Scott,
The question was restricted to just the ampersand for the sake of
clarity.
The solution will of course accommodate other entities.
Since there are examples out there of regexes that are generic and
operate on any string beginning with an ampersand, I thought perhaps one
that applies to a sed might find it's way here.
Unfortunately, I'm not at liberty to constrain the source(s) from which
the xml originates to pass better data.
The problem has only arisen since a move to industry standard parsing
move old legacy methods.
Peter
-----Original Message-----
From: rpg400-l-bounces@xxxxxxxxxxxx
[mailto:rpg400-l-bounces@xxxxxxxxxxxx] On Behalf Of Scott Klement
Sent: Thursday, 13 August 2009 2:20 a.m.
To: RPG programming on the IBM i / System i
Subject: Re: sed to translate
Hi Peter,
Peter Connell wrote:
sed 's/\&[^amp]/\&/g' test1.xml
I agree with what Adam said. This code says "translate anything
starting with & to &, except &a, &m or &p, which should be left
alone".
So what will you do if someone passes another character entity besides
&? Suppose they send you < for example, you'll wind up with
&lt; which is wrong. You don't really want to translate & only when
it's not & (which is what you seem to think). What you really want
to do is translate & when it's not part of a character entity -- which
is extremely hard to do!
Well-formed XML escapes the & character for a good reason. Whomever is
coding XML with unescaped & characters needs to be stopped. No XML
reader should accept XML that's not well-formed.
midrange.com
