Reliable and effective custom search & replace function - preg or str replace - php

Question

In a few different guises I've asked about this "filter" on here and WPSE. I'm now taking a different approach to it, and I'd like to make it solid and reliable.
My situation:
When I create a post in my WordPress CMS, I want to run a filter which searches for certain terms and replaces them with links.
I have the terms that I want to search for in two arrays: $glossary_terms and $species_terms.
$species_terms is a list of scientific names of fishes, such as Apistogramma panduro.
$glossary_terms is a list of fishkeeping glossary terms such as abdomen, caudal-fin and Gram's Method.
There are a few nuances worth noting:
Speed is not an issue, as I will be running this filter in the background rather than when a user visits the page or whan an author submits/edits a species profile or post.
Some of the post content being filtered may contain HTML with these terms in, like <img src="image.jpg" title="Apistogramma panduro male" />. Obviously these shouldn't be replaced.
Species are often referred to with an abbreviated Genus, so instead of Apistogramma panduro, you'll often see A. panduro. This means I need to search & replace all of the species terms as an abbreviation too - Apistogramma panduro, A. panduro, Satanoperca daemon, S. daemon etc.
If caudal-fin and caudal both exist in the glossary terms, caudal-fin should be replaced first.
I was contemplating simply adding a preg_replace which searched for the terms, but only with a space on the left, (i.e. ( )term) and a space, comma, exclamation, full-stop or hyphen on the right (i.e. term(, . ! - )) but that won't help me to not break the image HTML.
Example content
<br />
It looks very similar to fishes of the <i>B. foerschi</i> group/complex but its breeding strategy, adult size and observed behaviour preclude its inclusion in that assemblage.
Instead it appears to be a member of the <i>B. coccina</i> group which currently includes <i>B. brownorum</i>, <i>B. burdigala</i>, <i>B. coccina</i>, <i>B. livida</i>, <i>B. miniopinna</i>, <i>B. persephone</i>, <i>B. tussyae</i>, <i>B. rutilans</i> and <i>B. uberis</i>.
Of these it's most similar in appearance to <i>B. uberis</i> but can be distinguished by its noticeably shorter dorsal-fin base and overall blue-greenish (vs. green/reddish) colouration.
Members of this group are characterised by their small adult size (< 40 mm SL), a uniform red or black base body colour, the presence of a midlateral body blotch in some species and the fact they have 9 abdominal vertebrae compared with 10-12 in the other species groups. In addition all are obligate peat swamp dwellers (Tan and Ng, 2005).<br />
^^^ This example here has had the correct links manually inserted. The filter shouldn't break these links!
It looks very similar to fishes of the B. foerschi group/complex but its breeding strategy, adult size and observed behaviour preclude its inclusion in that assemblage.
Instead it appears to be a member of the B. coccina group which currently includes B. brownorum, B. burdigala, B. coccina, B. livida, B. miniopinna, B. persephone, B. tussyae, B. rutilans and B. uberis.
Of these it's most similar in appearance to B. uberis but can be distinguished by its noticeably shorter dorsal-fin base and overall blue-greenish (vs. green/reddish) colouration.
Members of this group are characterised by their small adult size (< 40 mm SL), a uniform red or black base body colour, the presence of a midlateral body blotch in some species and the fact they have 9 abdominal vertebrae compared with 10-12 in the other species groups. In addition all are obligate peat swamp dwellers (Tan and Ng, 2005).
^^^ Same example pre-formatting.
[caption id="attachment_542" align="alignleft" width="125" caption="Amazonas Magazine - now in English!"]<img class="size-thumbnail wp-image-542" title="Amazonas English" src="/wp-content/uploads/2011/12/Amazonas-English-1-288x381.jpg" alt="Amazonas English" width="125" height="165" />[/caption]
Edited by Hans-Georg Evers, the magazine 'Amazonas' has been widely-regarded as among the finest regular publications in the hobby since its launch in 2005, an impressive achievment considering it's only been published in German to date. The long-awaited English version is just about to launch, and we think a subscription should be top of any serious fishkeeper's Xmas list...
The magazine is published in a bi-monthly basis and the English version launches with the January/February 2012 issue with distributors already organised in the United States, Canada, the United Kingdom, South Africa, Australia, and New Zealand. There are also mobile apps availablen which allow digital subscribers to read on portable devices.
It's fair to say that there currently exists no better publication for dedicated hobbyists with each issue featuring cutting-edge articles on fishes, invertebrates, aquatic plants, field trips to tropical destinations plus the latest in husbandry and breeding breakthroughs by expert aquarists, all accompanied by excellent photography throughout.
U.S. residents can subscribe to the printed edition for just $29 USD per year, which also includes a free digital subscription, with the same offer available to Canadian readers for $41 USD or overseas subscribers for $49 USD. Please see the Amazonas website for further information and a sample digital issue!
Alternatively, subscribe directly to the print version here or digital version here.
^^^ This will likely only have a few Glossary terms in rather than any species links.
Example terms
$species_terms
339 => 'Aulonocara maylandi maylandi',
340 => 'Aulonocara maylandi kandeensis',
341 => 'Aulonocara sp. "walteri"',
342 => 'Aulonocara sp. "stuartgranti maleri"',
343 => 'Aulonocara stuartgranti',
344 => 'Benthochromis tricoti',
345 => 'Boulengerochromis microlepis',
346 => 'Buccochromis lepturus',
347 => 'Buccochromis nototaenia',
348 => 'Betta brownorum',
349 => 'Betta foerschi',
350 => 'Betta coccina',
351 => 'Betta uberis'
As you can see above, the general format for these scientific names is "Genus species", but can often include "sp." or "aff." (for species which aren't officially described) and "Genus species subspecies" formats.
$glossary_terms
1 => 'abdomen',
2 => 'caudal',
3 => 'caudal-fin',
4 => 'caudal-fin peduncle',
5 => 'Gram\'s Method'
If anyone can come up with a filter which meets all these conditions and requirements, I'd like to offer a bounty.
Thanks in advance,

