Asynchronous cURL executor simply based on resource and Generator
| PHP | ❓ | Feature Restriction |
|---|---|---|
| 7.0~ | 😄 | Full Support |
| 5.5~5.6 | 😧 | Generator is not so cool |
| ~5.4 | 💥 | Incompatible |
function curl_init_with(string $url, array $options = [])
{
$ch = curl_init();
$options = array_replace([
CURLOPT_URL => $url,
CURLOPT_RETURNTRANSFER => true,
], $options);
curl_setopt_array($ch, $options);
return $ch;
}
function get_xpath_async(string $url) : \Generator
{
$dom = new \DOMDocument;
@$dom->loadHTML(yield curl_init_with($url));
return new \DOMXPath($dom);
}
var_dump(Co::wait([
'Delay 5 secs' => function () {
echo "[Delay] I start to have a pseudo-sleep in this coroutine for about 5 secs\n";
for ($i = 0; $i < 5; ++$i) {
yield Co::DELAY => 1;
if ($i < 4) {
printf("[Delay] %s\n", str_repeat('.', $i + 1));
}
}
echo "[Delay] Done!\n";
},
"google.com HTML" => curl_init_with("https://google.com"),
"Content-Length of github.com" => function () {
echo "[GitHub] I start to request for github.com to calculate Content-Length\n";
$content = yield curl_init_with("https://github.com");
echo "[GitHub] Done! Now I calculate length of contents\n";
return strlen($content);
},
"Save mpyw's Gravatar Image URL to local" => function () {
echo "[Gravatar] I start to request for github.com to get Gravatar URL\n";
$src = (yield get_xpath_async('https://github.com/mpyw'))
->evaluate('string(//img[contains(@class,"avatar")]/@src)');
echo "[Gravatar] Done! Now I download its data\n";
yield curl_init_with($src, [CURLOPT_FILE => fopen('/tmp/mpyw.png', 'wb')]);
echo "[Gravatar] Done! Saved as /tmp/mpyw.png\n";
}
]));The requests are executed as parallelly as possible 😄
Note that there is only 1 process and 1 thread.
[Delay] I start to have a pseudo-sleep in this coroutine for about 5 secs
[GitHub] I start to request for github.com to calculate Content-Length
[Gravatar] I start to request for github.com to get Gravatar URL
[Delay] .
[Delay] ..
[GitHub] Done! Now I calculate length of contents
[Gravatar] Done! Now I download its data
[Delay] ...
[Gravatar] Done! Saved as /tmp/mpyw.png
[Delay] ....
[Delay] Done!
array(4) {
["Delay 5 secs"]=>
NULL
["google.com HTML"]=>
string(262) "<HTML><HEAD><meta http-equiv="content-type" content="text/html;charset=utf-8">
<TITLE>302 Moved</TITLE></HEAD><BODY>
<H1>302 Moved</H1>
The document has moved
<A HREF="https://www.google.co.jp/?gfe_rd=cr&ei=XXXXXX">here</A>.
</BODY></HTML>
"
["Content-Length of github.com"]=>
int(25534)
["Save mpyw's Gravatar Image URL to local"]=>
NULL
}
Install via Composer.
composer require mpyw/co:^1.5And require Composer autoloader in your scripts.
require __DIR__ . '/vendor/autoload.php';
use mpyw\Co\Co;
use mpyw\Co\CURLException;Wait for all the cURL requests to complete.
