--- AnyEvent-HTTP/HTTP.pm 2009/07/05 01:45:01 1.40 +++ AnyEvent-HTTP/HTTP.pm 2009/08/05 16:23:48 1.49 @@ -41,16 +41,16 @@ use strict; no warnings; -use Carp; +use Errno (); -use AnyEvent 4.452 (); +use AnyEvent 4.8 (); use AnyEvent::Util (); use AnyEvent::Socket (); use AnyEvent::Handle (); use base Exporter::; -our $VERSION = '1.12'; +our $VERSION = '1.42'; our @EXPORT = qw(http_get http_post http_head http_request); @@ -61,7 +61,7 @@ our $TIMEOUT = 300; # changing these is evil -our $MAX_PERSISTENT_PER_HOST = 2; +our $MAX_PERSISTENT_PER_HOST = 0; our $MAX_PER_HOST = 4; our $PROXY; @@ -96,9 +96,9 @@ object at least alive until the callback get called. If the object gets destroyed before the callbakc is called, the request will be cancelled. -The callback will be called with the response data as first argument -(or C if it wasn't available due to errors), and a hash-ref with -response headers as second argument. +The callback will be called with the response body data as first argument +(or C if an error occured), and a hash-ref with response headers as +second argument. All the headers in that hash are lowercased. In addition to the response headers, the "pseudo-headers" C, C and C @@ -110,8 +110,9 @@ joined together with a comma (C<,>), as per the HTTP spec. If an internal error occurs, such as not being able to resolve a hostname, -then C<$data> will be C, C<< $headers->{Status} >> will be C<599> -and the C pseudo-header will contain an error message. +then C<$data> will be C, C<< $headers->{Status} >> will be C<59x> +(usually C<599>) and the C pseudo-header will contain an error +message. A typical callback might look like this: @@ -139,7 +140,9 @@ The request headers to use. Currently, C may provide its own C, C, C and C headers -and will provide defaults for C and C. +and will provide defaults for C and C (this can be +suppressed by using C for these headers in which case they won't be +sent at all). =item timeout => $seconds @@ -151,7 +154,7 @@ Use the given http proxy for all requests. If not specified, then the default proxy (as specified by C<$ENV{http_proxy}>) is used. -C<$scheme> must be either missing or C for HTTP, or C for +C<$scheme> must be either missing, C for HTTP or C for HTTPS. =item body => $string @@ -187,6 +190,68 @@ The default for this option is C, which could be interpreted as "give me the page, no matter what". +=item on_header => $callback->($headers) + +When specified, this callback will be called with the header hash as soon +as headers have been successfully received from the remote server (not on +locally-generated errors). + +It has to return either true (in which case AnyEvent::HTTP will continue), +or false, in which case AnyEvent::HTTP will cancel the download (and call +the finish callback with an error code of C<598>). + +This callback is useful, among other things, to quickly reject unwanted +content, which, if it is supposed to be rare, can be faster than first +doing a C request. + +Example: cancel the request unless the content-type is "text/html". + + on_header => sub { + $_[0]{"content-type"} =~ /^text\/html\s*(?:;|$)/ + }, + +=item on_body => $callback->($partial_body, $headers) + +When specified, all body data will be passed to this callback instead of +to the completion callback. The completion callback will get the empty +string instead of the body data. + +It has to return either true (in which case AnyEvent::HTTP will continue), +or false, in which case AnyEvent::HTTP will cancel the download (and call +the completion callback with an error code of