/root/doris/be/src/util/coding.h
Line | Count | Source |
1 | | // Copyright (c) 2011-present, Facebook, Inc. All rights reserved. |
2 | | // This source code is licensed under both the GPLv2 (found in the |
3 | | // COPYING file in the root directory) and Apache 2.0 License |
4 | | // (found in the LICENSE.Apache file in the root directory). |
5 | | // Copyright (c) 2011 The LevelDB Authors. All rights reserved. |
6 | | // Use of this source code is governed by a BSD-style license that can be |
7 | | // found in the LICENSE file. See the AUTHORS file for names of contributors. |
8 | | |
9 | | #pragma once |
10 | | |
11 | | #include <bit> |
12 | | |
13 | | #ifndef __APPLE__ |
14 | | #include <endian.h> |
15 | | #endif |
16 | | #include <stdint.h> |
17 | | #include <string.h> |
18 | | |
19 | | #include "olap/olap_common.h" |
20 | | #include "util/slice.h" |
21 | | #include "vec/common/endian.h" |
22 | | |
23 | | namespace doris { |
24 | | |
25 | | // TODO(zc): add encode big endian later when we need it |
26 | | // use big endian when we have order requirement. |
27 | | // little endian is more efficient when we use X86 CPU, so |
28 | | // when we have no order needs, we prefer little endian encoding |
29 | | inline void encode_fixed8(uint8_t* buf, uint8_t val) { |
30 | | *buf = val; |
31 | | } |
32 | | |
33 | 554k | inline void encode_fixed16_le(uint8_t* buf, uint16_t val) { |
34 | 554k | val = to_endian<std::endian::little>(val); |
35 | 554k | memcpy(buf, &val, sizeof(val)); |
36 | 554k | } |
37 | | |
38 | 123M | inline void encode_fixed32_le(uint8_t* buf, uint32_t val) { |
39 | 123M | val = to_endian<std::endian::little>(val); |
40 | 123M | memcpy(buf, &val, sizeof(val)); |
41 | 123M | } |
42 | | |
43 | 424k | inline void encode_fixed64_le(uint8_t* buf, uint64_t val) { |
44 | 424k | val = to_endian<std::endian::little>(val); |
45 | 424k | memcpy(buf, &val, sizeof(val)); |
46 | 424k | } |
47 | | |
48 | 24.4k | inline void encode_fixed128_le(uint8_t* buf, uint128_t val) { |
49 | 24.4k | val = to_endian<std::endian::little>(val); |
50 | 24.4k | memcpy(buf, &val, sizeof(val)); |
51 | 24.4k | } |
52 | | |
53 | 6.71M | inline uint8_t decode_fixed8(const uint8_t* buf) { |
54 | 6.71M | return *buf; |
55 | 6.71M | } |
56 | | |
57 | 603k | inline uint16_t decode_fixed16_le(const uint8_t* buf) { |
58 | 603k | uint16_t res; |
59 | 603k | memcpy(&res, buf, sizeof(res)); |
60 | 603k | return to_endian<std::endian::little>(res); |
61 | 603k | } |
62 | | |
63 | 286M | inline uint32_t decode_fixed32_le(const uint8_t* buf) { |
64 | 286M | uint32_t res; |
65 | 286M | memcpy(&res, buf, sizeof(res)); |
66 | 286M | return to_endian<std::endian::little>(res); |
67 | 286M | } |
68 | | |
69 | 3.60G | inline uint64_t decode_fixed64_le(const uint8_t* buf) { |
70 | 3.60G | uint64_t res; |
71 | 3.60G | memcpy(&res, buf, sizeof(res)); |
72 | 3.60G | return to_endian<std::endian::little>(res); |
73 | 3.60G | } |
74 | | |
75 | 24.4k | inline uint128_t decode_fixed128_le(const uint8_t* buf) { |
76 | 24.4k | uint128_t res; |
77 | 24.4k | memcpy(&res, buf, sizeof(res)); |
78 | 24.4k | return to_endian<std::endian::little>(res); |
79 | 24.4k | } |
80 | | |
81 | | template <typename T> |
82 | 109M | void put_fixed32_le(T* dst, uint32_t val) { |
83 | 109M | uint8_t buf[sizeof(val)]; |
84 | 109M | encode_fixed32_le(buf, val); |
