]> git.proxmox.com Git - ceph.git/blob - ceph/src/boost/libs/beast/test/bench/parser/nodejs-parser/README.md
update sources to v12.2.3
[ceph.git] / ceph / src / boost / libs / beast / test / bench / parser / nodejs-parser / README.md
1 HTTP Parser
2 ===========
3
4 [![Build Status](https://api.travis-ci.org/nodejs/http-parser.svg?branch=master)](https://travis-ci.org/nodejs/http-parser)
5
6 This is a parser for HTTP messages written in C. It parses both requests and
7 responses. The parser is designed to be used in performance HTTP
8 applications. It does not make any syscalls nor allocations, it does not
9 buffer data, it can be interrupted at anytime. Depending on your
10 architecture, it only requires about 40 bytes of data per message
11 stream (in a web server that is per connection).
12
13 Features:
14
15 * No dependencies
16 * Handles persistent streams (keep-alive).
17 * Decodes chunked encoding.
18 * Upgrade support
19 * Defends against buffer overflow attacks.
20
21 The parser extracts the following information from HTTP messages:
22
23 * Header fields and values
24 * Content-Length
25 * Request method
26 * Response status code
27 * Transfer-Encoding
28 * HTTP version
29 * Request URL
30 * Message body
31
32
33 Usage
34 -----
35
36 One `http_parser` object is used per TCP connection. Initialize the struct
37 using `http_parser_init()` and set the callbacks. That might look something
38 like this for a request parser:
39 ```c
40 http_parser_settings settings;
41 settings.on_url = my_url_callback;
42 settings.on_header_field = my_header_field_callback;
43 /* ... */
44
45 http_parser *parser = malloc(sizeof(http_parser));
46 http_parser_init(parser, HTTP_REQUEST);
47 parser->data = my_socket;
48 ```
49
50 When data is received on the socket execute the parser and check for errors.
51
52 ```c
53 size_t len = 80*1024, nparsed;
54 char buf[len];
55 ssize_t recved;
56
57 recved = recv(fd, buf, len, 0);
58
59 if (recved < 0) {
60 /* Handle error. */
61 }
62
63 /* Start up / continue the parser.
64 * Note we pass recved==0 to signal that EOF has been received.
65 */
66 nparsed = http_parser_execute(parser, &settings, buf, recved);
67
68 if (parser->upgrade) {
69 /* handle new protocol */
70 } else if (nparsed != recved) {
71 /* Handle error. Usually just close the connection. */
72 }
73 ```
74
75 HTTP needs to know where the end of the stream is. For example, sometimes
76 servers send responses without Content-Length and expect the client to
77 consume input (for the body) until EOF. To tell http_parser about EOF, give
78 `0` as the fourth parameter to `http_parser_execute()`. Callbacks and errors
79 can still be encountered during an EOF, so one must still be prepared
80 to receive them.
81
82 Scalar valued message information such as `status_code`, `method`, and the
83 HTTP version are stored in the parser structure. This data is only
84 temporally stored in `http_parser` and gets reset on each new message. If
85 this information is needed later, copy it out of the structure during the
86 `headers_complete` callback.
87
88 The parser decodes the transfer-encoding for both requests and responses
89 transparently. That is, a chunked encoding is decoded before being sent to
90 the on_body callback.
91
92
93 The Special Problem of Upgrade
94 ------------------------------
95
96 HTTP supports upgrading the connection to a different protocol. An
97 increasingly common example of this is the WebSocket protocol which sends
98 a request like
99
100 GET /demo HTTP/1.1
101 Upgrade: WebSocket
102 Connection: Upgrade
103 Host: example.com
104 Origin: http://example.com
105 WebSocket-Protocol: sample
106
107 followed by non-HTTP data.
108
109 (See [RFC6455](https://tools.ietf.org/html/rfc6455) for more information the
110 WebSocket protocol.)
111
112 To support this, the parser will treat this as a normal HTTP message without a
113 body, issuing both on_headers_complete and on_message_complete callbacks. However
114 http_parser_execute() will stop parsing at the end of the headers and return.
115
116 The user is expected to check if `parser->upgrade` has been set to 1 after
117 `http_parser_execute()` returns. Non-HTTP data begins at the buffer supplied
118 offset by the return value of `http_parser_execute()`.
119
120
121 Callbacks
122 ---------
123
124 During the `http_parser_execute()` call, the callbacks set in
125 `http_parser_settings` will be executed. The parser maintains state and
126 never looks behind, so buffering the data is not necessary. If you need to
127 save certain data for later usage, you can do that from the callbacks.