C<598>). + +This callback is useful when the data is too large to be held in memory +(so the callback writes it to a file) or when only some information should +be extracted, or when the body should be processed incrementally. + +It is usually preferred over doing your own body handling via +C, but in case of streaming APIs, where HTTP is +only used to create a connection, C is the better +alternative, as it allows you to install your own event handler, reducing +resource usage. + +=item want_body_handle => $enable + +When enabled (default is disabled), the behaviour of AnyEvent::HTTP +changes considerably: after parsing the headers, and instead of +downloading the body (if any), the completion callback will be +called. Instead of the C<$body> argument containing the body data, the +callback will receive the L object associated with the +connection. In error cases, C will be passed. When there is no body +(e.g. status C<304>), the empty string will be passed. + +The handle object might or might not be in TLS mode, might be connected to +a proxy, be a persistent connection etc., and configured in unspecified +ways. The user is responsible for this handle (it will not be used by this +module anymore). + +This is useful with some push-type services, where, after the initial +headers, an interactive protocol is used (typical example would be the +push-style twitter API which starts a JSON/XML stream). + +If you think you need this, first have a look at C, to see if +that doesn't solve your problem in a better way. + =back Example: make a simple HTTP GET request for http://www.nethype.de/ @@ -251,11 +316,11 @@ _slot_schedule $_[0]; } -our $qr_nl = qr<\015?\012>; -our $qr_nlnl = qr<\015?\012\015?\012>; +our $qr_nl = qr{\015?\012}; +our $qr_nlnl = qr{(? 1, dh => undef, sslv2 => 1 }; -our $TLS_CTX_HIGH = { cache => 1, verify => 1, verify_cn => "https", dh => "skip4096" }; +our $TLS_CTX_LOW = { cache => 1, sslv2 => 1 }; +our $TLS_CTX_HIGH = { cache => 1, verify => 1, verify_peername => "https" }; sub http_request($$@) { my $cb = pop; @@ -282,8 +347,6 @@ my $proxy = $arg{proxy} || $PROXY; my $timeout = $arg{timeout} || $TIMEOUT; - $hdr{"user-agent"} ||= $USERAGENT; - my ($uscheme, $uauthority, $upath, $query, $fragment) = $url =~ m|(?:([^:/?#]+):)?(?://([^/?#]*))?([^?#]*)(?:\?([^#]*))?(?:#(.*))?|; @@ -291,9 +354,7 @@ my $uport = $uscheme eq "http" ? 80 : $uscheme eq "https" ? 443 - : return $cb->(undef, { Status => 599, Reason => "Only http and https URL schemes supported (not '$uscheme')", URL => $url }); - - $hdr{referer} ||= "$uscheme://$uauthority$upath"; # leave out fragment and query string, just a heuristic + : return $cb->(undef, { Status => 599, Reason => "Only http and https URL schemes supported", URL => $url }); $uauthority =~ /^(?: .*\@ )? ([^\@:]+) (?: : (\d+) )?$/x or return $cb->(undef, { Status => 599, Reason => "Unparsable URL", URL => $url }); @@ -301,6 +362,8 @@ my $uhost = $1; $uport = $2 if defined $2; + $hdr{host} = defined $2 ? "$uhost:$2" : "$uhost"; + $uhost =~ s/^\[(.*)\]$/$1/; $upath .= "?