blob: 621a22955e652e257c0f018e5a22918704b6a61f [file] [log] [blame]
Radek Krejci5aeea3a2018-09-05 13:29:36 +02001/**
2 * @file hash_table.h
3 * @author Radek Krejci <rkrejci@cesnet.cz>
4 * @author Michal Vasko <mvasko@cesnet.cz>
5 * @brief libyang hash table
6 *
7 * Copyright (c) 2015 - 2018 CESNET, z.s.p.o.
8 *
9 * This source code is licensed under BSD 3-Clause License (the "License").
10 * You may not use this file except in compliance with the License.
11 * You may obtain a copy of the License at
12 *
13 * https://opensource.org/licenses/BSD-3-Clause
14 */
15
16#ifndef LY_HASH_TABLE_H_
17#define LY_HASH_TABLE_H_
18
Radek Krejcie7b95092019-05-15 11:03:07 +020019#include <pthread.h>
20#include <stddef.h>
21#include <stdint.h>
22
Radek Krejci535ea9f2020-05-29 16:01:05 +020023#include "config.h"
Radek Krejcie7b95092019-05-15 11:03:07 +020024#include "log.h"
Radek Krejci5aeea3a2018-09-05 13:29:36 +020025
26/**
27 * @brief Compute hash from (several) string(s).
28 *
29 * Usage:
30 * - init hash to 0
31 * - repeatedly call dict_hash_multi(), provide hash from the last call
32 * - call dict_hash_multi() with key_part = NULL to finish the hash
33 */
34uint32_t dict_hash_multi(uint32_t hash, const char *key_part, size_t len);
35
Radek Krejcif2dc4c52018-11-08 09:04:13 +010036/*
37 * @brief Compute hash from a string.
38 */
39uint32_t dict_hash(const char *key, size_t len);
40
Radek Krejci5aeea3a2018-09-05 13:29:36 +020041/**
42 * @brief Callback for checking hash table values equivalence.
43 *
Michal Vasko90932a92020-02-12 14:33:03 +010044 * @param[in] val1_p Pointer to the first value, the one being searched (inserted/removed).
45 * @param[in] val2_p Pointer to the second value, the one stored in the hash table.
Radek Krejci5aeea3a2018-09-05 13:29:36 +020046 * @param[in] mod Whether the operation modifies the hash table (insert or remove) or not (find).
47 * @param[in] cb_data User callback data.
48 * @return 0 on non-equal, non-zero on equal.
49 */
50typedef int (*values_equal_cb)(void *val1_p, void *val2_p, int mod, void *cb_data);
51
52/** when the table is at least this much percent full, it is enlarged (double the size) */
53#define LYHT_ENLARGE_PERCENTAGE 75
54
55/** only once the table is this much percent full, enable shrinking */
56#define LYHT_FIRST_SHRINK_PERCENTAGE 50
57
58/** when the table is less than this much percent full, it is shrunk (half the size) */
59#define LYHT_SHRINK_PERCENTAGE 25
60
61/** never shrink beyond this size */
62#define LYHT_MIN_SIZE 8
63
64/**
65 * @brief Generic hash table record.
66 */
67struct ht_rec {
68 uint32_t hash; /* hash of the value */
69 int32_t hits; /* collision/overflow value count - 1 (a filled entry has 1 hit,
70 * special value -1 means a deleted record) */
71 unsigned char val[1]; /* arbitrary-size value */
72} _PACKED;
73
74/**
75 * @brief (Very) generic hash table.
76 *
77 * Hash table with open addressing collision resolution and
78 * linear probing of interval 1 (next free record is used).
79 * Removal is lazy (removed records are only marked), but
80 * if possible, they are fully emptied.
81 */
82struct hash_table {
83 uint32_t used; /* number of values stored in the hash table (filled records) */
84 uint32_t size; /* always holds 2^x == size (is power of 2), actually number of records allocated */
85 values_equal_cb val_equal; /* callback for testing value equivalence */
86 void *cb_data; /* user data callback arbitrary value */
87 uint16_t resize; /* 0 - resizing is disabled, *
88 * 1 - enlarging is enabled, *
89 * 2 - both shrinking and enlarging is enabled */
90 uint16_t rec_size; /* real size (in bytes) of one record for accessing recs array */
91 unsigned char *recs; /* pointer to the hash table itself (array of struct ht_rec) */
92};
93
94struct dict_rec {
95 char *value;
96 uint32_t refcount;
97};
98
99/**
100 * dictionary to store repeating strings
101 */
102struct dict_table {
103 struct hash_table *hash_tab;
104 pthread_mutex_t lock;
105};
106
107/**
108 * @brief Initiate content (non-zero values) of the dictionary
109 *
110 * @param[in] dict Dictionary table to initiate
111 */
112void lydict_init(struct dict_table *dict);
113
114/**
115 * @brief Cleanup the dictionary content
116 *
117 * @param[in] dict Dictionary table to cleanup
118 */
119void lydict_clean(struct dict_table *dict);
120
121/**
122 * @brief Create new hash table.
