Re: grep

[Stepan Kasal]

Hello,

On Mon, Sep 22, 2003 at 02:21:41PM -0400, Gerald S Stoller wrote:
> FreeBSD  4.3-RELEASE FreeBSD 4.3-RELEASE #0: Sat Apr 21 10:54:49 GMT 2001
> address@hidden:/usr/src/sys/compile/GENERIC  i386

unfortunately, I cannot a FreeBSD system at hand, so I cannot tell which
version of GNU grep this implies but I'll suppose you have 2.5 or 2.5.1,
similarily to my system:
        $ grep --version
        grep (GNU grep) 2.5.1

> the  grep -w  line output

The man page says exactly that -w means that the match may not be preceded
nor followed by a word constituent character.  The behaviour follows
this specification:
> 0-$ echo  "fgh( hkjh\nfgh(hkjh\nfgh((hkjh"  |  grep -w  "fgh("
> fgh( hkjh
> fgh((hkjh

Of course, typical usage of -w is that the pattern is a word and you
search exactly for that word.  If the pattern is not a word, things
look a bit strange, but they are correctly documented.

I think that when you need some finer things with "words", it's better
to make use of \<, \>, \b and \B.

>       In addition, in the  grep  command's pattern string, the  man  page
> (in  FreeBSD ) says that the characters '?', '+', and '*' are
> naturally metacharacters.  However, in practice I found that '+'
> must have a preceding backlash to be viewed as a metacharacter.

The manpage states clearly that it documents extended regular
expressions, the differences for "basic" regular expressions
are listed below.

> In  egrep  use, though, I find how it views "+" & "\+" to be

BTW: according to POSIX, "egrep" is obsolete; perhaps you could
acquire a habit of using "grep -E" instead.

Hope this helps,
        Stepan Kasal