Update documentation:

* http parse functions
* Parser concept
* More detail on concepts
* Better hyperlinking
This commit is contained in:
Vinnie Falco
2017-07-20 08:12:07 -07:00
parent 0d69868051
commit da313a002e
12 changed files with 346 additions and 249 deletions
+50
View File
@@ -0,0 +1,50 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:Body Body]
In this table:
* `X` is a type meeting the requirements of [*`Body`].
[table Body requirements
[[operation] [type] [semantics, pre/post-conditions]]
[
[`X::value_type`]
[]
[
The type of the `message::body` member.
If this is not movable or not copyable, the containing message
will be not movable or not copyable.
]
]
[
[`X:value_type{}`]
[]
[`DefaultConstructible`]
]
[
[`Body::reader`]
[]
[
If present, a type meeting the requirements of
[link beast.types.Reader [*`Reader`]].
Provides an implementation to parse the body.
]
]
[
[`Body::writer`]
[]
[
If present, a type meeting the requirements of
[link beast.types.Writer [*`Writer`]].
Provides an implementation to serialize the body.
]
]
]
[endsect]
+15
View File
@@ -0,0 +1,15 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:BufferSequence BufferSequence]
A `BufferSequence` is a type meeting either of the following requirements:
* [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/ConstBufferSequence.html [*`ConstBufferSequence`]]
* [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/MutableBufferSequence.html [*`MutableBufferSequence`]]
[endsect]
+41
View File
@@ -0,0 +1,41 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:Field Field]
A [*`Field`] represents a single HTTP header field/value pair.
In this table:
* `X` denotes a type meeting the requirements of [*`Field`].
* `a` denotes a value of type `X`.
[table Field requirements
[[operation][type][semantics, pre/post-conditions]]
[
[`a.name()`]
[`boost::string_ref`]
[
This function returns a value implicitly convertible to
`boost::string_ref` containing the case-insensitve field
name, without leading or trailing white space.
]
]
[
[`a.value()`]
[`boost::string_ref`]
[
This function returns a value implicitly convertible to
`boost::string_ref` containing the value for the field. The
value is considered canonical if there is no leading or
trailing whitespace.
]
]
]
[endsect]
+51
View File
@@ -0,0 +1,51 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:FieldSequence FieldSequence]
A [*`FieldSequence`] is an iterable container whose value type meets
the requirements of [link beast.types.Field [*`Field`]].
In this table:
* `X` denotes a type that meets the requirements of [*`FieldSequence`].
* `a` is a value of type `X`.
[table FieldSequence requirements
[[operation][type][semantics, pre/post-conditions]]
[
[`X::value_type`]
[]
[
A type that meets the requirements of `Field`.
]
]
[
[`X::const_iterator`]
[]
[
A type that meets the requirements of `ForwardIterator`.
]
]
[
[`a.begin()`]
[`X::const_iterator`]
[
Returns an iterator to the beginning of the field sequence.
]
]
[
[`a.end()`]
[`X::const_iterator`]
[
Returns an iterator to the end of the field sequence.
]
]
]
[endsect]
+58
View File
@@ -0,0 +1,58 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:Parser Parser]
A [*`Parser`] is used to deserialize HTTP/1 messages from [link beast.types.streams streams].
Objects of this type are used with [link beast.ref.http__parse http::parse] and
[link beast.ref.http__async_parse http::async_parse].
In this table:
* `X` denotes a type meeting the requirements of [*`Parser`].
* `a` denotes a value of type `X`.
* `b` is a value meeting the requirements of [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/ConvertibleToConstBuffer.html [*`ConvertibleToConstBuffer`]].
* `ec` is a value of type [link beast.ref.error_code `error_code&`].
[table Parser requirements
[[operation] [type] [semantics, pre/post-conditions]]
[
[`a.complete()`]
[`bool`]
[
Returns `true` when a complete HTTP/1 message has been parsed.
]
]
[
[`a.write(b, ec)`]
[`std::size_t`]
[
Parses the octets in the specified input buffer sequentially until
an error occurs, the end of the buffer is reached, or a complete
HTTP/1 message has been parsed. If an error occurs, `ec` is set
to the error code and parsing stops. This function returns the
number of bytes consumed in the input buffer.
]
]
[
[`a.write_eof(ec)`]
[`void`]
[
Indicates to the parser that no more octets will be available.
Typically this function is called when the end of stream is reached.
For example, if a call to `boost::asio::ip::tcp::socket::read_some`
generates a `boost::asio::error::eof` error. Some HTTP/1 messages
determine the end of the message body by an end of file marker or
closing of the connection.
]
]
]
[endsect]
+54
View File
@@ -0,0 +1,54 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:Reader Reader]
Parser implementations will construct the corresponding `reader` object
during the parse. This customization point allows the Body to determine
the strategy for storing incoming message body data.
