Question

This is a sentence sanitizer.

function sanitize_sentence($string) {
    $pats = array(
    '/([.!?]\s{2}),/',      # Abc.  ,Def
    '/\.+(,)/',             # ......,
    '/(!|\?)\1+/',          # abc!!!!!!!!, abc?????????
    '/\s+(,)/',             # abc   , def
    '/([a-zA-Z])\1\1/');    # greeeeeeen
    $fixed = preg_replace($pats,'$1',$string); # apply pats
    $fixed = preg_replace('/(?:(?<=\s)|^)[^a-z0-9]+(?:(?=\s)|$)/i', '',$fixed); # bad chunks
    $fixed = preg_replace( '/([!?,.])(\S)/', '$1 $2', $fixed); # spaces after punctuation, if it doesn't exist already
    $fixed = preg_replace( '/[^a-zA-Z0-9!?.]+$/', '.', $fixed); # end of string must end in period
    $fixed = preg_replace('/,(?!\s)/',', ',$fixed); # spaces after commas
    return $fixed;
}

This is the test sentence:

hello [[[[[[]]]]]] friend…..? how are you [}}}}}}

It should return:

hello friend…..? how are you

But instead it is returning:

hello friend. .. .. ? how are you.

So there are 2 problems and I can’t find a solution around them:

1. the set of periods are being separated into ".. .. ." for some reason. They should remain as "….." next to the question mark.
2. the end of the string must end in a period only and only if there is at least one of these characters anywhere in the string: !?,. (if at least one of those characters are not found in the string, that preg_replace should not be executed)

Examples for the second problem:

This sentence doesn’t need an ending period because the mentioned characters are nowhere to be found

This other sentence, needs it! Why? Because it contains at least one of the mentioned characters

(of course, the ending period should only be placed if it doesn’t exist yet)

Thanks for your help!

Answer 1

Here is the answer to your first problem. The third-to-last replacement is the problem:

$fixed = preg_replace( '/([!?,.])(\S)/', '$1 $2', $fixed); # spaces after punctuation, if it doesn't exist already

It will match the first period with the character class, and the second period as a non-space character. Then insert a space. Since matches cannot overlap, it will then match the third and forth period and insert a space and so on. This is probably best fixed like this:

$fixed = preg_replace( '/[!?,.](?![!?,.\s])/', '$0 ', $fixed);

Here is how you could go about your second requirement (replace the second-to-last preg_replace):

$fixed = trim($fixed);
$fixed = preg_replace( '/[!?.,].*(?<![.!?])$/', '$0.', $fixed);

First we get rid of leading and trailing whitespace to separate this concern from the trailing period. Then the preg_replace will try to find a punctuation character in the string and if it does, it matches everything until the end of the string. The replacement puts the match back in place and appends the period. Note the negative lookbehind. It asserts that the string does not already end with a sentence-ending punctuation character.