Re: find + sh + grep

[Eric Blake]

[re-adding the list, so that others may learn from this conversation]

On 10/25/2011 10:08 AM, Kirk Korver wrote:
> Eric,
>
> I wanted to thank you again for all of your help. I have one additional
> question. I will understand if you are too busy to respond. It is not
> blocking me, it is just for my personal education. You are more
> knowledgeable than anyone else I know.

/me blushes
Not quite true - part of becoming an "expert" is realizing that there is 
almost always someone out there better than you :)

> Here is what I type, and the result.
>
>
>
> grep '$Header:.*$' NoEnd.h
>
> $Header:

So my initial analysis wasn't quite right.  According to POSIX,

http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html

A <dollar-sign> ( '$' ) shall be an anchor when used as the last 
character of an entire BRE. The implementation may treat a <dollar-sign> 
as an anchor when used as the last character of a subexpression.

Therefore, I was wrong in stating that you have to use \$; a 
POSIX-compliant regex engine should correctly recognize that '$Header' 
uses $ where it cannot be an anchor, and thus treat it as a literal 
character without needing the \$.  [Now, how many regex implementations 
actually implement this part of POSIX, and how many get it wrong?]  In 
fact, a strict reading of POSIX makes it sound like \$ is undefined if $ 
would not otherwise be an anchor!

> So now on to my (mis)understandings
>
>
>
> 1) If I had typed    grep '\$Header:.*\$' NoEnd.h    I would not have a
> match. The single quote tells the shell to not change the contents, and the
> \$ is the dollar sign.

If $ can be an anchor, then \$ is the proper way to match a literal '$'. 
 At least GNU sed and grep treat \$ as a literal '$' everywhere, 
whether or not the $ could match an anchor.

> 2) In what I did type, the dollar sign is the 'end of line'character
>
> 3) I typed, find a line where there is an end of line, followed by Header:
> This is not my intent

Back to that pesky POSIX wording - if $ is not at the end of the regex 
or a subexpression of the regex, then it cannot be an anchor, therefore 
it does not match the end of line and instead matches a literal '$'.

> 4) I also tried     grep   "$Header:.*$" NoEnd.h    and got 3 matches. I
> then realized that the    $Header   was being interpreted as the environment
> variable  Header  which is currently not set, so this becomes      grep
> ':.*'   which matches all lines with a colon in them. I am not sure what the
> $  all by itself means. Short lesson there, understand better the difference
> between the single quote, and the double quote, and then type what I mean. J

Yes, in shell expansion, "$Header" is much different from 
'$Header'/"\$Header", which is in turn different from '\$Header'.

Also, in shell, "$" produces $, rather than a variable expansion, since 
there was no variable name to expand, but it's risky enough that you 
should generally escape that particular $.

> Now my question, I do not understand why there is a match, in what I
> initially typed. I am missing something about the regular expression. I
> believe that
>
> $Header:.*$
>
> means
>
> [end of line]Header:[any character][any number of times][end of line], which
> should not yield a match.

It actually means literal $, literal Header:, any number of characters, 
and end of line.

> Can you shed some light?

So my initial analysis wasn't quite right - but I still stand by the 
conclusion that you had a bad regex.  It was the combination of double 
shell expansion, inside "", that ate enough levels of \ that you ended 
up expanding $Header instead of searching for a literal $, so you were 
using a different regex than you had planned (:.*$ instead of $Header:.*$).

At any rate, thanks for forcing me to re-read POSIX and add better 
information to this thread.

> P.S. with an 801 number, do you live in SLC?

Yep.

--
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org