Add setUrl() method

[hackel]

Typically when working with APIs, all requests are sent to the same URL. It doesn't make sense to specify this over and over again. There should be a single call to set the URL to use for every request.

[zachborboa]

Thank you for the issue. Can you provide an example use case?

[hackel]

Anything in which you are performing multiple requests to the same URL. E.g.:

$curl->setUrl('http://somegreat/api');
$curl->post(array('name'=>'foo');
$curl->get(array('q'=>'keyword'));
$curl->patch(array('id'=>100, 'name'=>'bar');

It's not a requirement, I can certainly include the URL in every request (which I am doing now), it just seems like it might be a natural/elegant method to add. Even what I wrote above seems too procedural. Possibly chaining them like this might be better:

Curl::url('http://somegreat/api')
    ->setOpt(CURLOPT_AUTOREFERER, true)
    ->setCookie(...)
    ->get(...);

[zborboa-g]

Here is how this could be implemented

$curl = new Curl();
$curl->setUrl('https://api.example.com/inbox/');

// OR

$curl = new Curl('https://api.example.com/inbox/');

And the usage:

for ($i = 1; $i <= 10; $i++) {
    $curl->get(array(
        'page' => $i,
    ));
}

However, there would be a bit of ambiguity. For example, $curl->get() currently expects a url string or an array of urls as the first parameter. Using $curl->setUrl() and then calling $curl->get() with an array of data would be ambiguous as we would not know if the array passed is a set of urls or data.

[hackel]

Yes, I don't see an obvious way to maintain backward compatibility... You could possibly move the parallel processing into it's own get_parallel function or something, and then check if $url is an array in every function, and if so use it as $data. This might be too big a change, though.

[zborboa-g]

Actually I believe there is a way this could work by using Curl::is_array_assoc() on the first parameter. A user is going to do parallel requests using get() with an indexed array and not with an associated array.