Truncate html text while taking in consideration "full stops" (in CachePHP TextHelper->truncate)

Edit:

I ended up using CakePHP's truncate() function. It's much faster and supports unicode :D

But the question still remains:

How can I make the function auto-detect full 开发者_StackOverflow中文版stops (.) and cut it just after that? So basically $length would be semi-ignored. So if the new text would have a incomplete sentence, more words would appended until the sentence finishes (Or removed, depending on the length of the string from the cut-off until next/previous sentence)

Edit 2: I found out how to detect full stops. I replaced:

 if (!$exact) {
   $spacepos = mb_strrpos($truncate, ' ');

 ...

with

 if (!$exact) {
    $spacepos = mb_strrpos($truncate, '.');
 ...

edit - problem:

When I have tags like img that have dots inside their attributes, the text gets cutoff inside the tag:

 $text = '<p>Abc def abc def abc def abc def. Abc def <img src="test.jpg" /></p><p>abc def abc def abc def abc def.</p>';

 echo htmlentities(truncate($text));

How can I fix that? I'll open a bounty because the original question was already answered...

---

This snippet solves what you're looking for, and lists it's failures (full stops may not indicate sentence ends, and other punctuation can end sentences).

It will scan characters up to $maxLen and then effectively 'throw away' a partial sentence after the last full stop it finds.

In your case, you'd use this function just before you return $new_text.

---

To resolve the "full-stop in tag" issue, you can use something similar to the following to detect if the stop is within a tag:

$str_len       = strlen($summary);
$pos_stop      = strrpos($summary, '.');
$pos_tag_open  = strrpos($summary, '<', -($str_len - $pos_stop));
$pos_tag_close = strpos($summary, '>', $pos_tag_open);

if (($pos_tag_open < $pos_stop) && ($pos_stop < $pos_tag_close)) {
  // Inside tag! Search for the next nearest prior full-stop.
  $pos_stop = strrpos($summary, '.', -($str_len - $pos_tag_open));
}

echo htmlentities(substr($summary, 0, $pos_stop + 1));

Obviously, this code can be optimized (and pulled out into its own function), but you get the idea. I have a feeling there is a regex that might handle this a bit more efficiently.

Edit:

Indeed, there is a regex that can do this, using negative lookahead:

$text = '<p>Abc def abc def abc def abc def. Abc def <img src="test.jpg" />abc</p>';

$count = preg_match_all("/\.(?!([^<]+)?>)/", $text, $arr, PREG_OFFSET_CAPTURE);
$offset = $arr[0][$count-1][1];

echo substr($text, 0, $offset + 1)."\n";

This should be relatively efficient, at least in comparison with truncate() which also uses preg_match internally.

---

The regular expression above Truncate html text while taking in consideration "full stops" (in CachePHP TextHelper->truncate) might work.

But, considering the efficiency, in this case, we might truncate the string to a max_length first, and then do the preg to the truncated string. And yes, considerations must be done to the punctuation symbols.

Some more rules will create a proper logic to determine the ending of a sentence.

1. a space or a EOL after the punctuation symbol
2. First word after the picked punctuation, having an Upper case.
3. Multiple new lines (end of paragraph) after the punctuation symbol, etc.