The options will override static defaults.
static Co::wait(mixed $value, array $options = []) : mixed(mixed)$value
Any values to be parallelly resolved.(array<string, mixed>)$options
Associative array of options.
| Key | Default | Description |
|---|---|---|
throw |
true |
Whether to throw or capture CURLException or RuntimeException on top-level. |
pipeline |
false |
Whether to use HTTP/1.1 pipelining. At most 5 requests for the same destination are bundled into single TCP connection. |
multiplex |
true |
Whether to use HTTP/2 multiplexing. All requests for the same destination are bundled into single TCP connection. |
autoschedule |
false |
Whether to use automatic scheduling by CURLMOPT_MAX_TOTAL_CONNECTIONS. |
interval |
0.002 |
curl_multi_select() timeout seconds. 0 means real-time observation. |
concurrency |
6 |
Limit of concurrent TCP connections. 0 means unlimited.The value should be within 10 at most. |
Throwablewhich are not extended fromRuntimeException, such asErrorExceptionLogicExceptionare not captured. If you need to capture them, you have to write your own try-catch blocks in your functions.- HTTP/1.1 pipelining can be used only if the TCP connection is already established and verified that uses keep-alive session. It means that the first bundle of HTTP/1.1 requests CANNOT be pipelined. You can use it from second
yieldinCo::wait()call. - To use HTTP/2 multiplexing, you have to build PHP with libcurl 7.43.0+ and
--with-nghttp2. - To use
autoschedule, PHP 7.0.7 or later is required.
When autoschedule Disabled:
curl_multi_add_handle()call can be delayed.concurrencycontrolling withpipeline/multiplexCANNOT be correctly driven. You should set higherconcurrencyin those cases.
When autoschedule Enabled:
curl_multi_add_handle()is always immediately called.CURLINFO_TOTAL_TIMECANNOT be correctly calculated. "Total Time" includes the time waiting for other requests are finished.
The details of CURLIFNO_*_TIME timing charts are described at the bottom of this page.
(mixed)
Resolved values; in exception-safe context, it may contain...
CURLExceptionwhich has been raised internally.RuntimeExceptionwhich has been raised by user.
- Throws
CURLExceptionorRuntimeExceptionin exception-unsafe context.
Execute cURL requests along with Co::wait() call, without waiting resolved values.
The options are inherited from Co::wait().
This method is mainly expected to be used ...
- When you are not interested in responses.
- In
CURLOPT_WRITEFUNCTIONorCURLOPT_HEADERFUNCTIONcallbacks.
static Co::async(mixed $value, mixed $throw = null) : null(mixed)$value
Any values to be parallelly resolved.(mixed)$throw
OverridesthrowinCo::wait()options when you passedtrueorfalse.
(null)
CURLExceptionorRuntimeExceptioncan be thrown in exception-unsafe context.
Note that you CANNOT capture top-level exceptions unless you catch outside ofCo::wait()call.
Return if Co::wait() is running().
With this check, you can safely call Co::wait() or Co::async().
static Co::isRunning() : boolReturn a Generator that resolves with specific value.
static Co::any(array $value) : \Generator<mixed>
static Co::race(array $value) : \Generator<mixed>
static Co::all(array $value) : \Generator<mixed>| Family | Return Value | Exception |
|---|---|---|
Co::any() |
First Success | AllFailedException |
Co::race() |
First Success | First Failure |
- Jobs CANNOT be canceled.
Incomplete jobs remain even ifCo::any()orCo::race()is resolved. Co::all(...)is just a wrapper of(function () { return yield ...; })().
It should be only used withCo::race()orCo::any().
Co::wait(function () {
$group1 = Co::all([$ch1, $ch2, $ch3]);
$group2 = Co::all([$ch4, $ch5, $ch6]);
$group1or2 = Co::any([$group1, $group2]);
var_dump(yield $group1or2);
});Overrides/gets static default settings.
static Co::setDefaultOptions(array $options) : null
static Co::getDefaultOptions() : arrayThe all yielded/returned values are resolved by the following rules.
Yielded values are also resent to the Generator.
The rules will be applied recursively.
| Before | After |
|---|---|
| cURL resource | curl_multi_getconent() result or CURLException |
| Array | Array (with resolved children) or RuntimeException |
| Generator Closure Generator |
Return value (after all yields done) or RuntimeException |
"Generator Closure" means Closure that contains yield keywords.
Exception-unsafe context by default.
The following yield statement specifies exception-safe context.
