note LanX Let me try to write a <u>deeper conceptional answer</u>, because the handling of /e irritated me for long and I need to pin down the source of this irritation myself.<P> <h4>The short answer</h4><P> The RHS of <c>s/LHS/RHS/</c> undergoes a [http://perldoc.perl.org/perlop.html#Gory-details-of-parsing-quoted-constructs|variable interpolation] (like doublequotes) if and only if there are no <c>/e</c> modifiers used.<P> <table border=1><P> <tr> <th> modifiers </th> <th>substition</th> <th> equivalent </th> </tr> <tr> <td> no <c>e</c> </td> <td> VAR =~ s/LHS/RHS/ </td> <td> VAR =~ m/LHS/; MATCH = "RHS" </td> </tr> <tr> <td> one <c>e</c> </td> <td> VAR =~ s/LHS/RHS/e </td> <td> VAR =~ m/LHS/; MATCH = eval 'RHS' </td> </tr> <tr> <td> two <c>e</c> </td> <td> VAR =~ s/LHS/RHS/ee </td> <td> VAR =~ m/LHS/; MATCH = eval eval 'RHS' </td> </tr> <tr> <td> more <c>e</c> </td> <td colspan=2> Don't! Only for brainfuck and obfuscation, better call function </td> </tr><P> </table><P> <small>MATCH meaning the substring from VAR matching the LHS-pattern.</small><P> For completeness nota bene ...<P> <table border=1> <tr> <th> concept </th> <th>code</th> <th> equivalent </th> </tr> <tr> <td> disabling interpolation </td> <td> VAR =~ s'LHS'RHS' </td> <td> VAR =~ m/LHS/; MATCH = 'RHS' </td> </tr> <tr> <td> string interpolation </td> <td> $a = "$b - $c" </td> <td> $a = $b . ' - ' . $c </td> </tr> <tr> <td> variable evaluation </td> <td> $a = $b </td> <td> $a = eval '$b' </td> </tr> </table><P> <h4> The long answer ...</h4><P> <h5>Problem </h5><P> At first it's confusing that there is no difference between <c>s/(pattern)/$1/</c> and <c>s/(pattern)/$1/e</c><P> example: <c> DB<1> $_='abc'; s/a(b)c/$1/; print $_ b DB<2> $_='abc'; s/a(b)c/$1/e; print $_ b </c><P> <h5>Explanation</h5><P> <u>Line 1 w/o e-modifier but with interpolation:</u> <P> Variables in RHS undergo an interpolation like the result in double-quotes <c>"RHS"</c>!<P> the following code is equivalent <P> <c> DB<1> $_='abc'; s/a(b)c/$1/; print $_ b DB<3> $_='abc'; m/a(b)c/; $_ = "$1" ; print $_ b </c><P> <u>Line 2: with e-modifier but w/o interpolation:</u><P> RHS is evaluated without variable interpolation¹ like in simple quotes <c>eval 'RHS'</c><P> <c> DB<2> $_='abc'; s/a(b)c/$1/e; print $_ b DB<4> $_='abc'; m/a(b)c/; $_= eval '$1'; print $_ b </c><P> And <c>$_= eval '$1'</c> just means <c>$_= $1</c><P> <h5>I hope now double evals <c>/ee</c> can be better understood</h5><P> Like [ikegami] stated correctly those <c> s/LHS/RHS/ee;</c> and <c> s/LHS/eval RHS/e;</c> are equivalent <P> Those double <c>/ee</c> are used if you need a matchvariables (a substring from LHS) to be interpolated within an eval, b/c variable interpolation is missing!<P> Any code from LHS will be treated as uninterpolated string, as in these equivalent expamples<P> <c> DB<5> $_='warn 666'; s/(.*)/$1/e; print $_ warn 666 DB<6> $_='warn 666'; m/(.*)/; $_= eval '$1'; print $_ warn 666 </c><P> So if I need code-evaluation from parts of LHS I need a second <c>/e</c><P> <c> DB<7> $_='warn 666'; s/(.*)/$1/ee; print $_ 666 at (eval 171) .... # skipped error message 1 # $_ = return code from warn DB<8> $_='warn 666'; m/(.*)/; $_= eval eval '$1'; print $_ 666 at (eval 175). ... # skipped error message 1 # $_ = return code from warn </c><P> <h4>Resumee</h4><P> I think this explanation could be better worded ... <P> It already helped me a lot to clarify whats happening and I hope it helps you too or anyone else who tries to explain it better. =)<P>  <div class="pmsig"><div class="pmsig-708738"> <p>Cheers Rolf </div></div><P> 1) As a side note:<P> Variable interpolation is a powerful tool which allows <u>some (not all)</u> "evaluation like effects" like interpolation of hash-variables. <P> <c> DB<134> $hash{b}=666; DB<135> $_='abc'; s/a(b)c/$hash{$1}/; print $_ 666 DB<136> $_='abc'; m/a(b)c/; print "$hash{$1}"; 666 </c><P> 1006930 1006930