YAHC - Yet another HTTP client
use YAHC qw/yahc_reinit_conn/; my @hosts = ('www.booking.com', 'www.google.com:80'); my ($yahc, $yahc_storage) = YAHC->new({ host => \@hosts }); $yahc->request({ path => '/', host => 'www.reddit.com' }); $yahc->request({ path => '/', host => sub { 'www.reddit.com' } }); $yahc->request({ path => '/', host => \@hosts }); $yahc->request({ path => '/', callback => sub { ... } }); $yahc->request({ path => '/' }); $yahc->request({ path => '/', callback => sub { yahc_reinit_conn($_[0], { host => 'www.newtarget.com' }) if $_[0]->{response}{status} == 301; } }); $yahc->run;
YAHC is fast & minimal low-level asynchronous HTTP client intended to be used where you control both the client and the server. Is especially suits cases where set of requests need to be executed against group of machines.
It is NOT a general HTTP user agent, it doesn't support redirects, proxies and any number of other advanced HTTP features like (in roughly descending order of feature completeness) LWP::UserAgent, WWW::Curl, HTTP::Tiny, HTTP::Lite or Furl. This library is basically one step above manually talking HTTP over sockets.
YAHC supports SSL and socket reuse (latter is in experimental mode).
Each YAHC connection goes through following list of states in its lifetime:
+-----------------+ +<<-| INITALIZED <-<<+ v +-----------------+ ^ v | ^ v +-------v---------+ ^ +<<-+ RESOLVE DNS +->>+ v +-----------------+ ^ v | ^ v +-------v---------+ ^ +<<-+ CONNECTING +->>+ v +-----------------+ ^ v | ^ Path in v +-------v---------+ ^ Retry case of +<<-+ CONNECTED +->>+ logic failure v +-----------------+ ^ path v | ^ v +-------v---------+ ^ +<<-+ WRITING +->>+ v +-----------------+ ^ v | ^ v +-------v---------+ ^ +<<-+ READING +->>+ v +-----------------+ ^ v | ^ v +-------v---------+ ^ +>>-> USER ACTION +->>+ +-----------------+ | +-------v---------+ | COMPLETED | +-----------------+
There are three paths of workflow:
In normal situation a connection after being initialized goes through state:
- RESOLVE DNS (not implemented)
- CONNECTING - wait finishing of handshake
- CONNECTED
- WRITING - sending request body
- READING - awaiting and reading response
- USER ACTION - see below
- COMPLETED - all done, this is terminal state
SSL connection has extra state SSL_HANDSHAKE after CONNECTED state. State 'RESOLVE DNS' is not implemented yet.
In case of IO error during normal execution YAHC retries connection retries times. In practice this means that connection goes back to INITIALIZED state.
retries
If all retry attempts did not succeeded a connection goes to state 'USER ACTION' (see below).
'USER ACTION' state is called right before connection if going to enter 'COMPLETED' state (with either failed or successful results) and is meant to give a chance to user to interrupt the workflow.
'USER ACTION' state is entered in these circumstances:
HTTP response received. Note that non-200 responses are NOT treated as error.
unsupported HTTP response is received (such as response without Content-Length header)
retries limit reached
lifetime timeout has expired
provided callback has thrown exception
internal error has occured
When a connection enters this state callback CodeRef is called:
callback
$yahc->request({ ... callback => sub { my ( $conn, # connection 'object' $error, # one of YAHC::Error::* constants $strerror # string representation of error ) = @_; # Note that fields in $conn->{response} are not reliable # if $error != YAHC::Error::NO_ERROR() # HTTP response is stored in $conn->{response}. # It can be also accessed via yahc_conn_response(). my $response = $conn->{response}; my $status = $response->{status}; my $body = $response->{body}; } });
If there was no IO error yahc_conn_response return HashRef representing response. It contains the following key-value pairs.