$results = yield Co::SAFE => [$ch1, $ch2];This is equivalent to:
$results = yield [
function () use ($ch1) {
try {
return yield $ch1;
} catch (\RuntimeException $e) {
return $e;
}
},
function () use ($ch2) {
try {
return yield $ch2;
} catch (\RuntimeException $e) {
return $e;
}
},
];Exception-unsafe context by default.
The following setting specifies exception-safe context.
$result = Co::wait([$ch1, $ch2], ['throw' => false]);This is equivalent to:
$results = Co::wait([
function () use ($ch1) {
try {
return yield $ch1;
} catch (\RuntimeException $e) {
return $e;
}
},
function () use ($ch2) {
try {
return yield $ch2;
} catch (\RuntimeException $e) {
return $e;
}
},
]);Contexts are inherited from Co::wait().
The following setting overrides parent context as exception-safe.
Co::async($value, false);The following setting overrides parent context as exception-unsafe.
Co::async($value, true);The following yield statements delay the coroutine processing:
yield Co::DELAY => $seconds
yield Co::SLEEP => $seconds # AliasPHP 7.0+:
yield $foo;
yield $bar;
return $baz;PHP 5.5~5.6:
yield $foo;
yield $bar;
yield Co::RETURN_WITH => $baz;Although experimental aliases Co::RETURN_ Co::RET Co::RTN are provided,
Co::RETURN_WITH is recommended in terms of readability.
PHP 7.0+:
$a = yield $foo;
echo yield $bar;PHP 5.5~5.6:
$a = (yield $foo);
echo (yield $bar);Be careful that return triggers finally while yield Co::RETURN_WITH => does not.
try {
return '...';
} finally {
// Reachable
}try {
yield Co::RETURN_WITH => '...';
} finally {
// Unreachable
}Note that S is equal to Q when autoschedule is disabled.
| ID | When |
|---|---|
| Q | curl_multi_exec() immediately after curl_multi_add_handle() called |
| S | Processing started actually |
| DNS | DNS resolution completed |
| TCP | TCP connection established |
| TLS | TLS/SSL session established |
| HS | All HTTP request headers sent |
| BS | Whole HTTP request body sent |
| HR | All HTTP response headers received |
| BR | Whole HTTP response body received |
| Constant | Time |
|---|---|
| CURLINFO_NAMELOOKUP_TIME | DNS - S |
| CURLINFO_CONNECT_TIME | TCP - S |
| CURLINFO_APPCONNECT_TIME | TLS - S |
| CURLINFO_PRETRANSFER_TIME | HS - S |
| CURLINFO_STARTTRANSFER_TIME | HR - S |
| CURLINFO_TOTAL_TIME | BR - Q |
| ID | When |
|---|---|
| Q | curl_multi_exec() immediately after curl_multi_add_handle() called |
| S | Processing started actually |
| DNS(1) | DNS resolution completed |
| TCP(1) | TCP connection established |
| TLS(1) | TLS/SSL session established |
| HS(1) | All HTTP request headers sent |
| BS(1) | Whole HTTP request body sent |
| HR(1) | All HTTP response headers received |
| DNS(2) | DNS resolution completed |
| TCP(2) | TCP connection established |
| TLS(2) | TLS/SSL session established |
| HS(2) | All HTTP request headers sent |
| BS(2) | Whole HTTP request body sent |
| HR(2) | All HTTP response headers received |
| BR(2) | Whole HTTP response body received |
| Constant | Time |
|---|---|
| CURLINFO_REDIRECT_TIME | HR(1) - Q |
| CURLINFO_NAMELOOKUP_TIME | DNS(2) - HR(1) |
| CURLINFO_CONNECT_TIME | TCP(2) - HR(1) |
| CURLINFO_APPCONNECT_TIME | TLS(2) - HR(1) |
| CURLINFO_PRETRANSFER_TIME | HS(2) - HR(1) |
| CURLINFO_STARTTRANSFER_TIME | HR(2) - HR(1) |
| CURLINFO_TOTAL_TIME | BR(2) - Q |