85 | 109M | dst->append((char*)buf, sizeof(buf)); |
86 | 109M | } _ZN5doris14put_fixed32_leINS_10faststringEEEvPT_j Line | Count | Source | 82 | 104M | void put_fixed32_le(T* dst, uint32_t val) { | 83 | 104M | uint8_t buf[sizeof(val)]; | 84 | 104M | encode_fixed32_le(buf, val); | 85 | 104M | dst->append((char*)buf, sizeof(buf)); | 86 | 104M | } |
_ZN5doris14put_fixed32_leINSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEEEvPT_j Line | Count | Source | 82 | 4.76M | void put_fixed32_le(T* dst, uint32_t val) { | 83 | 4.76M | uint8_t buf[sizeof(val)]; | 84 | 4.76M | encode_fixed32_le(buf, val); | 85 | 4.76M | dst->append((char*)buf, sizeof(buf)); | 86 | 4.76M | } |
|
87 | | |
88 | | template <typename T> |
89 | 37.1k | void put_fixed64_le(T* dst, uint64_t val) { |
90 | 37.1k | uint8_t buf[sizeof(val)]; |
91 | 37.1k | encode_fixed64_le(buf, val); |
92 | 37.1k | dst->append((char*)buf, sizeof(buf)); |
93 | 37.1k | } _ZN5doris14put_fixed64_leINSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEEEvPT_m Line | Count | Source | 89 | 12.6k | void put_fixed64_le(T* dst, uint64_t val) { | 90 | 12.6k | uint8_t buf[sizeof(val)]; | 91 | 12.6k | encode_fixed64_le(buf, val); | 92 | 12.6k | dst->append((char*)buf, sizeof(buf)); | 93 | 12.6k | } |
_ZN5doris14put_fixed64_leINS_10faststringEEEvPT_m Line | Count | Source | 89 | 24.4k | void put_fixed64_le(T* dst, uint64_t val) { | 90 | 24.4k | uint8_t buf[sizeof(val)]; | 91 | 24.4k | encode_fixed64_le(buf, val); | 92 | 24.4k | dst->append((char*)buf, sizeof(buf)); | 93 | 24.4k | } |
|
94 | | |
95 | | // Returns the length of the varint32 or varint64 encoding of "v" |
96 | 57 | inline int varint_length(uint64_t v) { |
97 | 57 | int len = 1; |
98 | 57 | while (v >= 128) { |
99 | 0 | v >>= 7; |
100 | 0 | len++; |
101 | 0 | } |
102 | 57 | return len; |
103 | 57 | } |
104 | | |
105 | | template <typename T> |
106 | 24.4k | void put_fixed128_le(T* dst, uint128_t val) { |
107 | 24.4k | uint8_t buf[sizeof(val)]; |
108 | 24.4k | encode_fixed128_le(buf, val); |
109 | 24.4k | dst->append((char*)buf, sizeof(buf)); |
110 | 24.4k | } |
111 | | |
112 | | extern uint8_t* encode_varint32(uint8_t* dst, uint32_t value); |
113 | | extern uint8_t* encode_varint64(uint8_t* dst, uint64_t value); |
114 | | |
115 | 4.01M | inline uint8_t* encode_varint64(uint8_t* dst, uint64_t v) { |
116 | 4.01M | static const unsigned int B = 128; |
117 | 9.03M | while (v >= B) { |
118 | | // Fetch low seven bits from current v, and the eight bit is marked as compression mark. |
119 | | // v | B is optimised from (v & (B-1)) | B, because result is assigned to uint8_t and other bits |
120 | | // is cleared by implicit conversion. |
121 | 5.02M | *(dst++) = v | B; |
122 | 5.02M | v >>= 7; |
123 | 5.02M | } |
124 | 4.01M | *(dst++) = static_cast<unsigned char>(v); |
125 | 4.01M | return dst; |
126 | 4.01M | } |
127 | | |
128 | | extern const uint8_t* decode_varint32_ptr_fallback(const uint8_t* p, const uint8_t* limit, |
129 | | uint32_t* value); |
130 | | |
131 | | inline const uint8_t* decode_varint32_ptr(const uint8_t* ptr, const uint8_t* limit, |
132 | 237M | uint32_t* value) { |
133 | 237M | if (ptr < limit) { |
134 | 237M | uint32_t result = *ptr; |
135 | 237M | if ((result & 128) == 0) { |
136 | 237M | *value = result; |
137 | 237M | return ptr + 1; |
138 | 237M | } |
139 | 237M | } |