yahc_conn_response
HashRef
proto => :Str status => :StatusCode body => :Str head => :HashRef
In case of a error or non-200 HTTP response yahc_retry_conn or yahc_reinit_conn may be called to give the request more chances to complete successfully (for example by following redirects or providing new target hosts). Also, note that in case of a error data returned by yahc_conn_response cannot be trusted. For example, if an IO error happened during receiving HTTP body headers would state 200 response code.
yahc_retry_conn
yahc_reinit_conn
YAHC lowercases headers names returned in head. This is done to comply with RFC which identify HTTP headers as case-insensitive.
head
In some cases connection cannot be retried anymore and callback is called for information purposes only. This case can be distinguished by $error having YAHC::Error::TERMINAL_ERROR() bit set. One can use yahc_terminal_error helper to detect such case.
$error
yahc_terminal_error
Note that callback should NOT throw exception. If so the connection will be immediately closed.
This method creates YAHC object and accompanying storage object:
my ($yahc, $yahc_storage) = YAHC->new();
This is a radical way of solving all possible memleak because of cyclic references in callbacks. Since all references of callbacks are kept in $yahc_storage object it's fine to use YAHC object inside request callback:
my $yahc->request({ callback => sub { $yahc->stop; # this is fine!!! }, });
However, user has to guarantee that both $yahc and $yahc_storage objects are kept in the same scope. So, they will be destroyed at the same time.
new can be passed with all parameters supported by request. They will be inherited by all requests.
new
request
Additionally, new supports three parameters: socket_cache, account_for_signals, and loop.
socket_cache
account_for_signals
loop
socket_cache option controls socket reuse logic. By default socket cache is disabled. If user wants YAHC reuse sockets he should set socket_cache to a HashRef.
my ($yahc, $yahc_storage) = YAHC->new({ socket_cache => {} });
In this case YAHC maintains unused sockets keyed on join($;, $$, $host, $port, $scheme). We use $; so we can use the $socket_cache->{$$, $host, $port, $scheme} idiom to access the cache.
join($;, $$, $host, $port, $scheme)
$;
$socket_cache->{$$, $host, $port, $scheme}
It's up to user to control the cache. It's also up to user to set necessary request headers for keep-alive. YAHC does not cache socket in cases of an error, HTTP/1.0 and when server explicitly instructs to close connection (i.e. header 'Connection' = 'close').
By default, each YAHC object will use its own EV eventloop. This is normally preferred since it allows for more accurate timing metrics.
However, if the process is already using an eventloop, having an inner loop means the outer one stays waiting until the inner one is done.
To get around this, one can specify the eventloop that YAHC will use:
my ($yahc, $storage) = YAHC->new({ loop => EV::default_loop(), # use the default EV eventloop });
Using the above, YAHC will be sharing the same eventloop as everyone else, so some operations are now riskier and should be avoided; For example, in most scenarios account_for_signals should not be used alongside loop, as only whatever is entering the eventloop should set the signal handlers.
Another parameter account_for_signals requires special attention! Here is why:
excerpt from EV documentation http://search.cpan.org/~mlehmann/EV-4.22/EV.pm#PERL_SIGNALS
While Perl signal handling (%SIG) is not affected by EV, the behaviour with EV is as the same as any other C library: Perl-signals will only be handled when Perl runs, which means your signal handler might be invoked only the next time an event callback is invoked.
In practise this means that none of set %SIG handlers will be called until EV calls one of perl callbacks. Which, in some cases, may take a long time. By setting account_for_signals YAHC adds EV::check watcher with empty callback effectively making EV calling the callback on every iteration. The trickery comes at some performance cost. This is what EV documentation says about it:
EV::check
... you can also force a watcher to be called on every event loop iteration by installing a EV::check watcher. This ensures that perl gets into control for a short time to handle any pending signals, and also ensures (slightly) slower overall operation.