$query" if length $query; @@ -342,6 +405,8 @@ if ($proxy) { ($rpath, $rhost, $rport, $rscheme) = ($url, @$proxy); + $rscheme = "http" unless defined $rscheme; + # don't support https requests over https-proxy transport, # can't be done with tls as spec'ed, unless you double-encrypt. $rscheme = "http" if $uscheme eq "https" && $rscheme eq "https"; @@ -349,7 +414,10 @@ ($rhost, $rport, $rscheme, $rpath) = ($uhost, $uport, $uscheme, $upath); } - $hdr{host} = $uhost; + # leave out fragment and query string, just a heuristic + $hdr{referer} ||= "$uscheme://$uauthority$upath" unless exists $hdr{referer}; + $hdr{"user-agent"} ||= $USERAGENT unless exists $hdr{"user-agent"}; + $hdr{"content-length"} = length $arg{body}; my %state = (connect_guard => 1); @@ -361,7 +429,12 @@ $state{connect_guard} = AnyEvent::Socket::tcp_connect $rhost, $rport, sub { $state{fh} = shift - or return (%state = (), $cb->(undef, { Status => 599, Reason => "$!", URL => $url })); + or do { + my $err = "$!"; + %state = (); + return $cb->(undef, { Status => 599, Reason => $err, URL => $url }); + }; + pop; # free memory, save a tree return unless delete $state{connect_guard}; @@ -404,7 +477,7 @@ # send request $state{handle}->push_write ( "$method $rpath HTTP/1.0\015\012" - . (join "", map "\u$_: $hdr{$_}\015\012", keys %hdr) + . (join "", map "\u$_: $hdr{$_}\015\012", grep defined $hdr{$_}, keys %hdr) . "\015\012" . (delete $arg{body}) ); @@ -425,7 +498,7 @@ # headers, could be optimized a bit $state{handle}->unshift_read (line => $qr_nlnl, sub { - for ("$_[1]\012") { + for ("$_[1]") { y/\015//d; # weed out any \015, as they show up in the weirdest of places. # things seen, not parsed: @@ -433,7 +506,7 @@ $hdr{lc $1} .= ",$2" while /\G - ([^:\000-\037]+): + ([^:\000-\037]*): [\011\040]* ((?: [^\012]+ | \012[\011\040] )*) \012 @@ -446,13 +519,45 @@ substr $_, 0, 1, "" for values %hdr; + # redirect handling + # microsoft and other shitheads don't give a shit for following standards, + # try to support some common forms of broken Location headers. + if ($hdr{location} !~ /^(?: $ | [^:\/?\#]+ : )/x) { + $hdr{location} =~ s/^\.\/+//; + + my $url = "$rscheme://$uhost:$uport"; + + unless ($hdr{location} =~ s/^\///) { + $url .= $upath; + $url =~ s/\/[^\/]*$//; + } + + $hdr{location} = "$url/$hdr{location}"; + } + + my $redirect; + + if ($recurse) { + if ($hdr{Status} =~ /^30[12]$/ && $method ne "POST") { + # apparently, mozilla et al. just change POST to GET here + # more research is needed before we do the same + $redirect = 1; + } elsif ($hdr{Status} == 303) { + # even http/1.1 is unclear on how to mutate the method + $method = "GET" unless $method eq "HEAD"; + $redirect = 1; + } elsif ($hdr{Status} == 307 && $method =~ /^(?:GET|HEAD)$/) { + $redirect = 1; + } + } + my $finish = sub { - $state{handle}->destroy; + $state{handle}->destroy if $state{handle}; %state = (); # set-cookie processing if ($arg{cookie_jar}) { - for ($hdr{"set-cookie"}) { + for ($_[1]{"set-cookie"}) { # parse NAME=VALUE my @kv; @@ -500,58 +605,77 @@ } } - # microsoft and other shitheads don't give a shit for following standards, - # try to support some common forms of broken Location headers. - if ($_[1]{location} !~ /^(?: $ | [^:\/?\#]+ : )/x) { - $_[1]{location} =~ s/^\.\/+//; - - my $url = "$rscheme://$uhost:$uport"; - - unless ($_[1]{location} =~ s/^\///) { - $url .= $upath; - $url =~ s/\/[^\/]*$//; - } - - $_[1]{location} = "$url/$_[1]{location}"; - } - - if ($_[1]{Status} =~ /^30[12]$/ && $recurse && $method ne "POST") { - # apparently, mozilla et al. just change POST to GET here - # more research is needed before we do the same - http_request ($method => $_[1]{location}, %arg, recurse => $recurse - 1, $cb); - } elsif ($_[1]{Status} == 303 && $recurse) { - # even http/1.1 is unclear on how to mutate the method - $method = "GET" unless $method eq "HEAD"; - http_request ($method => $_[1]{location}, %arg, recurse => $recurse - 1, $cb); - } elsif ($_[1]{Status} == 307 && $recurse && $method =~ /^(?:GET|HEAD)$/) { - http_request ($method => $_[1]{location}, %arg, recurse => $recurse - 1, $cb); + if ($redirect && exists $hdr{location}) { + # we ignore any errors, as it is very common to receive + # Content-Length != 0 but no actual body + # we also access %hdr, as $_[1] might be an erro + http_request ($method => $hdr{location}, %arg, recurse => $recurse - 1, $cb); } else { $cb->($_[0], $_[1]); } }; - if ($hdr{Status} =~ /^(?:1..