zendframework / zend-mail / 24

38%

Pull #7

travis-ci

Protocol\Smtp->data(): be more memory friendly (stream version)

The only problem with this version is the dependency on "php://temp"
(not sure if it's enable on all systems, fopen/fwrite can fail)

Before this patch, we were removing \r, now we're keeping them,
in any case \r should be encoded

With huge mails (50M, 660000 lines), i'm hitting memory_limit, so i've done some tests
looking at:
- peak memory usage for big mails (50M, 660000 lines)
- average time(100 runs) for big mails
- average time(1000 runs) for small mails (50k, 1700 lines)

with php5.6/php7

- str_replace("\r","") + explode("\n")
(what we use now)
260M/200M|0,23s/0,08s|0,29ms/0,18ms
```
$data = str_replace("\r", '', $data);
$data = explode("\n", $data);
foreach ($data as $line) {
    if (isset($line[0]) && $line[0] === '.') {
        $line = '.' . $line;
    }
    //send
}
```

- preg_split("\r?\n")
200M/160M|1,5s/0,14s|1,75ms/0,22ms
```
foreach (preg_split('/\r?\n/', $data) as $line) {
    if (isset($line[0]) && $line[0] === '.') {
        $line = '.' . $line;
    }
    //send
}
```

- strpos("\n") + $data[]==="\r" + substr()
50M/50M|0,32s/0,16s|0,86ms/0,41ms
```
$start = 0;
while (($stop = strpos($data, "\n", $start)) !== false) {
    $len = $stop-$start;
    if ($len > 0 && $data[$stop-1] === "\r") {
        $len -= 1;
    }
    $line = substr($data, $start, $len);

    if (isset($line[0]) && $line[0] === '.') {
        $line = '.' . $line;
    }
    //send
    $start = $stop + 1;
}
//last line
$line = substr($data, $start);
if (isset($line[0]) && $line[0] === '.') {
    $line = '.' . $line;
}
//send
```

- str_replace("\r","") + strpos("\n") + substr()
150M/100M|0,28s/0,14s|0,76ms/0,31ms
```
$start = 0;
$data = str_replace("\r", '', $data);
while (($stop = strpos($data, "\n", $start)) !== false) {
    $line = substr($data, $start, $stop-$start);
    if (isset($line[0]) && $line[0] === '.') {
        $line = '.' . $line;
    }
    //send
    $start = $stop + 1;
}
//last line
$li... (continued)

10 of 10 new or added lines in 1 file covered. (100.0%)

1533 of 3787 relevant lines covered (40.48%)

9.45 hits per line