parse_pdf.php

<?php
// Run this script in CLI mode, e.g: php parse_pdf.php full_path_to_input_file.txt

$time_start = microtime(true);

$input_file = $argv[1];

$handle_input = fopen($input_file, "r");
$handle_output = fopen($input_file . ".csv", "w+");

if(!$handle_input)
{
	exit("unable to input file");
}

if(!$handle_output)
{
	exit("unable to output file");
}

$skip_line = true;
$row = array("Date", "Doc No.", "Debit", "Credit", "Balance", "Details");

while (($line = fgets($handle_input)) !== false)
{
	$trimmed_line = ltrim($line);
	if (strlen($trimmed_line) < 20 && preg_match("/^\d{2}\/\d{2}\/\d{4}/", $trimmed_line))
	{
		foreach($row as $key => &$cell)
		{
			$cell = str_replace(array("\r\n", "\n", "\r"), "", $cell);
			$cell = trim(preg_replace('/[\t\n\r\s]+/', " ", $cell));
		}

		fputcsv($handle_output, $row);

		$row = array("date" => "", "docno" => "", "debit" => "", "credit" => "", "balance" => "", "details" => "");
		$row["date"] .= substr($trimmed_line, 0, 11);

		$skip_line = false;
		continue;
	}

	if (strpos($trimmed_line, "Postal address") === 0)
	{
		$skip_line = true;
	}

	if($skip_line === true)
	{
		continue;
	}

	$row["docno"] .= trim(substr($line, 0, 11));
	$row["debit"] .= trim(substr($line, 40, 19));
	$row["credit"] .= trim(substr($line, 59, 30));
	$row["balance"] .= trim(substr($line, 90, 19));
	$row["details"] .= trim(substr($line, 110), "\n\r");
}

foreach($row as $key => &$cell)
{
	$cell = str_replace(array("\r\n", "\n", "\r"), "", $cell);
	$cell = trim(preg_replace('/[\t\n\r\s]+/', " ", $cell));
}
fputcsv($handle_output, $row);

fclose($handle_input);
fclose($handle_output);

$time_end = microtime(true);
$execution_time = number_format($time_end - $time_start, 2);

echo "Execution time: " . $execution_time . "s" . PHP_EOL;
echo "Peak memory usage: " . number_format((memory_get_peak_usage(true)/1024/1024), 2) . "MB" . PHP_EOL;