Net::HTTP(3)

📦
version 6.24 · 12 versions
, in fedora 43

💖 Net::HTTP — the low-level HTTP client 💖

okay bestie, sit down, hydrate, because we're doing Net::HTTP, a page from the User Contributed Perl Documentation section (that's the 3 in Net::HTTP(3), iykyk ✨). It's a manual page and it's giving very "I talk to servers directly and I'm not sorry" energy. Let's get into it 💅

✨ VERSION ✨

version 6.24 — that's the one, no cap 🌸

🦋 SYNOPSIS 🦋

okay so here's the whole vibe in code form, copy it exactly bestie:

use Net::HTTP;
my $s = Net::HTTP->new(Host => "www.perl.com") || die $@;
$s->write_request(GET => "/", 'User-Agent' => "Mozilla/5.0");
my($code, $mess, %h) = $s->read_response_headers;

while (1) {
   my $buf;
   my $n = $s->read_entity_body($buf, 1024);
   die "read failed: $!" unless defined $n;
   last unless $n;
   print $buf;
}

look at her. iconic. she connects, she asks, she reads, she loops. periodt 🔥

🌈 DESCRIPTION 🌈

so the Net::HTTP class is a low-level HTTP client. One instance of Net::HTTP = one connection to an HTTP server. The HTTP protocol itself is described in RFC 2616, and this class supports HTTP/1.0 and HTTP/1.1 🫶

Also, plot twist: Net::HTTP is a sub-class of one of IO::Socket::IP (IPv6+IPv4), IO::Socket::INET6 (IPv6+IPv4), or IO::Socket::INET (IPv4 only). So you can mix the methods below with reading and writing from the socket directly. That said — and I need you to hear me — this is not necessary a good idea, unless you know what you are doing 😭 (and no, bestie, "I read a blog post once" does not count as knowing what you are doing).

The following methods are provided (in addition to those of IO::Socket::INET):

  • $s = Net::HTTP->new( %options ) — the Net::HTTP constructor method. It takes the same options as IO::Socket::INET's, plus these:
 Host:            Initial host attribute value
 KeepAlive:       Initial keep_alive attribute value
 SendTE:          Initial send_te attribute_value
 HTTPVersion:     Initial http_version attribute value
 PeerHTTPVersion: Initial peer_http_version attribute value
 MaxLineLength:   Initial max_line_length attribute value
 MaxHeaderLines:  Initial max_header_lines attribute value

The Host option is also the default for IO::Socket::INET's PeerAddr. The PeerPort defaults to 80 if you don't provide it. And the PeerPort can also be embedded in the PeerAddr by preceding it with a :, closing the IPv6 address on brackets [] if necessary: "192.0.2.1:80", "[2001:db8::1]:80", "any.example.com:80" 💖

The Listen option provided by IO::Socket::INET's constructor method is not allowed. Nope. Not the vibe. 🙈

If it's unable to connect to the given HTTP server, then the constructor returns undef and $@ contains the reason. After a successful connect, a Net:HTTP object is returned 👑

  • $s->host — get/set the default value of the Host header to send. The $host must not be set to an empty string (or undef) for HTTP/1.1.
  • $s->keep_alive — get/set the keep-alive value. If this value is TRUE then the request will be sent with headers indicating that the server should try to keep the connection open so that multiple requests can be sent.

The actual headers set will depend on the value of the http_version and peer_http_version attributes.

  • $s->send_te — get/set the value indicating if the request will be sent with a TE header, to indicate the transfer encodings that the server can choose to use. The list of encodings announced as accepted by this client depends on availability of the following modules: Compress::Raw::Zlib for deflate, and IO::Compress::Gunzip for gzip 🎀
  • $s->http_version — get/set the HTTP version number that this client should announce. This value can only be set to "1.0" or "1.1". The default is "1.1".
  • $s->peer_http_version — get/set the protocol version number of our peer. This value will initially be "1.0", but will be updated by a successful read_response_headers() method call.
  • $s->max_line_length — get/set a limit on the length of response line and response header lines. The default is 8192. A value of 0 means no limit 🦋
  • $s->max_header_length — get/set a limit on the number of header lines that a response can have. The default is 128. A value of 0 means no limit 🌸
  • $s->format_request($method, $uri, %headers, [$content]) — format a request message and return it as a string. If the headers do not include a Host header, then a header is inserted with the value of the host attribute. Headers like Connection and Keep-Alive might also be added depending on the status of the keep_alive attribute.

