<?php

/**

* Cleans up artifacts of text blocks for an experiment.

* Uses default values for cleanup parameters.

*

* @param array $textBlocks An array of text block data. Each element should contain at least a 'text' key.

* @return array The cleaned text block data.

*/

function cleanupTextBlocks(array $textBlocks): array

{

$defaultParams = [

'removeLeadingWhitespace' => true,

'removeTrailingWhitespace' => true,

'removeExcessNewlines' => true,

'normalizeWhitespace' => true,

'removeHTMLTags' => false, // Default: Don't remove HTML

];

$params = array_merge($defaultParams, $argv); // Allow overriding with command-line arguments (optional)

$cleanedBlocks = [];

foreach ($textBlocks as $block) {

$text = $block['text'] ?? ''; // Default to empty string if 'text' key is missing

//Remove leading/trailing whitespace

$text = trim($text, $params['removeLeadingWhitespace'] ? "\x00\x09\x0a\x0d\x0c" : "");

$text = trim($text, $params['removeTrailingWhitespace'] ? "\x00\x09\x0a\x0d\x0c" : "");

// Remove excessive newlines

$text = preg_replace('/\s*\n+/', "\n", $text, 1); // Replace multiple newlines with single newline

// Normalize whitespace

$text = preg_replace('/\s+/', ' ', $text); // Replace multiple spaces with single space

// Remove HTML tags (optional)

if ($params['removeHTMLTags']) {

$text = strip_tags($text);

}

$cleanedBlocks[] = ['text' => $text]; // Store cleaned text

}

return $cleanedBlocks;

}

//Example usage (can be tested with command line arguments):

/*

$textBlocks = [

['text' => ' This is a test.\n\nWith some extra whitespace. '],

['text' => '<div>This has HTML</div>'],

['text' => 'No HTML here'],

];

$cleaned = cleanupTextBlocks($textBlocks);

print_r($cleaned);

*/

?>