140 | 18.4E | return decode_varint32_ptr_fallback(ptr, limit, value); |
141 | 237M | } |
142 | | |
143 | | extern const uint8_t* decode_varint64_ptr(const uint8_t* p, const uint8_t* limit, uint64_t* value); |
144 | | |
145 | | template <typename T> |
146 | 66.7M | void put_varint32(T* dst, uint32_t v) { |
147 | 66.7M | uint8_t buf[16]; |
148 | 66.7M | uint8_t* ptr = encode_varint32(buf, v); |
149 | 66.7M | dst->append((char*)buf, static_cast<size_t>(ptr - buf)); |
150 | 66.7M | } |
151 | | |
152 | | template <typename T> |
153 | | void put_varint64(T* dst, uint64_t v) { |
154 | | uint8_t buf[16]; |
155 | | uint8_t* ptr = encode_varint64(buf, v); |
156 | | dst->append((char*)buf, static_cast<size_t>(ptr - buf)); |
157 | | } |
158 | | |
159 | | template <typename T> |
160 | 4.01M | void put_length_prefixed_slice(T* dst, const Slice& value) { |
161 | 4.01M | put_varint32(dst, value.get_size()); |
162 | 4.01M | dst->append(value.get_data(), value.get_size()); |
163 | 4.01M | } |
164 | | |
165 | | template <typename T> |
166 | 4.01M | void put_varint64_varint32(T* dst, uint64_t v1, uint32_t v2) { |
167 | 4.01M | uint8_t buf[16]; |
168 | 4.01M | uint8_t* ptr = encode_varint64(buf, v1); |
169 | 4.01M | ptr = encode_varint32(ptr, v2); |
170 | 4.01M | dst->append((char*)buf, static_cast<size_t>(ptr - buf)); |
171 | 4.01M | } |
172 | | |
173 | | // parse a varint32 from the start of `input` into `val`. |
174 | | // on success, return true and advance `input` past the parsed value. |
175 | | // on failure, return false and `input` is not modified. |
176 | 799k | inline bool get_varint32(Slice* input, uint32_t* val) { |
177 | 799k | const uint8_t* p = (const uint8_t*)input->data; |
178 | 799k | const uint8_t* limit = p + input->size; |
179 | 799k | const uint8_t* q = decode_varint32_ptr(p, limit, val); |
180 | 799k | if (q == nullptr) { |
181 | 0 | return false; |
182 | 799k | } else { |
183 | 799k | *input = Slice(q, limit - q); |
184 | 799k | return true; |
185 | 799k | } |
186 | 799k | } |
187 | | |
188 | | // parse a varint64 from the start of `input` into `val`. |
189 | | // on success, return true and advance `input` past the parsed value. |
190 | | // on failure, return false and `input` is not modified. |
191 | 283k | inline bool get_varint64(Slice* input, uint64_t* val) { |
192 | 283k | const uint8_t* p = (const uint8_t*)input->data; |
193 | 283k | const uint8_t* limit = p + input->size; |
194 | 283k | const uint8_t* q = decode_varint64_ptr(p, limit, val); |
195 | 283k | if (q == nullptr) { |
196 | 0 | return false; |
197 | 283k | } else { |
198 | 283k | *input = Slice(q, limit - q); |
199 | 283k | return true; |
200 | 283k | } |
201 | 283k | } |
202 | | |
203 | | // parse a length-prefixed-slice from the start of `input` into `val`. |
204 | | // on success, return true and advance `input` past the parsed value. |
205 | | // on failure, return false and `input` may or may not be modified. |
206 | 288k | inline bool get_length_prefixed_slice(Slice* input, Slice* val) { |
207 | 288k | uint32_t len; |
208 | 288k | if (get_varint32(input, &len) && input->get_size() >= len) { |
209 | 285k | *val = Slice(input->get_data(), len); |
210 | 285k | input->remove_prefix(len); |
211 | 285k | return true; |
212 | 285k | } else { |
213 | 3.15k | return false; |
214 | 3.15k | } |
215 | 288k | } |
216 | | |
217 | | } // namespace doris |