In this table:
* `X` denotes a type meeting the requirements of [*`Reader`].
* `a` denotes a value of type `X`.
* `p` is any pointer.
* `n` is a value convertible to `std::size_t`.
* `ec` is a value of type `error_code&`.
* `m` denotes a value of type `message const&` where
`std::is_same<decltype(m.body), Body::value_type>:value == true`
[table Reader requirements
[[operation] [type] [semantics, pre/post-conditions]]
[
[`X a(m);`]
[]
[
`a` is constructible from `m`. The lifetime of `m` is
guaranteed to end no earlier than after `a` is destroyed.
]
]
[
[`a.write(p, n, ec)`]
[`void`]
[
Deserializes the input sequence into the body.
If `ec` is set, the deserialization is aborted and the error
is returned to the caller.
]
]
]
[note Definitions for required `Reader` member functions should be declared
inline so the generated code becomes part of the implementation. ]
[endsect]
+96
View File
@@ -0,0 +1,96 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:Streambuf Streambuf]
A [*`Streambuf`] represents a logical octet sequence divided in two sections,
the input sequence and the output sequence. Octets are written to the output
sequence, then moved to the input sequence where they are available for
reading. When some or all of the input sequence is no longer needed, it may
be consumed.
The interface to this concept is intended to permit the following
implementation strategies:
* A single contiguous octet array, which is reallocated as necessary to
accommodate changes in the size of the octet sequence.
* A sequence of one or more octet arrays, where each array is of the same
size. Additional octet array objects are appended to the sequence to
accommodate changes in the size of the octet sequence.
* A sequence of one or more octet arrays of varying sizes. Additional octet
array objects are appended to the sequence to accommodate changes in the
size of the character sequence. This is the implementation approached
currently offered by [link beast.ref.basic_streambuf `basic_streambuf`].
In the table below:
* `X` denotes a class meeting the requirements of [*`Streambuf`]
* `a` denotes a value of type `X`
* `n` denotes a value convertible to `std::size_t`
* `U`, `T` denote unspecified types.
[table Streambuf requirements
[[operation] [type] [semantics, pre/post-conditions]]
[
[`X::const_buffers_type`]
[`T`]
[`T` meets the requirements for [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/ConstBufferSequence.html `ConstBufferSequence`].]
]
[
[`X::mutable_buffers_type`]
[`U`]
[`U` meets the requirements for [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/MutableBufferSequence.html `MutableBufferSequence`].]
]
[
[`a.commit(n)`]
[`void`]
[Moves bytes from the output sequence to the input sequence.]
]
[
[`a.consume(n)`]
[`void`]
[Removes bytes from the input sequence.]
]
[
[`a.data()`]
[`T`]
[Returns a list of buffers that represents the input sequence.]
]
[
[`a.prepare(n)`]
[`U`]
[Returns a list of buffers that represents the output sequence, with
the given size.]
]
[
[`a.size()`]
[`std::size_t`]
[Returns the size of the input sequence.]
]
[
[`a.max_size()`]
[`std::size_t`]
[Returns the maximum size of the stream buffer.]
]
[
[`read_size_helper(a, n)`]
[`std::size_t`]
[
Returns the suggested number of bytes to read into the output
sequence where `n` is an upper limit on this value. One possible
implementation is to return the number of bytes that may be prepared
without causing a dynamic allocation or `n`, whichever is smaller.
Calls to `read_size_helper` will be made without namespace
qualification, to allow the rules for argument dependent lookup to
take effect.
]
]
]
[endsect]
+34
View File
@@ -0,0 +1,34 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:streams Streams]
Stream types represent objects capable of performing synchronous or
asynchronous I/O. They are based on concepts from `boost::asio`.
[heading:Stream Stream]
A type modeling [*`Stream`] meets either or both of the following requirements:
* [link beast.types.streams.AsyncStream [*`AsyncStream`]]
* [link beast.types.streams.SyncStream [*`SyncStream`]]
[heading:AsyncStream AsyncStream]
A type modeling [*`AsyncStream`] meets the following requirements:
* [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/AsyncReadStream.html [*`AsyncReadStream`]]
* [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/AsyncWriteStream.html [*`AsyncWriteStream`]]
[heading:SyncStream SyncStream]
A type modeling [*`SyncStream`] meets the following requirements:
* [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/SyncReadStream.html [*`SyncReadStream`]]
* [@http://www.boost.org/doc/libs/1_60_0/doc/html/boost_asio/reference/SyncWriteStream.html [*`SyncWriteStream`]]
[endsect]
+179
View File
@@ -0,0 +1,179 @@
[/
Copyright (c) 2013-2016 Vinnie Falco (vinnie dot falco at gmail dot com)
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
]
[section:Writer Writer]
A `Writer` serializes the message body. The implementation creates an instance
of this type when serializing a message, and calls into it zero or more times
to provide buffers containing the data. The interface of `Writer` is intended
to allow serialization in these scenarios:
* A body that does not entirely fit in memory.
