Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!pasteur!ucbvax!decwrl!labrea!sri-unix!quintus!ok
From: ok@quintus.UUCP (Richard A. O'Keefe)
Newsgroups: comp.unix.questions
Subject: Re: Picking a character from a word
Message-ID: <905@cresswell.quintus.UUCP>
Date: 27 Apr 88 10:30:28 GMT
References: <13125@brl-adm.ARPA> <50995@sun.uucp>
Organization: Quintus Computer Systems, Mountain View, CA
Lines: 24

In article <50995@sun.uucp>, guy@gorodish.Sun.COM (Guy Harris) writes:
> [I wrote]
> >    Unfortunately, "expr substr" is a BSD-ism which has yet to find its way
> >    into the SVID.
> Actually, "expr substr" is an AT&T-ism that they never documented and that they
> thought better of in S3 or so and deleted.

Thanks for the information.  And thanks for the warning that 
	expr <string1> <relop> <string2>
won't work when <string1> is one of 'length', 'substr', 'index'.
However, deleting those operators did _NOT_ fix the general problem:
it doesn't work too well if <string1> = "(".

It should be noted that test(1) has similar problems: according to the SVID,
	test $string
is supposed to succeed if $string is non-empty, e.g.
	test "a"
is true, and
	test ""
is false.  But let $string be "-z", and you get an error message!  You
would expect that using "-n $string" would eliminate this ambiguity,
but it introduces another: try string="="!.

The world is still waiting for a UNIX utility like test(1) or expr(1)
which works all the time.