So, if your code or the codes surrounding your code use %SIG handlers it's wise to set account_for_signals.
protocol => "HTTP/1.1", # (or "HTTP/1.0") scheme => "http" or "https" host => see below, port => ..., method => "GET", path => "/", query_string => "", head => [], body => "", # timeouts connect_timeout => undef, request_timeout => undef, drain_timeout => undef, lifetime_timeout => undef, # burst control backoff_delay => undef, # callbacks init_callback => undef, connecting_callback => undef, connected_callback => undef, writing_callback => undef, reading_callback => undef, callback => undef, # SSL options ssl_options => {},
Notice how YAHC does not take a full URI string as input, you have to specify the individual parts of the URL. Users who need to parse an existing URI string to produce a request should use the URI module to do so.
For example, to send a request to http://example.com/flower?color=red, pass the following parameters:
http://example.com/flower?color=red
$yach->request({ host => "example.com", port => "80", path => "/flower", query_string => "color=red" });
YAHC doesn't escape any values for you, it just passes them through as-is. You can easily produce invalid requests if e.g. any of these strings contain a newline, or aren't otherwise properly escaped.
Notice that you do not need to put the leading "?" character in the query_string. You do, however, need to properly uri_escape the content of query_string.
"?"
query_string
uri_escape
The value of head is an ArrayRef of key-value pairs instead of a HashRef, this way you can decide in which order the headers are sent, and you can send the same header name multiple times. For example:
ArrayRef
head => [ "Content-Type" => "application/json", "X-Requested-With" => "YAHC", ]
Will produce these request headers:
Content-Type: application/json X-Requested-With: YAHC
host parameter can accept one of following values:
host
1) string - represents target host. String may have following formats: hostname:port, ip:port. 2) ArrayRef of strings - YAHC will cycle through items selecting new host for each attempt. 3) CodeRef. The subroutine is invoked for each attempt and should at least return a string (hostname or IP address). It can also return array containing: ($host, $ip, $port, $scheme). This option effectively give a user control over host selection for retries. The CodeRef is passed with connection "object" which can be fed to yahc_conn_* family of functions.
The value of connect_timeout, request_timeout and drain_timeout is in floating point seconds, and is used as the time limit for connecting to the host (reaching CONNECTED state), full request time (reaching COMPLETED state) and sending request to remote site (reaching READING state) respectively.
connect_timeout
request_timeout
drain_timeout
lifetime_timeout has special purpose. Its task is to provide upper bound timeout for a request lifetime. In other words, if a request comes with multiple retries connect_timeout, request_timeout and drain_timeout are per attempt. lifetime_timeout covers all attempts. If by the time lifetime_timeout expires a connection is not in COMPLETED state a error is generated. Note that after this error the connection cannot be retried anymore. So, it's forced to go to COMPLETED state.
lifetime_timeout
The default value for all is undef, meaning no timeout limit.
undef
backoff_delay can be used to introduce delay between retries. This is a great way to avoid load spikes on server side. Following example creates new request which would be retried twice doing three attempts in total. Second and third attempts will be delay by one second each.
backoff_delay
$yach->request({ host => "example.com", retries => 2, backoff_delay => 1, });
backoff_delay can be set in two ways:
1) floating point seconds - define constant delay between retires. 2) CodeRef. The subroutine is invoked on each retry and should return floating point seconds. This option is useful for having exponentially growing delay or, for instance, jitted delays.
The default value is undef, meaning no delay.
The value of init_callback, connecting_callback, connected_callback, writing_callback, reading_callback is a reference to a subroutine which is called upon reaching corresponding state. Any exception thrown in the subroutine will be ignored.
init_callback
connecting_callback
connected_callback
writing_callback
reading_callback
The value of callback defines main request callback which is called when a connection enters 'USER ACTION' state (see 'USER ACTION' state above).
Also see LIMITATIONS
Performing HTTPS requires the value of ssl_options extended by two parameters set to current hostname:
ssl_options
SSL_verifycn_name => $hostname, IO::Socket::SSL->can_client_sni ? ( SSL_hostname => $hostname ) : (),
Apart of this changes, the value is directly passed to IO::Socket::SSL::start_SSL(). For more details refer to IO::Socket::SSL documentation https://metacpan.org/pod/IO::Socket::SSL.
