example.php

<?php
/*
 * Running the example program:
 * $ php example.php
 * filesize: 10000
 * histogram: 0:40, ..., f:40, 10:39, ..., ff:39
 * delta histogram: 1:9960, -255:39
 *
 * Without output buffering:
 * $ strace php example.php 2>&1 | grep 'write(4' | wc -l
 * 10000
 *
 * With output buffering:
 * $ strace php example.php 2>&1 | grep 'write(4' | wc -l
 * 3      # 4096 + 4096 + 1808 = 10000 bytes
 */
function calc_hist($contents, $fsize)
{
    $hist = array_pad([], 256, 0);
    for ($i = 0; $i < $fsize; $i++) {
        $key = ord($contents[$i]);
        $hist[$key]++;
    }
    return $hist;
}

function print_hist($hist)
{
    $toks = [];
    foreach ($hist as $key => $val) {
        $toks[] = dechex($key) . ":$val";
    }
    echo 'histogram: ' . implode(', ', $toks) . "\n";
}

function calc_deltahist($contents, $fsize)
{
    $deltahist = [];
    for ($i = 1; $i < $fsize; $i++) {
        $from = ord($contents[$i-1]);
        $to = ord($contents[$i]);
        $delta = $to - $from;
        if (!isset($deltahist[$delta]))
            $deltahist[$delta] = 0;
        $deltahist[$delta]++;
    }
    return $deltahist;
}

function print_deltahist($deltahist)
{
    $toks = [];
    foreach ($deltahist as $key => $val) {
        $toks[] = "$key:$val";
    }
    echo 'delta histogram: ' . implode(', ', $toks) . "\n";
}


$fname = '/tmp/example.txt';
$fp = fopen($fname, 'w');

/*
 * Requires that extension=output_buffer is added to /etc/php/8.3/cli/php.ini
 */
stream_filter_append($fp, 'output.buffer');

/*
 * This simulates some legacy/library code that does a lot of small
 * writes and has a terrible performance but which we cannot easily
 * change or don't want to touch for other reasons
 */
for ($i = 0; $i < 10000; $i++) {
    fwrite($fp, chr($i));
}
fclose($fp);

echo "filesize: " . filesize($fname) . "\n";  // should be 10000

$handle = fopen($fname, 'r');
$fsize = filesize($fname);
$contents = fread($handle, $fsize);
fclose($handle);

$hist = calc_hist($contents, $fsize);
print_hist($hist);

$deltahist = calc_deltahist($contents, $fsize);
print_deltahist($deltahist);
?>