First order of business: If you read Writeup Formatting Tips you'll see that wrapping code in <code> ...... </code> tags is the way to prevent your code from getting mangled by the website's linking syntax. Following that protocol, your first regex must look like this:

s/^([^a]*)def/$1xyz/
[download]

...which can't possibly match "abcdef", because it anchors the match to the start of the string, and an 'a' character comes between the start of the string and the literal 'def', blocking the match. I guess that "works fine" because it does fail to match "abcdef".

The second regex is this:

s/^(?<!abc)def/$1xyz/
[download]

If your goal is to replace 'xyz' if it is not preceded by 'abc', that one would work, except for the anchoring to the beginning of the string. You're explicitly stating that you don't want to match 'def' if 'abc' comes before it. But by anchoring to the beginning of the string, and since assertions don't consume what they match, the only legal string would start with "def". Also, since you're not consuming the prefix (lookahead and lookbehind assertions don't consume), and since you're not capturing (there are no capturing parens), there's no need to have the $1 in the replacement.

By adding capturing parens to the beginning of the regex, with three wildcard metacharacers, it works out better. You'll find that s/^(...)(?<!abc)def/$1xyz/ rejects 'abcdef', but 'xxxdef' matches, becoming 'xxxxyz', which I think is what you want.

s/^(...)(?<!abc)def/$1xyz/;
[download]

Dave writes "...which can't possibly match "abcdef", "

Dave if you read closely I was looking NOT have a preceeding "a" in that example- I was constrasting a CHAR, vs a set of CHARS...

SO I was CLOSE , except for the anchor . THANKS to you and Kenneth- BOTH of you got my ++ vote...

use strict;
use warnings;
use YAPE::Regex::Explain;

print YAPE::Regex::Explain->new(qr/^(?<!abc)def/)->explain;
[download]

The regular expression:

(?-imsx:^(?<!abc)def)

matches as follows:
  
NODE                     EXPLANATION
----------------------------------------------------------------------
(?-imsx:                 group, but do not capture (case-sensitive)
                         (with ^ and $ matching normally) (with . not
                         matching \n) (matching whitespace and #
                         normally):
----------------------------------------------------------------------
  ^                        the beginning of the string
----------------------------------------------------------------------
  (?<!                     look behind to see if there is not:
----------------------------------------------------------------------
    abc                      'abc'
----------------------------------------------------------------------
  )                        end of look-behind
----------------------------------------------------------------------
  def                      'def'
----------------------------------------------------------------------
)                        end of grouping
----------------------------------------------------------------------
[download]

Look aheads and look behinds are zero-width. This means that your regular expression is a subset of /^def/, which is clearly not what you intend. There are a number of ways to get what you express above. Probably the most natural would be s/(?<!^abc)\Kdef/xyz/. Moving the ^ inside the negative look behind means it only matters if the a starts the string. \K means 'keep everything before this' (see Character Classes and other Special Escapes)

(note SB = SQUARE BRACKET, this EDITOR is removing them for some reason)

That's because you're not using  <code> ... </code> tags! Please use code tags! (Can also be abbreviated  <c> ... </c>) Please see How do I post a question effectively?, Writeup Formatting Tips and Markup in the Monastery.

In contrast to Re: Regex question is this one of those look (ahead|behind)s ?, produced using PPIx::Regexp::xplain because YAPE::Regex::Explain is dead; file ppixregexplain.pl needs _desc.pl, the explanation

Also interesting (but tad more pita to install) is wxPPIxregexplain.pl