If $content is given (and it is non-empty), then a Content-Length header is automatically added unless it was already present ✨

  • $s->write_request($method, $uri, %headers, [$content]) — format and send a request message. Arguments are the same as for format_request(). Returns true if successful.
  • $s->format_chunk( $data ) — returns the string to be written for the given chunk of data.
  • $s->write_chunk($data) — will write a new chunk of request entity body data. This method should only be used if the Transfer-Encoding header with a value of chunked was sent in the request. Note, writing zero-length data is a no-op. Use the write_chunk_eof() method to signal end of entity body data.

Returns true if successful.

  • $s->format_chunk_eof( %trailers ) — returns the string to be written for signaling EOF when a Transfer-Encoding of chunked is used.
  • $s->write_chunk_eof( %trailers ) — will write eof marker for chunked data and optional trailers. Note that trailers should not really be used unless is was signaled with a Trailer header.

Returns true if successful.

  • ($code, $mess, %headers) = $s->read_response_headers( %opts ) — read response headers from server and return it. The $code is the 3 digit HTTP status code (see HTTP::Status) and $mess is the textual message that came with it. Headers are then returned as key/value pairs. Since key letter casing is not normalized and the same key can even occur multiple times, assigning these values directly to a hash is not wise 😭. Only the $code is returned if this method is called in scalar context.

As a side effect this method updates the peer_http_version attribute.

Options might be passed in as key/value pairs. There are currently only two options supported: "laxed" and "junk_out".

The "laxed" option will make read_response_headers() more forgiving towards servers that have not learned how to speak HTTP properly 🙈. The "laxed" option is a boolean flag, and is enabled by passing in a TRUE value. The "junk_out" option can be used to capture bad header lines when "laxed" is enabled. The value should be an array reference. Bad header lines will be pushed onto the array.

The "laxed" option must be specified in order to communicate with pre-HTTP/1.0 servers that don't describe the response outcome or the data they send back with a header block. For these servers peer_http_version is set to "0.9" and this method returns (200, "Assumed OK").

The method will raise an exception (die) if the server does not speak proper HTTP or if the max_line_length or max_header_length limits are reached. If the "laxed" option is turned on and max_line_length and max_header_length checks are turned off, then no exception will be raised and this method will always return a response code 👑

  • $n = $s->read_entity_body($buf, $size); — reads chunks of the entity body content. Basically the same interface as for read() and sysread(), but the buffer offset argument is not supported yet. This method should only be called after a successful read_response_headers() call.

The return value will be undef on read errors, 0 on EOF, -1 if no data could be returned this time, otherwise the number of bytes assigned to $buf. The $buf is set to "" when the return value is -1.

You normally want to retry this call if this function returns either -1 or undef with $! as EINTR or EAGAIN (see Errno). EINTR can happen if the application catches signals and EAGAIN can happen if you made the socket non-blocking.

This method will raise exceptions (die) if the server does not speak proper HTTP. This can only happen when reading chunked data 🍓

  • %headers = $s->get_trailers — after read_entity_body() has returned 0 to indicate end of the entity body, you might call this method to pick up any trailers.
  • $s->_rbuf — get/set the read buffer content. The read_response_headers() and read_entity_body() methods use an internal buffer which they will look for data before they actually sysread more from the socket itself. If they read too much, the remaining data will be left in this buffer.
  • $s->_rbuf_length — returns the number of bytes in the read buffer. This should always be the same as:
    length($s->_rbuf)

but might be more efficient 💅

🧚 SUBCLASSING 🧚

The read_response_headers() and read_entity_body() will invoke the sysread() method when they need more data. Subclasses might want to override this method to control how reading takes place.

The object itself is a glob. Subclasses should avoid using hash key names prefixed with http_ and io_ ⭐

💜 SEE ALSO 💜

LWP, IO::Socket::INET, Net::HTTP::NB

🌸 AUTHOR 🌸

Gisle Aas <gisle@activestate.com>

This software is copyright (c) 2001 by Gisle Aas.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself 🫶

and that's the whole page bestie!! 💖 this one came out of perl v5.42.0 and was dated 2025-09-01, so if you're reading it way later, hi, you're a time traveler, slay 🌈✨ bye!! 👑

📦 where this page comes from