Here is a short function you can use to diff two arrays. It implements the LCS algorithm:
function computeDiff($from, $to)
{
$diffValues = array();
$diffMask = array();
$dm = array();
$n1 = count($from);
$n2 = count($to);
for ($j = -1; $j < $n2; $j++) $dm[-1][$j] = 0;
for ($i = -1; $i < $n1; $i++) $dm[$i][-1] = 0;
for ($i = 0; $i < $n1; $i++)
{
for ($j = 0; $j < $n2; $j++)
{
if ($from[$i] == $to[$j])
{
$ad = $dm[$i - 1][$j - 1];
$dm[$i][$j] = $ad + 1;
}
else
{
$a1 = $dm[$i - 1][$j];
$a2 = $dm[$i][$j - 1];
$dm[$i][$j] = max($a1, $a2);
}
}
}
$i = $n1 - 1;
$j = $n2 - 1;
while (($i > -1) || ($j > -1))
{
if ($j > -1)
{
if ($dm[$i][$j - 1] == $dm[$i][$j])
{
$diffValues[] = $to[$j];
$diffMask[] = 1;
$j--;
continue;
}
}
if ($i > -1)
{
if ($dm[$i - 1][$j] == $dm[$i][$j])
{
$diffValues[] = $from[$i];
$diffMask[] = -1;
$i--;
continue;
}
}
{
$diffValues[] = $from[$i];
$diffMask[] = 0;
$i--;
$j--;
}
}
$diffValues = array_reverse($diffValues);
$diffMask = array_reverse($diffMask);
return array('values' => $diffValues, 'mask' => $diffMask);
}
It generates two arrays:
- values array: a list of elements as they appear in the diff.
- mask array: contains numbers. 0: unchanged, -1: removed, 1: added.
If you populate an array with characters, it can be used to compute inline difference. Now just a single step to highlight the differences:
function diffline($line1, $line2)
{
$diff = computeDiff(str_split($line1), str_split($line2));
$diffval = $diff['values'];
$diffmask = $diff['mask'];
$n = count($diffval);
$pmc = 0;
$result = '';
for ($i = 0; $i < $n; $i++)
{
$mc = $diffmask[$i];
if ($mc != $pmc)
{
switch ($pmc)
{
case -1: $result .= '</del>'; break;
case 1: $result .= '</ins>'; break;
}
switch ($mc)
{
case -1: $result .= '<del>'; break;
case 1: $result .= '<ins>'; break;
}
}
$result .= $diffval[$i];
$pmc = $mc;
}
switch ($pmc)
{
case -1: $result .= '</del>'; break;
case 1: $result .= '</ins>'; break;
}
return $result;
}
Eg.:
echo diffline('StackOverflow', 'ServerFault')
Will output:
S<del>tackO</del><ins>er</ins>ver<del>f</del><ins>Fau</ins>l<del>ow</del><ins>t</ins>
StackOerverfFaulowt
Additional notes:
- The diff matrix requires (m+1)*(n+1) elements. So you can run into out of memory errors if you try to diff long sequences. In this case diff larger chunks (eg. lines) first, then diff their contents in a second pass.
- The algorithm can be improved if you trim the matching elements from the beginning and the end, then run the algorithm on the differing middle only. A latter (more bloated) version contains these modifications too.