This pull contains a number of improvements:

• An anonymous function to handle writing to php://memory to mitigate (https://bugs.php.net/bug.php?id=43468).
• Support for PECL-style Header output (multi-dimensional associative arrays).
• Support for setting a group of headers.
• Support for setting Headers in the PECL-style (multi-dimensional associative arrays).
• Support for automatic JSON encoding of the POST contents if the 'Content-Type' is application/json.
• Supports URLs with place-holders (Restful URLs).
• Added methods:
  • setRequestType() - Sets the CURLOPT_CUSTOMREQUEST to Uppercase Request Type, and sets CURLOPT_HTTPGET, CURLOPT_POST, CURLOPT_NOBODY depending on the type.
  • setURL() - Sets the base_url to the provided URL with optional place-holder substitution, sets url to the base_url with optional GET request appended, sets CURLOPT_URL to the resulting url.
  • buildBaseURL() - Performs place-holder substitution on a provided URL with the provided $url_subs.
  • normalizeContentType() - Filters a set of Headers and resolves a Lower Case normalized representation of the Content Type (xml/json at this time).
  • more.

Got this error with last commits on master branch. Last commit at this moment (42d84d7). Use MultiCurl with proxy and MultiCurl::stop() method.

curl_multi_info_read(): supplied resource is not a valid cURL handle resource {"exception":"[object] (ErrorException(code: 0): curl_multi_info_read(): supplied resource is not a valid cURL handle resource at /var/www/html/vendor/php-curl-class/php-curl-class/src/Curl/MultiCurl.php:998)"}

Post-redirect-get is a workflow where the an user submits data using a POST request (on a login page for example) and receive a "303: See other" response in return (for example, to be redirected to its home page after a successful login), redirection which is supposed to be followed using GET requests.

This workflow is used by lots of websites, and is actually consistent with the HTTP/1.1 specs.

Using the cURL php extension, it is possible to force requests resulting from a redirect after a POST request to also be done using using POST. It is achieved by setting the CURLOPT_CUSTOMREQUEST option to "POST" (source). And this is what php-curl-class currently does, preventing post-redirect-get requests:

public function post($url, $data = array())
{
    if (is_array($url)) {
        $data = $url;
        $url = $this->baseUrl;
    }

    $this->setURL($url);
    $this->setOpt(CURLOPT_CUSTOMREQUEST, 'POST');
    $this->setOpt(CURLOPT_POST, true);
    $this->setOpt(CURLOPT_POSTFIELDS, $this->buildPostData($data));
    return $this->exec();
}

Fixing this is quite easy, we just need to reset the CURLOPT_CUSTOMREQUEST option by setting it to null (the request will still be a POST due to the CURLOPT_POST option). One downside though: before PHP 5.5.11 (april 2014), it was not possible to reset this field (see this patch).

Example usage:

$curl = new Curl();
$curl->setRetry(function($instance) {
    // Retry when request times out.
    return $instance->retries < 3 && $instance->error === CURLE_OPERATION_TIMEOUTED;
});

#200

A user may wish to disable the XML parsing, since when downloading to a file, such parsing shouldn't do anything, since rawResponse wouldn't contain a string anyway.

Also should be able to disable XML or JSON parsing by giving null as a handler.

There has a bug in unsetHeader()

The unsetHeader bad work, for example

$curl->setHeader('Host','example.com'); $curl->unsetHeader('Host'); $curl->get('http://127.0.0.1/');

This request is make with Header "Host" = '' (empty string), because unsetHeader bad work

with the my fix the header 'Host' with be deleted and curl auto-retrive the right header. horever is already possible make request with header 'Host' = '' (empty string) with $curl->setHeader('Host', '' );

also remove a bad comment about fuction

• \* @return string
  

  in setHeader()

What is the expected behavior? return a response

What is the actual behavior? the response is always empty, but rawresponse is not

What steps will reproduce the problem? i am using it in the most stripped down basic usecase. just like in your examples. But the response is always empty, even tho the rawresponse is not. the response is a correctly formatted json. no errors given.