Answer 1

I think it's better to use DOMDocument functionality than regexps. Here is a working prototype:
// Each dynamically constructed regexp will contain at most 70 subpatterns
define('GROUPS_PER_REGEXPS', 70);
$speciesTerms = array(
339 => '(?:Aulonocara|A\.) maylandi maylandi',
340 => '(?:Aulonocara|A\.) maylandi kandeensis',
344 => '(?:Benthochromis|B\.) tricoti',
345 => '(?:Boulengerochromis|B\.) microlepis',
);
function matchTerms($text) {
// Globals are not good. I left it for the simplicity
global $speciesTerms;
$result = array();
$t = 0;
$speciesCount = count($speciesTerms);
reset($speciesTerms);
while ($t < $speciesCount) {
// Maps capturing group identifiers to term ids
$termMapping = array();
// Dynamically construct regexp
$groups = '';
$c = 1;
while (list($termId, $termPattern) = each($speciesTerms)) {
if (!empty($groups)) {
$groups .= '|';
}
// Match word boundaries, so we don't capture "B. tricotisomeramblingstring"
$groups .= '(\b' . $termPattern . '\b)';
$termMapping[$c++] = $termId;
if (++$t % GROUPS_PER_REGEXPS == 0) {
break;
}
}
$regexp = "/$groups/m";
preg_match_all($regexp, $text, $matches, PREG_OFFSET_CAPTURE);
for ($i = 1; $i < $c; $i++) {
foreach ($matches[$i] as $matchData) {
// matchData[0] holds matched string, e.g. Benthochromis tricoti
// matchData[1] holds offset, e.g. 15
if (isset($matchData[0]) && !empty($matchData[0])) {
$result[] = array(
'text' => $matchData[0],
'offset' => $matchData[1],
'id' => $termMapping[$i],
);
}
}
}
}
// Sort by offset in descending order
usort($result, function($a, $b) {
return $a['offset'] > $b['offset'] ? -1 : 1;
});
return $result;
}
$doc = DOMDocument::loadHTML($html);
// Stack will be used to avoid recursive functions
$stack = new SplStack;
$stack->push($doc);
while (!$stack->isEmpty()) {
$node = $stack->pop();
if ($node->nodeType == XML_TEXT_NODE && $node->parentNode instanceof DOMElement) {
// $node represents text node
// and it's inside a tag (second condition in the statement above)
// Check that this text is not wrapped in <a> tag
// as we don't want to wrap it twice
if ($node->parentNode->tagName != 'a') {
$matches = matchTerms($node->wholeText);
foreach ($matches as $match) {
// Create new link element in the DOM
$link = $doc->createElement('a', $match['text']);
$link->setAttribute('href', 'species/' . $match['id']);
$link->setAttribute('class', 'link_species');
// Save the text after the link
$remainingText = $node->splitText($match['offset'] + strlen($match['text']));
// Save the text before the link
$linkText = $node->splitText($match['offset']);
// Replace $linkText with $link node
// i.e. 'something' becomes 'something'
$node->parentNode->replaceChild($link, $linkText);
}
}
}
if ($node->hasChildNodes()) {
foreach ($node->childNodes as $childNode) {
$stack->push($childNode);
}
}
}
$body = $doc->getElementsByTagName('body');
echo $doc->saveHTML($body->item(0));
Implementation details
I've only showed how to replace species terms, glossary terms will be same. Links are formed in form "species/$id". Abbreviations are handled correctly. DOMDocument is a very reliable parser, it can deal with broken markup and is fast.