|204|304)$/ or $method eq "HEAD") { - $finish->(undef, \%hdr); + my $len = $hdr{"content-length"}; + + if (!$redirect && $arg{on_header} && !$arg{on_header}(\%hdr)) { + $finish->(undef, { Status => 598, Reason => "Request cancelled by on_header", URL => $url }); + } elsif ( + $hdr{Status} =~ /^(?:1..|[23]04)$/ + or $method eq "HEAD" + or (defined $len && !$len) + ) { + # no body + $finish->("", \%hdr); } else { - if (exists $hdr{"content-length"}) { - $_[0]->unshift_read (chunk => $hdr{"content-length"}, sub { - # could cache persistent connection now - if ($hdr{connection} =~ /\bkeep-alive\b/i) { - # but we don't, due to misdesigns, this is annoyingly complex - }; + # body handling, four different code paths + # for want_body_handle, on_body (2x), normal (2x) + # we might read too much here, but it does not matter yet (no pers. connections) + if (!$redirect && $arg{want_body_handle}) { + $_[0]->on_eof (undef); + $_[0]->on_error (undef); + $_[0]->on_read (undef); + + $finish->(delete $state{handle}, \%hdr); - $finish->($_[1], \%hdr); - }); + } elsif ($arg{on_body}) { + $_[0]->on_error (sub { $finish->(undef, { Status => 599, Reason => $_[2], URL => $url }) }); + if ($len) { + $_[0]->on_eof (undef); + $_[0]->on_read (sub { + $len -= length $_[0]{rbuf}; + + $arg{on_body}(delete $_[0]{rbuf}, \%hdr) + or $finish->(undef, { Status => 598, Reason => "Request cancelled by on_body", URL => $url }); + + $len > 0 + or $finish->("", \%hdr); + }); + } else { + $_[0]->on_eof (sub { + $finish->("", \%hdr); + }); + $_[0]->on_read (sub { + $arg{on_body}(delete $_[0]{rbuf}, \%hdr) + or $finish->(undef, { Status => 598, Reason => "Request cancelled by on_body", URL => $url }); + }); + } } else { - # too bad, need to read until we get an error or EOF, - # no way to detect winged data. - $_[0]->on_error (sub { - # delete ought to be more efficient, as we would have to make - # a copy otherwise as $_[0] gets destroyed. - $finish->(delete $_[0]{rbuf}, \%hdr); - }); $_[0]->on_eof (undef); - $_[0]->on_read (sub { }); + + if ($len) { + $_[0]->on_error (sub { $finish->(undef, { Status => 599, Reason => $_[2], URL => $url }) }); + $_[0]->on_read (sub { + $finish->((substr delete $_[0]{rbuf}, 0, $len, ""), \%hdr) + if $len <= length $_[0]{rbuf}; + }); + } else { + $_[0]->on_error (sub { + $! == Errno::EPIPE || !$! + ? $finish->(delete $_[0]{rbuf}, \%hdr) + : $finish->(undef, { Status => 599, Reason => $_[2], URL => $url }); + }); + $_[0]->on_read (sub { }); + } } } }); @@ -624,17 +748,14 @@ The default value for the C header (the default is C). -=item $AnyEvent::HTTP::MAX_PERSISTENT - -The maximum number of persistent connections to keep open (default: 8). - -Not implemented currently. - -=item $AnyEvent::HTTP::PERSISTENT_TIMEOUT +=item $AnyEvent::HTTP::MAX_PER_HOST -The maximum time to cache a persistent connection, in seconds (default: 2). +The maximum number of concurrent connections to the same host (identified +by the hostname). If the limit is exceeded, then the additional requests +are queued until previous connections are closed. -Not implemented currently. +The default value for this is C<4>, and it is highly advisable to not +increase it. =item $AnyEvent::HTTP::ACTIVE