用递归和正则表达式替换字符串中的文本

Question

I am using tags to replace text before displaying output in a browser, similar to Wordpress' short codes.

Example string: Hi, this is a block of text {{block:welcome}} and this is a system variable {{variable:system_version}}

I have functions to replace these blocks accordingly, and I realize a foreach or while function will be the best way to deal with it, but unfortunately, replacing one {{...}} may introduce another. Hence, I opted for recursion until no more are found. Typical recursion is only once, but I have had two in one scenario. Maybe calling the function 3 times will work, but it sounds "wrong".

Now that is where the problem occurs: I do NOT want to replace them when they appear in:

1) A page where the URL you are calling contains something
2) Any form element such as `<input>` or `<textarea>`.

I need help on how to exclude from #2 above by means of a regex.

My regex currently look like this: ^\{\{((?!keep).)*$ (I realize it may still be wrong, or need modification - does not quite work yet).

If the item contains "keep", e.g., {{block:welcome:keep}} it should not be replaced, but when doing so, the recursion never stops, as I keep finding items to replace, and thus run out of memory, or get maximum nested level errors.

The reason why I want to do this, is because I do not want the content replaced when on an ADMIN page, or when you are editing form content.

Someone willing to give it a crack? I am using PHP, if that matters.

Thanks!

EDIT 1

Since @Pablo's answer was given to me in chat, I decided to edit my question to reflect why his answer was marked as the correct one.

My regex now look like this: /(?:<(?:textarea|select)[\s\S]*?>[\s\S]*?)?({{variable:(.*?)}})[\s\S]*?(?:<\/(?:textarea|select)>)?|(?:<(?:input)[\s\S]*?)?{{variable:(.*?)}}(?:[\s\S]*?>)?/im

I then check if the match contains an input, select or textarea, and if so, replace the {{ with something else temporarily, and then do my replacement, and when done, change the "something else" back to {{ as Pablo suggested. My regex is thanks to the answer on this question: Text replacement: PHP/regex.

If the above edit does not belong, feel free to remove.

Answer 1

Instead of looking for the perfect RegEx I suggest looking into using preg_replace_callback(). It should allow you to use a simpler RegEx while having more control over the search and replace algorithm for your templating engine. Consider the following example:

1. resolvePlaceholder() generates the replacing content
2. interpolate() parses a template string. It supports nested parsing up to 4 levels.
3. Stop recursive parsing for tags starting with !.

---

<?php

function resolvePlaceholder($name)
{
    $store = [
        'user:first'              => 'John',
        'user:last'               => 'Doe',
        'user:full_name'          => '{{user:first}} {{user:last}}',
        'block:welcome'           => 'Welcome {{user:full_name}}',
        'variable:system_version' => '2019.1',
        'nest-test'               => '{{level1}}',
        'level1'                  => '{{level2}}',
        'level2'                  => '{{level3}}',
        'level3'                  => '{{level4}}',
        'level4'                  => '{{level5}}',
        'level5'                  => 'Nesting Limit Test Failed',
        'user-template'           => 'This is a user template with {{weird-placeholder}} that will not be replaced in edit mode {{user:first}}',
    ];

    return $store[$name] ?? '';
}

function interpolate($text, $level = 1)
{
    // Limit interpolation recursion
    if ($level > 5) {
        return $text;
    }

    // Replace placeholders
    return preg_replace_callback('/{{([^}]*)}}/', function ($match) use ($level) {
        list($tag, $name) = $match;
        // Do not replace tags with :keep
        if (strpos($name, ':keep')) {
            // Remove :keep?
            return $tag;
        }

        if (strpos($name, '!') === 0) {
            return resolvePlaceholder(trim($name, '!'));
        }

        return interpolate(resolvePlaceholder($name), $level + 1);
    }, $text);
}

$sample = 'Hi, this is a block of text {{block:welcome}} and this is a system variable {{variable:system_version}}. ' .
    'This is a placeholder {{variable:web_url:keep}}. Nest value test {{nest-test}}. User Template: {{!user-template}}';

echo interpolate($sample);
// Hi, this is a block of text Welcome John Doe and this is a system variable 2019.1. This is a placeholder {{variable:web_url:keep}}. Nest value test {{level5}}. User Template: This is a user template with {{weird-placeholder}} that will not be replaced in edit mode {{user:first}}