IO::Socket::SSL::start_SSL()
Given connection HashRef or conn_id move connection to COMPLETED state (avoiding 'USER ACTION' state) and drop it from internal pool. The function takes two parameters: first is either a connection id or connection HashRef. Second one is a boolean flag indicating whether connection's socket should closed or it might be reused.
Start YAHC's loop. The loop stops when all connection complete.
Note that run can accept two extra parameters: until_state and list of connections. These two parameters tell YAHC to break the loop once specified connections reach desired state.
run
For example:
$yahc->run(YAHC::State::READING(), $conn_id);
Will loop until connection '$conn_id' move to state READING meaning that the data has been sent to remote side. In order to gather response one should later call:
$yahc->run(YAHC::State::COMPLETED(), $conn_id);
or simply:
$yahc->run();
Leaving list of connection empty makes YAHC waiting for all connection reaching needed until_state.
Note that waiting one particular connection to finish doesn't mean that others are not executed. Instead, all active connections are looped at the same time, but YAHC breaks the loop once waited connection reaches needed state.
Same as run but with EV::RUN_ONCE set. For more details check https://metacpan.org/pod/EV
Same as run but with EV::RUN_NOWAIT set. For more details check https://metacpan.org/pod/EV
Return true if YAHC is running, false otherwise.
Return underlying EV loop object.
Break running EV loop if any.
yahc_reinit_conn reinitialize given connection. The attempt counter is reset to 0. The function accepts HashRef as second argument. By passing it one can change host, port, scheme, body, head and others parameters. The format and meaning of these parameters is same as in request method.
One of use cases of yahc_reinit_conn, for example, is to handle redirects:
use YAHC qw/yahc_reinit_conn/; my ($yahc, $yahc_storage) = YAHC->new(); $yahc->request({ host => 'domain_which_returns_301.com', callback => sub { ... my $conn = $_[0]; yahc_reinit_conn($conn, { host => 'www.newtarget.com' }) if $_[0]->{response}{status} == 301; ... } }); $yahc->run;
yahc_reinit_conn is meant to be called inside callback i.e. when connection is in 'USER ACTION' state.
Retries given connection. yahc_retry_conn should be called only if yahc_conn_attempts_left returns positive value. Otherwise, it exits silently. The function accepts HashRef as second argument. By passing it one can change backoff_delay parameter. See docs for request for more details about backoff_delay.
yahc_conn_attempts_left
Intended usage is to retry transient failures or to try different host:
use YAHC qw/ yahc_retry_conn yahc_conn_attempts_left /; my ($yahc, $yahc_storage) = YAHC->new(); $yahc->request({ retries => 2, host => [ 'host1', 'host2' ], callback => sub { ... my $conn = $_[0]; if ($_[0]->{response}{status} == 503 && yahc_conn_attempts_left($conn)) { yahc_retry_conn($conn); return; } ... } }); $yahc->run;
yahc_retry_conn is meant to be called inside callback similarly to yahc_reinit_conn.
Return id of given connection.
Return state of given connection.
Return selected host and port for current attempt for given connection. Format "host:port". Default port values are omitted.
Same as yahc_conn_target but return full URL
yahc_conn_target
Let user assosiate arbitrary data with a connection. Be aware of not creating cyclic reference!
Return errors appeared in given connection. Note that the function returns all errors, not only ones happened during current attempt. Returned value is ArrayRef of ArrayRefs. Later one represents a error and contains following items:
error number (see YAHC::Error constants) error string ArrayRef of host, ip, port, scheme time when the error happened attempt when the error happened
yahc_conn_register_error adds new record in connection's error list. This functions is used internally for keeping track of all low-level errors during connection's lifetime. It can be also used by users for high-level errors such as 50x responses. The function takes $conn, $error which is one of YAHC::Error constants and error description. Error description can be passed in sprintf manner. For example:
yahc_conn_register_error
$conn
YAHC::Error
$yahc->request({ ... callback => sub { ... my $conn = $_[0]; my $status = $conn->{response}{status} || 0; if ($status == 503 || $status == 504) { yahc_conn_register_error( $conn, YAHC::Error::RESPONSE_ERROR(), "server returned %d", $status ); yahc_retry_conn($conn); return; } ... } });
Return last error appeared in connection. See yahc_conn_errors.
yahc_conn_errors
Given a error return 1 if the error has YAHC::Error::TERMINAL_ERROR() bit set. Otherwise return 0.
