-zA-Z0-9\.\-_]+@"
echo "@foo.bar-match@" | grep '@[a-zA-Z0-9\.\-_]+@' # different quotes
echo "@foo.bar-match@" | grep -e '@[a-zA-Z0-9\.\-_]+@'
echo "@foo.bar-match@" | grep -- '@[a-zA-Z0-9\.\-_]+@' # maybe the shell tries to interpret
echo "@foo.bar-match@" | grep '@[a-zA-Z0-9\.\\\-_]+@'
echo "@foo.bar-match@" | grep -e '\@[a-zA-Z0-9\.\-\_]+\@'
echo "@foo.bar-match@" | grep '@[a-zA-Z0-9\.[-]_]+@'

And a bunch more, actually, to add a bit of flavor. (e.g. combine -e and --), switch quotes, combine various ways of escaping; I'm a bit lost... why does grep not "see" this what should match...

Maybe one of the gurus immediately sees (as regexpes always give me headaches)

Ramurd · 08-22-2014, 04:14 AM

Maybe I should phrase the definition of a word here:
- A string starting and ending with '@'
-- Containing, between these two '@':
--- at least one of:
---- lowercase letter (a-z)
---- uppercase letter (A-Z)
---- number (0-9)
---- period (.)
---- dash (-)
---- underscore (_)

Ramurd · 08-22-2014, 05:08 AM

ok, this is weird...

Code:

echo "@foo.bar-match@" | grep -E '@[[:alnum:]_.\-]+@'

does recognize this string

Code:

echo "@foo.bar-match@" | grep -E '@[[:alnum:]_\.\-]+@'

as well
and

Code:

echo "@foo.bar=match@" | grep -E '@[[:alnum:]_.\-]+@'

does not match (which is intended)

Funny how the non-escaped dot does not have the special meaning 'any character'; I kinda forgot that one :-)
It seems that the location of the dash in the "character set" makes the difference... Never would've thought it would.

grail · 08-22-2014, 05:43 AM

I am not sure I understand the issue, suffice it to say, anything inside [] does not get interpreted.

Code:

echo "@foo.bar-match@" | grep -E '@[[:alnum:]_.-]+@'

The only other point I would make, is keeping the hyphen at the end also lets [] know that you are not providing a range, like [a-z]

rknichols · 08-22-2014, 11:07 AM

As stated in the grep manpage under "Character Classes and Bracket Expressions":

"Most meta-characters lose their special meaning inside bracket expressions. To include a literal ] place it first in the list. Similarly, to include a literal ^ place it anywhere but first. Finally, to include a literal - place it last."

Ramurd · 08-25-2014, 01:53 AM

I did grab the man page and tried to search the entire interwebs. While specifically looking for this issue I still overlooked that part in the manpage; However, I did read this line:

"Most meta-characters lose their special meaning inside bracket expressions"; but I could not find the specification of "most"; I guess I need new glasses or learn to read... or...

Quote:

suffice it to say, anything inside [] does not get interpreted.

And that's the thing actually; I read "most" not "anything" in the manpage, which put me on the wrong leg (if that's a proper saying)

Still thanks for the replies! it's fixed now :-) Marking thread as solved.

rknichols · 08-25-2014, 09:19 AM

Quote:

Originally Posted by Ramurd

I did grab the man page and tried to search the entire interwebs. While specifically looking for this issue I still overlooked that part in the manpage; However, I did read this line:

"Most meta-characters lose their special meaning inside bracket expressions"; but I could not find the specification of "most"; I guess I need new glasses or learn to read... or...

The rest of that paragraph details the only meta-characters that still have special meaning:

"To include a literal ] place it first in the list. Similarly, to include a literal ^ place it anywhere but first. Finally, to include a literal - place it last."

Those are the only three.