Re: s cmd lossage

[rob@plan9.att.com]

some time ago - i'm catching up on old sam mail - quanstro@sartre.minerva.bah.com said:

	this s command lost with the latest (straight from at&t)
	version of sam.
		s:(([A-Z][a-z]*[ ]?)+)[ 	]+([0-9]+):NAME	\1\nPHONE 205 \2\n:g
					^- space tab

	what happened was that \2 was set to \1.

i've tried it here and it works as written.  let me explain what's going on,
because i think you're also seeing it work correctly but it's confusing you.

i tried this source:
	ABCD	01234
and got
	NAME	ABCD
	PHONE	205 D
which is correct.  as it says in the manual, the \digit operators on the
right side of a substitution refer to the text matched by the subexpression
beginning at the digit-th left parenthesis. here \1 would refer to the match
of
	(([A-Z][a-z]*[ ]?)+)
which would be
	ABCD
and \2 would refer to the most recent match of
	([A-Z][a-z]*[ ]?)
which is
	D
confusion comes because of the nesting -- whose meaning is defined
by the manual -- and the repetition operator (+) -- whose meaning is
not but should be clear from any thought about the implementation.
referring to the implementation is the last refuge of the writer of
incomplete documentation, but i believe the behavior is reasonable.
you wrote a near-nonsense expression and got near-nonsense
results.  i see no bug here.

now you may have some input text that shows other behavior, but
if so please interpret the answer carefully before deciding there's a bug.
i don't deny there could be one, but nested repeated regexps can be
fertile sources of confusion as well as errors.

-rob