123 *
124 * @param[in] size Starting size of the hash table (capacity of values), must be power of 2.
125 * @param[in] val_size Size in bytes of value (the stored hashed item).
126 * @param[in] val_equal Callback for checking value equivalence.
127 * @param[in] cb_data User data always passed to \p val_equal.
128 * @param[in] resize Whether to resize the table on too few/too many records taken.
129 * @return Empty hash table, NULL on error.
130 */
131struct hash_table *lyht_new(uint32_t size, uint16_t val_size, values_equal_cb val_equal, void *cb_data, int resize);
132
133/**
134 * @brief Set hash table value equal callback.
135 *
136 * @param[in] ht Hash table to modify.
137 * @param[in] new_val_equal New callback for checking value equivalence.
138 * @return Previous callback for checking value equivalence.
139 */
140values_equal_cb lyht_set_cb(struct hash_table *ht, values_equal_cb new_val_equal);
141
142/**
143 * @brief Set hash table value equal callback user data.
144 *
145 * @param[in] ht Hash table to modify.
146 * @param[in] new_cb_data New data for values callback.
147 * @return Previous data for values callback.
148 */
149void *lyht_set_cb_data(struct hash_table *ht, void *new_cb_data);
150
151/**
152 * @brief Make a duplicate of an existing hash table.
153 *
154 * @param[in] orig Original hash table to duplicate.
155 * @return Duplicated hash table \p orig, NULL on error.
156 */
157struct hash_table *lyht_dup(const struct hash_table *orig);
158
159/**
160 * @brief Free a hash table.
161 *
162 * @param[in] ht Hash table to be freed.
163 */
164void lyht_free(struct hash_table *ht);
165
166/**
167 * @brief Find a value in a hash table.
168 *
169 * @param[in] ht Hash table to search in.
170 * @param[in] val_p Pointer to the value to find.
171 * @param[in] hash Hash of the stored value.
172 * @param[out] match_p Pointer to the matching value, optional.
173 * @return 0 on success, 1 on not found.
174 */
175int lyht_find(struct hash_table *ht, void *val_p, uint32_t hash, void **match_p);
176
177/**
178 * @brief Find another equal value in the hash table.
179 *
180 * @param[in] ht Hash table to search in.
181 * @param[in] val_p Pointer to the previously found value in \p ht.
182 * @param[in] hash Hash of the previously found value.
183 * @param[out] match_p Pointer to the matching value, optional.
184 * @return 0 on success, 1 on not found.
185 */
186int lyht_find_next(struct hash_table *ht, void *val_p, uint32_t hash, void **match_p);
187
188/**
189 * @brief Insert a value into a hash table.
190 *
191 * @param[in] ht Hash table to insert into.
192 * @param[in] val_p Pointer to the value to insert. Be careful, if the values stored in the hash table
193 * are pointers, \p val_p must be a pointer to a pointer.
194 * @param[in] hash Hash of the stored value.
195 * @param[out] match_p Pointer to the stored value, optional
Radek Krejci0ae092d2018-09-20 16:43:19 +0200196 * @return LY_ERR value (LY_EEXIST if the value is already present).
Radek Krejci5aeea3a2018-09-05 13:29:36 +0200197 */
Radek Krejci0ae092d2018-09-20 16:43:19 +0200198LY_ERR lyht_insert(struct hash_table *ht, void *val_p, uint32_t hash, void **match_p);
Radek Krejci5aeea3a2018-09-05 13:29:36 +0200199
200/**
201 * @brief Insert a value into hash table. Same functionality as lyht_insert()
202 * but allows to specify a temporary val equal callback to be used in case the hash table
203 * will be resized after successful insertion.
204 *
205 * @param[in] ht Hash table to insert into.
206 * @param[in] val_p Pointer to the value to insert. Be careful, if the values stored in the hash table
207 * are pointers, \p val_p must be a pointer to a pointer.
208 * @param[in] hash Hash of the stored value.
209 * @param[in] resize_val_equal Val equal callback to use for resizing.
210 * @param[out] match_p Pointer to the stored value, optional
211 * @return LY_ERR return value (LY_EEXIST if the value is already present).
212 */
213LY_ERR lyht_insert_with_resize_cb(struct hash_table *ht, void *val_p, uint32_t hash, values_equal_cb resize_val_equal,
214 void **match_p);
215
216/**
217 * @brief Remove a value from a hash table.
218 *
219 * @param[in] ht Hash table to remove from.
220 * @param[in] value_p Pointer to value to be removed. Be careful, if the values stored in the hash table
221 * are pointers, \p value_p must be a pointer to a pointer.
222 * @param[in] hash Hash of the stored value.
223 * @return LY_ERR return value (LY_EINVAL if value was not found).
224 */
225LY_ERR lyht_remove(struct hash_table *ht, void *val_p, uint32_t hash);
226
227#endif /* LY_HASH_TABLE_H_ */