是否可以知道主题字符串中匹配的位置

Question

I have a file name where information has to be replaced. Here is a subject sample :

FileA-2014-11-01_K_1_A2_383.xxx

As many files are to be processed, this filename is first matched by a regex, say :

/[a-zA-Z]*-\d{4}-\d{2}-\d{2}_(\w)_(\d)_A2_(\d*)\.xxx$/

This regex will give me, using preg_match, the values to be replaced, here :

• K=>A
• 1=>2
• 383=>666

My first try was to naively use "str_replace", but it fails when patterns are repeated in the string : here i will get :

FileA-2024-22-02_A_2_A2_666.xxx

So the date is also modified by the str_replace (as it was told to do..)

So, i wonder if there is a way to know where is a given match in the string to have a clean replacement. I'm now trying to revert the regex to be able to capture non-replacement blocks, and then insert replaced data. That regex would be :

/([a-zA-Z]*-\d{4}-\d{2}-\d{2}_)\w(_)\d(_A2_)\d*(\.xxx)$/

With that one, i'm able to keep non-replaced parts. I now have to find a kind of index to know the replacement position in the string. I guess I can achieve this way, but is seems somewhat complicated and error prone. Given I only have the initial regex and the map for to=>from replacement, is there a way to do that in a better way?

[EDIT : solution]

<?php

$filename = "FileA-2014-11-01_K_1_A2_383.xxx";
$expected = "FileA-2014-11-01_A_2_A2_666.xxx";

$regex = "/[a-zA-Z]*-\d{4}-\d{2}-\d{2}_(\w)_(\d)_A2_(\d*)\.xxx$/";


global $replacements;

$replacements["K"] = "A";
$replacements["1"] = "2";
$replacements["383"] = "666";


$result = preg_replace_callback($regex, function($matches){
    global $replacements;
    print_r($matches);
    // ended here. no way.
}, $filename);


if(strcmp($result,$expected)==0)
    echo "preg_replace_callback() : Yep
";
else
    echo "preg_replace_callback() : Nop
";


preg_match($regex, $filename, $matches, PREG_OFFSET_CAPTURE);

// remove useless global string match
array_shift($matches);

$result = $filename;
foreach($matches as $matchInfo){

    $match    = $matchInfo[0];
    $position = $matchInfo[1];

    $matchLength= strlen($match);

    $beforeReplacementPart = substr($result, 0, $position);
    $afterReplacementPart = substr($result, ($position + $matchLength));
    $result = $beforeReplacementPart . $replacements[$match] . $afterReplacementPart;

}


if(strcmp($result,$expected)==0)
    echo "preg_match() and substr game : Yep
";
else
    echo "preg_match() and substr game : Nop
";

Answer 1

A regex that matches that filename:

$re  = '/[a-zA-Z]*-\d{4}-\d{2}-\d{2}_(\w)_(\d)_A2_(\d*)\.xxx$/';
$str = 'FileA-2014-11-01_K_1_A2_383.xxx';

If you add PREG_OFFSET_CAPTURE as the fourth parameter ($flags) to the call to preg_match(), it will also return the offset of each captured string in the third parameter:

preg_match($re, $str, $matches, PREG_OFFSET_CAPTURE);

A print_r($matches) will reveal:

Array
(
    [0] => Array
        (
            [0] => FileA-2014-11-01_K_1_A2_383.xxx
            [1] => 0
        )
    [1] => Array
        (
            [0] => K
            [1] => 17
        )
    [2] => Array
        (
            [0] => 1
            [1] => 19
        )
    [3] => Array
        (
            [0] => 383
            [1] => 24
        )
)

$matches[0] is the part that matched the entire regex. $matches[1] is the first capturing sub-expression, $matches[2] is the second and so on.

$matches[1][0] is the fragment from the input string that matched the first regex sub-expression (\w) and $matches[1][1] is the offset in the input string where it was found. The same for $matches[N][0] and $matches[N][1] for the N^th sub-expression.

If you need to do a simple replacement then you don't need to bother about offsets but use preg_replace() or, if the replacement expression is complex or dynamic, preg_replace_callback().

Using preg_replace() you need to capture the parts you want to keep:

$re  = '/([a-zA-Z]*-\d{4}-\d{2}-\d{2}_)\w_\d_A2_\d*(\.xxx)$/';
$str = 'FileA-2014-11-01_K_1_A2_383.xxx';

$new = preg_replace($re, '$1A_2_A2_666$2', $str);
echo($new."
");

In the replacement string, $1 and $2 denote the sub-expressions from the regex. We marked them for capturing in order to re-use them in the replacement string.