?: in regexp allows not to count this subpattern as a capturing group (documentation on subpatterns). Without proper counting of subpatterns, we can't retrieve the termId. The idea is that we build a big regexp pattern by joining all regexps specified in $speciesTerms array and separating them with a pipe |. Final regexp for the first two species would be (spaces for clarity):
First capturing group Alternation Second capturing group
( (?:Aulonocara|A\.) maylandi maylandi ) | ( (?:Aulonocara|A\.) maylandi kandeensis )
So, the text "Examples: Aulonocara maylandi maylandi, A. maylandi kandeensis" will give following matches:
$matches[1] = array('Aulonocara maylandi maylandi') // Captured by the first group
$matches[2] = array('A. maylandi kandeensis') // Captured by the second group
We can clearly say that all elements in matches[1] are referring to the species Aulonocara maylandi maylandi or A. maylandi maylandi which has id = 339.
In short: Use (?:) if you're using subpatterns in $speciesTerms.
UPDATE
Each dynamically created regexp has a limit on maximal number of subpatterns, which is defined as a const at the top. This allows avoiding PCRE limit on number of subpatterns in regexp.
Important notes:
If you have a lot of terms you should rewrite matchTerms, because regexp has a limit on a number of subpatterns. In this case it's optimal to prebuild array of regexps out of every N terms.
matchTerms generates regexp at every call, obviously it can be done only once
It's possible to use advanced regexps in speciesTerms
strlen => mb_strlen if you're using multibyte encodings
Supplied $html will be wrapped in a <body> tag (unless it's already wrapped)

Answer 2

It would be better to parse the HTML rather than trying to use regular expressions. Regex is good when you have something specific you want to match, but gets quirky when you're trying to NOT match certain things.
Using http://simplehtmldom.sourceforge.net/ :
function addLinks(&$p, $species, $terms) {
// much easier to say "not in an anchor tag" with parsed content than with regex
if ($p->tag != 'a') {
// pull out existing elements so they aren't replaced
$children = array();
$x = 0;
foreach ($p->children as &$e) {
$children[] = $e->outertext;
$e->outertext = '---child-'.$x.'---';
$x++;
}
foreach($species as $s) {
$p->innertext = str_replace(
$s,
''.$s.'',
$p->innertext);
}
foreach($term as $t) {
$p->innertext = str_replace(
$t,
'<a href="glossary/'.
strtolower($t[0]).'/'.
strtolower(str_replace(' ','-',$t)).'">'.$t.'</a>',
$p->innertext);
}
// restore previous child elements
foreach($children as $x => $e) {
$p->innertext = str_replace('---child-'.$x.'---', $e, $p->innertext);
}
foreach ($p->children() as &$e) {
addLinks($e, $species, $terms);
}
}
}
$html = new simple_html_dom();
// you may have to wrap $content in a div. not exactly sure how partial content is handled
$html->load($content);
addLinks($html, $species_terms, $glossary_terms);
$content = $html->save();
I haven't used simple_html_dom a whole lot, but that should get you pointed in the right direction.