Manipulate a string that is 30 million characters long

Question

I am downloading a CSV file from another server as a data feed from a vendor.

I am using curl to get the contents of the file and saving that into a variable called $contents.

I can get to that part just fine, but I tried exploding by \r and \n to get an array of lines but it fails with an 'out of memory' error.

I echo strlen($contents) and it's about 30.5 million chars.

I need to manipulate the values and insert them into a database. What do I need to do to avoid memory allocation errors?

Solution 1

PHP is choking because it's running out memory. Instead of having curl populate a PHP variable with the contents of the file, use the

CURLOPT_FILE

option to save the file to disk instead.

//pseudo, untested code to give you the idea

$fp = fopen('path/to/save/file', 'w');
curl_setopt($ch, CURLOPT_FILE, $fp);
curl_exec ($ch);
curl_close ($ch);
fclose($fp);

Then, once the file is saved, instead of using the file or file_get_contents functions (which would load the entire file into memory, killing PHP again), use fopen and fgets to read the file one line at a time.

Solution 2

As other answers said :

• you can't have all that in memory
• a solution would be to use CURLOPT_FILE

But, you might not want to really create a file you could want to work with data in memory... Using it as soon as it "arrives".

One possible solution might be defining your own stream wrapper, and use this one, instead of a real file, with CURLOPT_FILE

First of all, see :

• stream_wrapper_register
• The streamWrapper class
• Example class registered as stream wrapper

And now, let's go with an example.

First, let's create our stream wrapper class :

class MyStream {
    protected $buffer;

    function stream_open($path, $mode, $options, &$opened_path) {
        // Has to be declared, it seems...
        return true;
    }

    public function stream_write($data) {
        // Extract the lines ; on y tests, data was 8192 bytes long ; never more
        $lines = explode("\n", $data);

        // The buffer contains the end of the last line from previous time
        // => Is goes at the beginning of the first line we are getting this time
        $lines[0] = $this->buffer . $lines[0];

        // And the last line os only partial
        // => save it for next time, and remove it from the list this time
        $nb_lines = count($lines);
        $this->buffer = $lines[$nb_lines-1];
        unset($lines[$nb_lines-1]);

        // Here, do your work with the lines you have in the buffer
        var_dump($lines);
        echo '<hr />';

        return strlen($data);
    }
}

What I do is :

• work on the chunks of data (I use var_dump, but you'd do your usual stuff instead) when they arrive
• Note that you don't get "full lines": the end of a line is the beginning of a chunk, and the beginning of that same line was at the end of the previous chunk so, you have to keep some parts of a chunk between the calls to stream_write

Next, we register this stream wrapper, to be used with the pseudo-protocol "test" :

// Register the wrapper
stream_wrapper_register("test", "MyStream")
    or die("Failed to register protocol");

And, now, we do our curl request, like we would do when writting to a "real" file, like other answers suggested :

// Open the "file"
$fp = fopen("test://MyTestVariableInMemory", "r+");

// Configuration of curl
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, "http://www.rue89.com/");
curl_setopt($ch, CURLOPT_HEADER, 0);
curl_setopt($ch, CURLOPT_BUFFERSIZE, 256);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_FILE, $fp);    // Data will be sent to our stream ;-)

curl_exec($ch);

curl_close($ch);

// Don't forget to close the "file" / stream
fclose($fp);

Note we don't work with a real file, but with our pseudo-protocol.

This way, each time a chunk of data arrives, MyStream::stream_write method will get called, and will be able to work on a small amount of data (when I tested, I always got 8192 bytes, whatever value I used for CURLOPT_BUFFERSIZE)

A few notes :

• You need to test this more than I did, obviously
• my stream_write implementation will probably not work if lines are longer than 8192 bytes up to you to patch it ;-)
• It's only meant as a few pointers, and not a fully-working solution: you have to test (again), and probably code a bit more!

Still, I hope this helps ;-)
Have fun !

Solution 3

Darren Cook comment to Pascal MARTIN response is really interesting. In modern PHP+Curl versions, the CURLOPT_WRITEFUNCTION option can be set so CURL invokes a callback for each received "chunk" of data. Specifically, the "callable" will received two parameters, the first one with the invoking curl object, and the second one with the data chunk. The funcion should return strlen($data) in order for curl to continue sending more data.

Callables can be methods in PHP. Using all this, I've developed a possible solution that I find more readable that the previous one (although Pascal Martin response is really great, things have changed since then). I've used public attributes for simplicity, but I'm sure readers could adapt and improve the code. You can even abort the CURL request when a number of lines (or bytes) have been reached. I hope this would be useful for others.

<?
class SplitCurlByLines {

    public function curlCallback($curl, $data) {

        $this->currentLine .= $data;
        $lines = explode("\n", $this->currentLine);
        // The last line could be unfinished. We should not
        // proccess it yet.
        $numLines = count($lines) - 1;
        $this->currentLine = $lines[$numLines]; // Save for the next callback.

        for ($i = 0; $i < $numLines; ++$i) {
            $this->processLine($lines[$i]); // Do whatever you want
            ++$this->totalLineCount; // Statistics.
            $this->totalLength += strlen($lines[$i]) + 1;
        }
        return strlen($data); // Ask curl for more data (!= value will stop).

    }

    public function processLine($str) {
        // Do what ever you want (split CSV, ...).
        echo $str . "\n";
    }

    public $currentLine = '';
    public $totalLineCount = 0;
    public $totalLength = 0;

} // SplitCurlByLines

// Just for testing, I will echo the content of Stackoverflow
// main page. To avoid artifacts, I will inform the browser about
// plain text MIME type, so the source code should be vissible.
Header('Content-type: text/plain');

$splitter = new SplitCurlByLines();

// Configuration of curl
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, "http://stackoverflow.com/");
curl_setopt($ch, CURLOPT_WRITEFUNCTION, array($splitter, 'curlCallback'));

curl_exec($ch);

// Process the last line.
$splitter->processLine($splitter->currentLine);

curl_close($ch);

error_log($splitter->totalLineCount . " lines; " .
 $splitter->totalLength . " bytes.");
?>

Solution 4

You might want to consider saving it to a temporary file, and then reading it one line at a time using fgets or fgetcsv.

This way you avoid the initial big array you get from exploding such a large string.

Solution 5

1. Increase memory_limit in php.ini.
2. Read data using fopen() and fgets().

View more solutions