[RESOLVED] Better way with regex to do this, currently using explode

Hiccup

Hey guys,

I have a script that currenlty does the job using the explode function, but I'm wondering if there's some funky regex that will do it faster.

Right now I'm just using explode on the periods and looping through each piece to find what I want. It works, but I'd like to optimize this as much as I can.

I have a block of text and I want to pull 2 sentences from it. The first sentence starts with the search term and the second sentence is the one immediately following it.

For example:

The search term is "Diet pills are". I want to pull any sentences starting with that term and then the sentence the follows it as my results.

Here's some same text:

Diet pills are in great demand in the market. About half of the population is overweight nowadays and so they look for a solution - most often keeping in tune with the trends of time a quick solution. So, people mostly use diet pills to reduce their weight. There are many diet pills available in market.But are these pills really effective? It is true that pills helps in reducing the weight. But do you know that diet pills mostly have negative side effects. Some people are also addicted to it.These pills are not regulated by any official body in real terms. The law actually doesn't require pills to be examined by FDA before making them available to public. Diet pills are the #1 choice for weight loss these days. And for good reason.

So from that paragraph, I only want the bold parts. Any regex using preg and php that can do that?

NogDog

First the caveat: changing an explode()-based solution to a regexp solution in no way guarantees better performance. The fact that you could conceivably change several lines of code into one preg_*() command does not necessarily mean that it compiles faster nor that it executes faster.

But, if you want to try it, it could be something like (untested):

preg_match('#^([^.]\.\s*){2}.*([^.]\.\s*){2}$#s', $text, $matches);
echo "Start: " . $matches[1];
echo "End: " . $matches[2];

Hiccup

Good point. I'll run some tests and see which is faster.

Thanks for the help.

🙂

ixalmida

A more conventional method, if you're just looking for the first two matches:

// Find string 1...
$pos1 = strpos($mystring, "Diet pills are");
// Clip the string at the first instance...
$substr = substr($mystring, $pos1);
// Get the entire sentence...
$str1 = strtok($substr, ".");

// Find string 2 (offset is arbitrary, as long as it skips the first character)...
$pos2 = strpos($substr, "Diet pills are", 10);
// Clip the string at the second instance...
$substr = substr($substr, $pos2);
// Get the entire sentence...
$str2 = strtok($substr, ".");

You may want to test this against your current function (and/or the preg_match() expression Nog suggested) and see which is faster My guess is that the above function might be faster as the input string grows in size since it searches from the beginning of the string and then stops.

There is one flaw with all of these functions: a period doesn't always end a sentence. If you were being really thorough, you would want to eliminate expressions like: "e.g.", "i.e.", decimals, etc. A regular expression might be best if you wanted to be that thorough.

Hiccup

Thank you I will test that one too. The first regex doesn't have a spot for the search term, but I think I can fit it in there. (I'm regex retarded.)

NogDog

Hiccup;10925953 wrote:
Thank you I will test that one too. The first regex doesn't have a spot for the search term, but I think I can fit it in there. (I'm regex retarded.)

Sorry, I didn't read your original post closely enough. For whatever reason I thought you wanted the first two and last two sentences of the supplied text. Somehow I missed the part about what you really wanted to search for and capture. :rolleyes:

NogDog

Again, untested:

preg_match_all('#'.preg_quote($search_term).'([^.]*\.){,2}#i', $text, $matches);
print_r($matches[0]);

ixalmida

Just for the record, I think there are numerous "good" ways to do most things and in most cases the only reason to sweat one versus the other is if there is a noticeable performance impact.

nrg_alpha

@OP,
As NogDog mentioned, regex isn't necessarily going to be faster. In fact, in my experience, clean, efficiently written non-regex solutions are actually faster on the whole then their regex counterparts. I suppose it all depends on the situation.

Given the assumption of sentences always ending in a period (not realsitic of course!), I like your selection of tools for choice (that is to say, strpos and strtok). Another alternative could be:

$str = "Diet pills are in great demand in the market. About half of the population is overweight nowadays and so they look for a solution - most often keeping in tune with the trends of time a quick solution. So, people mostly use diet pills to reduce their weight. There are many diet pills available in market.But are these pills really effective? It is true that pills helps in reducing the weight. But do you know that diet pills mostly have negative side effects. Some people are also addicted to it.These pills are not regulated by any official body in real terms. The law actually doesn't require pills to be examined by FDA before making them available to public. Diet pills are the #1 choice for weight loss these days. And for good reason.";
$searchTerm = 'Diet pills are';
$strToken = strtok($str, '.');

while($strToken !== false){
    if(strpos($strToken, $searchTerm) !== false) $searchFoundArray[] = trim($strToken) . '.' . $strToken = strtok('.') . '.';
    $strToken = strtok('.');
}

echo '<pre>'.print_r($searchFoundArray, true);

But yeah, regex would be the better choice here I think, simply given the possible different punctuations involved.