Return timeline of given connection. See more about timeline in description of new method.
Return request of given connection. See request.
Return response of given connection. See request.
Return current attempt starting from 1. The function can also return 0 if no attempts were made yet.
Return number of attempts left.
Return socket_cache id for given connection. Should be used to generate key for socket_cache. If connection is not initialized yet undef is returned.
YAHC provides set of constants for errors. Each constant returns bitmask which can be used to detect presence of a particular error, for example, in callback. There is one exception: YAHC::Error::NO_ERROR() return 0 indicating no error during request execution.
Error handling code can look like following:
$yahc->request({ ... callback => sub { my ( $conn, # connection 'object' $error, # one of YAHC::Error::* constants $strerror # string representation of error ) = @_; if ($error & YAHC::Error::TIMEOUT()) { # A timeout has happend. Use one of YAHC::Error::*_TIMEOUT() # constants for more clarification } elsif ($error & YAHC::Error::SSL_ERROR()) { # We had some issues with SSL. $error might have # YAHC::Error::READ_ERROR() or YAHC::Error::WRITE_ERROR() # indicating whether is was read or write error. } elsif (...) { # etc } } });
The list of error constants. The names are self-explanatory in many cases:
YAHC::Error::NO_ERROR()
Return value 0 (not a bitmask)> meaning no error
YAHC::Error::REQUEST_TIMEOUT()
YAHC::Error::CONNECT_TIMEOUT()
YAHC::Error::DRAIN_TIMEOUT()
YAHC::Error::LIFETIME_TIMEOUT()
YAHC::Error::TIMEOUT()
YAHC::Error::RETRY_LIMIT()
The connection has exhausted all available retries. This error is usually returned to callback. Check connection's errors via yahc_conn_errors to inspect the reasons of failures for each individual attempt.
YAHC::Error::CONNECT_ERROR()
YAHC::Error::READ_ERROR()
YAHC::Error::WRITE_ERROR()
YAHC::Error::SSL_ERROR()
YAHC::Error::REQUEST_ERROR()
not used
YAHC::Error::RESPONSE_ERROR()
Server returned unparsable response
YAHC::Error::CALLBACK_ERROR()
Usually represents exception in one of the callbacks
YAHC::Error::TERMINAL_ERROR()
This bit is set when connection cannot be retried anymore and is forced to complete
YAHC::Error::INTERNAL_ERROR()
https://github.com/ikruglov/YAHC
Note that YAHC has astonishing reduction in performance if any parameters participating in building HTTP message has UTF8 flag set. Those fields are protocol, host, port, method, path, query_string, head, body and maybe others.
protocol
port
method
path
body
Just one example (check scripts/utf8_test.pl for code). Simple HTTP request with 10MB of payload:
elapsed without utf8 flag: 0.039s elapsed with utf8 flag: 0.540s
Because of this YAHC warns if detected UTF8-flagged payload. The user needs to make sure that *all* data passed to YAHC is unflagged binary strings.
State 'RESOLVE DNS' is not implemented yet.
Ivan Kruglov <ivan.kruglov@yahoo.com>
Copyright (c) 2013-2017 Ivan Kruglov <ivan.kruglov@yahoo.com>.
<ivan.kruglov@yahoo.com>
This module derived lots of ideas, code and docs from Hijk https://github.com/gugod/Hijk. This module was originally developed for Booking.com.
The MIT License
BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR, OR CORRECTION.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
To install YAHC, copy and paste the appropriate command in to your terminal.
cpanm
cpanm YAHC
CPAN shell
perl -MCPAN -e shell install YAHC
For more information on module installation, please visit the detailed CPAN module installation guide.