Code:

$curl->response

    Hello, i have the following questions.
    I send the first request with CURLOPT_HTTPHEADER and this value set Content-Type is multipart/form-data.
    
    `$ch = curl_init();`
    `curl_setopt($ch, CURLOPT_HTTPHEADER, array('Content-Type:multipart/form-data'));`
    `curl_setopt($ch, CURLOPT_URL, 'http://www.baidu.com');`
    `curl_setopt($ch, CURLOPT_POST, true);`
    `curl_exec($ch);`
    `curl_close($ch);`

    The next curl post request i use php-curl-class and not set CURLOPT_HTTPHEADER, but the server cannot receive the data.

    `$curl = new \Curl\Curl();`
    `$curl->post([server_host], array('name'=>'name'));`
    `var_dump($curl1->response);// output array(0) { }`
    `exit;`

    If i don't use php-curl-class or use other classlib the server can right receive the data.

    `$ch2 = curl_init();`
    `curl_setopt($ch2, CURLOPT_URL, [server_host]);`
    `curl_setopt($ch2, CURLOPT_POST, true);`
    `curl_setopt($ch2, CURLOPT_POSTFIELDS, array('name'=>'name'));`
    `curl_exec($ch2);// output {"name":"name"}`
    `curl_close($ch2);`
    `exit;`

    No matter use which kinds of method, the second request header Content-Type value both is multipart/form-data, but use the php-curl-class can't receive the data.

If the CURLFile Class doesn´t exists, the class_exists call in Curl::buildPostData (Line: 155) produce a fatal error, cause of the missing 2nd parameter "autoload".

A couple of fixes for downloads of "volume"

1. I noticed that $mc = new MultiCurl; unset($mc); does not call __destruct (in fact, $c = new Curl; unser($c); doesn't either) as the $ch is still referenced by curl(), etc, so, for convenience, I added autoClose that will automatically call $ch->close(), which will destruct the objects and free up memory consumed by them.
2. When calling many downloads, I wanted to be able to manage the concurrency of the downloads, so I added concurrency(). Now, only X downloads will process at once. This manages an internal queue, and it will process all by default (concurrency of 0).
3. Also when managing many downloads (think, thousands) I hit a file pointer ceiling. I mitigated this by deferring the creating of the file pointers until initHandle() is called, which will stay within the concurrency limits.

Hope this finds you well.

Often times API or websites rate limits certain amounts of requests per minute. In multi_curl() this causes problem. A built in rate limiter would be extremely useful in such cases. Guzzle has something called guzzle-rate-limiter-middleware

Probably will be like:

MultiCurl::perSecond(3); // Max. 3 requests per second

MultiCurl::perMinute(5); // Max. 5 requests per minute

What is the expected behavior? Having some base class that will keep shared methods so Curl and the MuliCurl classes will extend it.

What is the actual behavior? The Curl and the MultiCurl haveing lots of methods which are exactly the same.

What steps will reproduce the problem?

Code:

What is the expected behavior?

getCookie() : expect arg to accept a decoded name expect the return value to be decoded

getCookies() expect decoded names/values

responseCookies public property : expect decoded names/values

What is the actual behavior?

must pass an encoded name to getCookie() and it returns a raw/encoded value

What steps will reproduce the problem?

response headers include

Set-Cookie: dingus=foo%3Bbar; path=/
Set-Cookie: a%3Bb=foo%3Bbar; path=/

Code:

$curl->getCookie('dingus') // returns "foo%3Bbar"   I would expect "foo;bar"
$curl->getCookie('a;b'); // returns null, I would expect "foo;bar"
$curl->getCookie('a%3Bb'); // returns "foo%3Bbar"  -> ugh

likewise, I would expect getCookies (and public responseCookies property) to return array of decoded keys/values

feature or bug?

I am missing a better error control on Curl class. May be, having the posibility to decorate the class with some methods to control errors and throw the needed exceptions. That would not change the class behaviour and would add a good feature. Do you think it could be interesting for the project?