A PHP script to convert text files of Prince Edward Island Hansard to a CSV

hansard-to-csv.php

<?php

$files = popen('find ./documents -name "*.txt"', 'r');

$out = fopen('hansard.csv', 'w');
fwrite($out, "date,text,speaker\n");

while (!feof($files)) {
	$filename = chop(fgets($files, 4096));
	$date = basename($filename);
	$date = str_replace("-hansard.txt", "", $date);
	if ($filename != '') {
		$fp = fopen($filename, 'r');
		$text = array();
		while(!feof($fp)) {
			$line = [];
			$line['date'] = $date;
			$line['text'] = chop(fgets($fp, 4096));
			$text[] = $line;
		}

		$started = FALSE;

		foreach($text as $lineno => $line) {
			// This line begins with a member's name, like Mr. Smith:
			if (
					(preg_match("/^(([A-Z][a-zA-Z\.\-]+\s?\b){1,4}):/", $line['text'], $matches)) or
					(preg_match("/^(Leader of the Opposition):/", $line['text'], $matches)) or
					(preg_match("/^(Chair \(.*\)):/", $line['text'], $matches)) or
					(preg_match("/^(Clerk Assistant \(.*\)):/", $line['text'], $matches))
				 ) {
				if (!preg_match("/Published by Order of the Legislature/", $line['text'])) {
					$text[$lineno]['speaker'] = $matches[1];
					$text[$lineno]['text'] = str_replace($matches[0] . ' ' , '', $text[$lineno]['text']);
					$currentline = $lineno;
					$started = TRUE;
				}
				else {
					unset($text[$lineno]);
				}
			}
			// This line does not begin with a member's name
			else {
				if (preg_match("/^.*HANSARD PEI.*/", $line['text'])) {
					unset($text[$lineno]);
				}
				else if (preg_match("/^\d+/", $line['text'])) {
					unset($text[$lineno]);
				}
				else if ($started) {
					$text[$currentline]['text'] .= ' ' . $line['text'];
					unset($text[$lineno]);
				}
				else {
					unset($text[$lineno]);
				}
			}
		}
		foreach($text as $lineno => $line) {
			$text[$lineno]['wordcount'] = str_word_count($line['text']);
		}

		foreach($text as $lineno => $line) {
			fputcsv($out, array_values($line));
		}
	}
}

fclose($out);