128
129 There are two types of callbacks:
130
131 * notification `typedef int (*http_cb) (http_parser*);`
132 Callbacks: on_message_begin, on_headers_complete, on_message_complete.
133 * data `typedef int (*http_data_cb) (http_parser*, const char *at, size_t length);`
134 Callbacks: (requests only) on_url,
135 (common) on_header_field, on_header_value, on_body;
136
137 Callbacks must return 0 on success. Returning a non-zero value indicates
138 error to the parser, making it exit immediately.
139
140 For cases where it is necessary to pass local information to/from a callback,
141 the `http_parser` object's `data` field can be used.
142 An example of such a case is when using threads to handle a socket connection,
143 parse a request, and then give a response over that socket. By instantiation
144 of a thread-local struct containing relevant data (e.g. accepted socket,
145 allocated memory for callbacks to write into, etc), a parser's callbacks are
146 able to communicate data between the scope of the thread and the scope of the
147 callback in a threadsafe manner. This allows http-parser to be used in
148 multi-threaded contexts.
149
150 Example:
151 ```
152 typedef struct {
153 socket_t sock;
154 void* buffer;
155 int buf_len;
156 } custom_data_t;
157
158
159 int my_url_callback(http_parser* parser, const char *at, size_t length) {
160 /* access to thread local custom_data_t struct.
161 Use this access save parsed data for later use into thread local
162 buffer, or communicate over socket
163 */
164 parser->data;
165 ...
166 return 0;
167 }
168
169 ...
170
171 void http_parser_thread(socket_t sock) {
172 int nparsed = 0;
173 /* allocate memory for user data */
174 custom_data_t *my_data = malloc(sizeof(custom_data_t));
175
176 /* some information for use by callbacks.
177 * achieves thread -> callback information flow */
178 my_data->sock = sock;
179
180 /* instantiate a thread-local parser */
181 http_parser *parser = malloc(sizeof(http_parser));
182 http_parser_init(parser, HTTP_REQUEST); /* initialise parser */
183 /* this custom data reference is accessible through the reference to the
184 parser supplied to callback functions */
185 parser->data = my_data;
186
187 http_parser_settings settings; / * set up callbacks */
188 settings.on_url = my_url_callback;
189
190 /* execute parser */
191 nparsed = http_parser_execute(parser, &settings, buf, recved);
192
193 ...
194 /* parsed information copied from callback.
195 can now perform action on data copied into thread-local memory from callbacks.
196 achieves callback -> thread information flow */
197 my_data->buffer;
198 ...
199 }
200
201 ```
202
203 In case you parse HTTP message in chunks (i.e. `read()` request line
204 from socket, parse, read half headers, parse, etc) your data callbacks
205 may be called more than once. Http-parser guarantees that data pointer is only
206 valid for the lifetime of callback. You can also `read()` into a heap allocated
207 buffer to avoid copying memory around if this fits your application.
208
209 Reading headers may be a tricky task if you read/parse headers partially.
210 Basically, you need to remember whether last header callback was field or value
211 and apply the following logic:
212
213 (on_header_field and on_header_value shortened to on_h_*)
214 ------------------------ ------------ --------------------------------------------
215 | State (prev. callback) | Callback | Description/action |
216 ------------------------ ------------ --------------------------------------------
217 | nothing (first call) | on_h_field | Allocate new buffer and copy callback data |
218 | | | into it |
219 ------------------------ ------------ --------------------------------------------
220 | value | on_h_field | New header started. |
221 | | | Copy current name,value buffers to headers |
222 | | | list and allocate new buffer for new name |
223 ------------------------ ------------ --------------------------------------------
224 | field | on_h_field | Previous name continues. Reallocate name |
225 | | | buffer and append callback data to it |
226 ------------------------ ------------ --------------------------------------------
227 | field | on_h_value | Value for current header started. Allocate |
228 | | | new buffer and copy callback data to it |
229 ------------------------ ------------ --------------------------------------------
230 | value | on_h_value | Value continues. Reallocate value buffer |
231 | | | and append callback data to it |
232 ------------------------ ------------ --------------------------------------------
233
234
235 Parsing URLs
236 ------------
237
238 A simplistic zero-copy URL parser is provided as `http_parser_parse_url()`.
239 Users of this library may wish to use it to parse URLs constructed from
240 consecutive `on_url` callbacks.
241
242 See examples of reading in headers:
243
244 * [partial example](http://gist.github.com/155877) in C
245 * [from http-parser tests](http://github.com/joyent/http-parser/blob/37a0ff8/test.c#L403) in C
246 * [from Node library](http://github.com/joyent/node/blob/842eaf4/src/http.js#L284) in Javascript