* A body produced incrementally from coroutine output.
* A body represented by zero or more buffers already in memory.
* A body as a series of buffers when the content size is not known ahead of time.
* Body data generated on demand from other threads.
* Body data computed algorithmically.
In this table:
* `X` denotes a type meeting the requirements of `Writer`.
* `a` denotes a value of type `X`.
* `m` denotes a value of type `message const&` where
`std::is_same<decltype(m.body), Body::value_type>:value == true`.
* `rc` is an object of type [link beast.ref.http__resume_context resume_context].
* `ec` is a value of type `error_code&`.
* `wf` is a [*write function]: a function object of unspecified type provided
by the implementation which accepts any value meeting the requirements
of `ConstBufferSequence` as its single parameter.
[table Writer requirements
[[operation] [type] [semantics, pre/post-conditions]]
[
[`X a(m);`]
[]
[
`a` is constructible from `m`. The lifetime of `m` is
guaranteed to end no earlier than after `a` is destroyed.
]
]
[
[`a.init(ec)`]
[`void`]
[
Called immediately after construction.
If `ec` is set, the serialization is aborted and the error
is propagated to the caller.
]
]
[
[`a.content_length()`]
[`std::uint64_t`]
[
If this member is present, it is called after initialization
and before calls to provide buffers. The serialized message will
have the Content-Length field set to the value returned from
this function. If this member is absent, the serialized message
body will be chunk-encoded for HTTP versions 1.1 and later, else
the serialized message body will be sent unmodified, with the
error `boost::asio::error::eof` returned to the caller, to notify
they should close the connection to indicate the end of the message.
]
]
[
[`a(rc, ec, wf)`]
[`boost::tribool`]
[
Called repeatedly after `init` succeeds.
`wf` is a function object which takes as its single parameter,
any value meeting the requirements of `ConstBufferSequence`.
Buffers provided by the `writer` to this [*write function] must
remain valid until the next member function of `writer` is
invoked (which may be the destructor). This function returns `true`
to indicate all message body data has been written, or `false`
if there is more body data. If the return value is
`boost::indeterminate`, the implementation will suspend the operation
until the writer invokes `rc`. It is the writers responsibility when
returning `boost::indeterminate`, to acquire ownership of the
`resume_context` via move construction and eventually call it or else
undefined behavior results.
]
]
]
[note Definitions for required `Writer` member functions should be declared
inline so the generated code becomes part of the implementation. ]
Exemplar:
```
struct writer
{
public:
/** Construct the writer.
The msg object is guaranteed to exist for the lifetime of the writer.
Exceptions:
No-throw guarantee.
@param msg The message whose body is to be written.
*/
template<bool isRequest, class Body, class Headers>
explicit
writer(message<isRequest, Body, Headers> const& msg);
/** Initialize the writer.
Called once immediately after construction.
The writer can perform initialization which may fail.
@param ec Contains the error code if any errors occur.
*/
void
init(error_code& ec);
/** Returns the content length.
If this member is present, the implementation will set the
Content-Length field accordingly. If absent, the implementation will
use chunk-encoding or terminate the connection to indicate the end
of the message.
*/
std::size_t
content_length() const;
/** Write zero or one buffer representing the message body.
Postconditions:
If return value is `true`:
* Callee does not take ownership of resume.
* Callee made zero or one calls to `write`.
* There is no more data remaining to write.
If return value is `false`:
* Callee does not take ownership of resume.
* Callee made one call to `write`.
If return value is boost::indeterminate:
* Callee takes ownership of `resume`.
* Caller suspends the write operation
until `resume` is invoked.
When the caller takes ownership of resume, the
asynchronous operation will not complete until the
caller destroys the object.
@param resume A functor to call to resume the write operation
after the writer has returned boost::indeterminate.
@param ec Set to indicate an error. This will cause an
asynchronous write operation to complete with the error.
@param write A functor the writer will call to provide the next
set of buffers. Ownership of the buffers is not transferred,
the writer must guarantee that the buffers remain valid until the
next member function is invoked, which may be the destructor.
@return `true` if there is data, `false` when done,
boost::indeterminate to suspend.
@note Undefined behavior if the callee takes ownership
of resume but does not return boost::indeterminate.
*/
template<class WriteFunction>
boost::tribool
operator()(resume_context&&, error_code&, WriteFunction&& write);
};
```
[endsect]