http/header/mod.rs
1//! HTTP header types
2//!
3//! The module provides [`HeaderName`], [`HeaderMap`], and a number of types
4//! used for interacting with `HeaderMap`. These types allow representing both
5//! HTTP/1 and HTTP/2 headers.
6//!
7//! # `HeaderName`
8//!
9//! The `HeaderName` type represents both standard header names as well as
10//! custom header names. The type handles the case insensitive nature of header
11//! names and is used as the key portion of `HeaderMap`. Header names are
12//! normalized to lower case. In other words, when creating a `HeaderName` with
13//! a string, even if upper case characters are included, when getting a string
14//! representation of the `HeaderName`, it will be all lower case. This allows
15//! for faster `HeaderMap` comparison operations.
16//!
17//! The internal representation is optimized to efficiently handle the cases
18//! most commonly encountered when working with HTTP. Standard header names are
19//! special cased and are represented internally as an enum. Short custom
20//! headers will be stored directly in the `HeaderName` struct and will not
21//! incur any allocation overhead, however longer strings will require an
22//! allocation for storage.
23//!
24//! ## Limitations
25//!
26//! `HeaderName` has a max length of 32,768 for header names. Attempting to
27//! parse longer names will result in a panic.
28//!
29//! # `HeaderMap`
30//!
31//! `HeaderMap` is a map structure of header names highly optimized for use
32//! cases common with HTTP. It is a [multimap] structure, where each header name
33//! may have multiple associated header values. Given this, some of the APIs
34//! diverge from [`HashMap`].
35//!
36//! ## Overview
37//!
38//! Just like `HashMap` in Rust's stdlib, `HeaderMap` is based on [Robin Hood
39//! hashing]. This algorithm tends to reduce the worst case search times in the
40//! table and enables high load factors without seriously affecting performance.
41//! Internally, keys and values are stored in vectors. As such, each insertion
42//! will not incur allocation overhead. However, once the underlying vector
43//! storage is full, a larger vector must be allocated and all values copied.
44//!
45//! ## Deterministic ordering
46//!
47//! Unlike Rust's `HashMap`, values in `HeaderMap` are deterministically
48//! ordered. Roughly, values are ordered by insertion. This means that a
49//! function that deterministically operates on a header map can rely on the
50//! iteration order to remain consistent across processes and platforms.
51//!
52//! ## Adaptive hashing
53//!
54//! `HeaderMap` uses an adaptive hashing strategy in order to efficiently handle
55//! most common cases. All standard headers have statically computed hash values
56//! which removes the need to perform any hashing of these headers at runtime.
57//! The default hash function emphasizes performance over robustness. However,
58//! `HeaderMap` detects high collision rates and switches to a secure hash
59//! function in those events. The threshold is set such that only denial of
60//! service attacks should trigger it.
61//!
62//! ## Limitations
63//!
64//! `HeaderMap` can store a maximum of 32,768 headers (header name / value
65//! pairs). Attempting to insert more will result in a panic.
66//!
67//! [`HeaderName`]: struct.HeaderName.html
68//! [`HeaderMap`]: struct.HeaderMap.html
69//! [multimap]: https://en.wikipedia.org/wiki/Multimap
70//! [`HashMap`]: https://doc.rust-lang.org/std/collections/struct.HashMap.html
71//! [Robin Hood hashing]: https://en.wikipedia.org/wiki/Hash_table#Robin_Hood_hashing
72
73mod map;
74mod name;
75mod value;
76
77pub use self::map::{
78 AsHeaderName, Drain, Entry, GetAll, HeaderMap, IntoHeaderName, IntoIter, Iter, IterMut, Keys,
79 MaxSizeReached, OccupiedEntry, VacantEntry, ValueDrain, ValueIter, ValueIterMut, Values,
80 ValuesMut,
81};
82pub use self::name::{HeaderName, InvalidHeaderName};
83pub use self::value::{HeaderValue, InvalidHeaderValue, ToStrError};
84
85// Use header name constants
86#[rustfmt::skip]
87pub use self::name::{
88 ACCEPT,
89 ACCEPT_CHARSET,
90 ACCEPT_ENCODING,
91 ACCEPT_LANGUAGE,
92 ACCEPT_RANGES,
93 ACCESS_CONTROL_ALLOW_CREDENTIALS,
94 ACCESS_CONTROL_ALLOW_HEADERS,
95 ACCESS_CONTROL_ALLOW_METHODS,
96 ACCESS_CONTROL_ALLOW_ORIGIN,
97 ACCESS_CONTROL_EXPOSE_HEADERS,
98 ACCESS_CONTROL_MAX_AGE,
99 ACCESS_CONTROL_REQUEST_HEADERS,
100 ACCESS_CONTROL_REQUEST_METHOD,
101 AGE,
102 ALLOW,
103 ALT_SVC,
104 AUTHORIZATION,
105 CACHE_CONTROL,
106 CACHE_STATUS,
107 CDN_CACHE_CONTROL,
108 CONNECTION,
109 CONTENT_DISPOSITION,
110 CONTENT_ENCODING,
111 CONTENT_LANGUAGE,
112 CONTENT_LENGTH,
113 CONTENT_LOCATION,
114 CONTENT_RANGE,
115 CONTENT_SECURITY_POLICY,
116 CONTENT_SECURITY_POLICY_REPORT_ONLY,
117 CONTENT_TYPE,
118 COOKIE,
119 DNT,
120 DATE,
121 ETAG,
122 EXPECT,
123 EXPIRES,
124 FORWARDED,
125 FROM,
126 HOST,
127 IF_MATCH,
128 IF_MODIFIED_SINCE,
129 IF_NONE_MATCH,
130 IF_RANGE,
131 IF_UNMODIFIED_SINCE,
132 LAST_MODIFIED,
133 LINK,
134 LOCATION,
135 MAX_FORWARDS,
136 ORIGIN,
137 PRAGMA,
138 PROXY_AUTHENTICATE,
139 PROXY_AUTHORIZATION,
140 PUBLIC_KEY_PINS,
141 PUBLIC_KEY_PINS_REPORT_ONLY,
142 RANGE,
143 REFERER,
144 REFERRER_POLICY,
145 REFRESH,
146 RETRY_AFTER,
147 SEC_WEBSOCKET_ACCEPT,
148 SEC_WEBSOCKET_EXTENSIONS,
149 SEC_WEBSOCKET_KEY,
150 SEC_WEBSOCKET_PROTOCOL,
151 SEC_WEBSOCKET_VERSION,
152 SERVER,
153 SET_COOKIE,
154 STRICT_TRANSPORT_SECURITY,
155 TE,
156 TRAILER,
157 TRANSFER_ENCODING,
158 UPGRADE,
159 UPGRADE_INSECURE_REQUESTS,
160 USER_AGENT,
161 VARY,
162 VIA,
163 WARNING,
164 WWW_AUTHENTICATE,
165 X_CONTENT_TYPE_OPTIONS,
166 X_DNS_PREFETCH_CONTROL,
167 X_FRAME_OPTIONS,
168 X_XSS_PROTECTION,
169};
170
171/// Maximum length of a header name
172///
173/// Generally, 64kb for a header name is WAY too much than would ever be needed
174/// in practice. Restricting it to this size enables using `u16` values to
175/// represent offsets when dealing with header names.
176const MAX_HEADER_NAME_LEN: usize = (1 << 16) - 1;