From bc4e83f2fa591205b9eb41fd6c3497d4b1f26430 Mon Sep 17 00:00:00 2001 From: Baptiste Daroussin Date: Sun, 25 Dec 2016 03:44:01 +0100 Subject: Use khash instead of uthash, it is a bit faster and consumer less memory --- external/klib/khash.h | 627 +++++++++++++++++++++++++++++++ external/uthash/uthash.h | 958 ----------------------------------------------- 2 files changed, 627 insertions(+), 958 deletions(-) create mode 100644 external/klib/khash.h delete mode 100644 external/uthash/uthash.h (limited to 'external') diff --git a/external/klib/khash.h b/external/klib/khash.h new file mode 100644 index 00000000..e44f3288 --- /dev/null +++ b/external/klib/khash.h @@ -0,0 +1,627 @@ +/* The MIT License + + Copyright (c) 2008, 2009, 2011 by Attractive Chaos + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice shall be + included in all copies or substantial portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + SOFTWARE. +*/ + +/* + An example: + +#include "khash.h" +KHASH_MAP_INIT_INT(32, char) +int main() { + int ret, is_missing; + khiter_t k; + khash_t(32) *h = kh_init(32); + k = kh_put(32, h, 5, &ret); + kh_value(h, k) = 10; + k = kh_get(32, h, 10); + is_missing = (k == kh_end(h)); + k = kh_get(32, h, 5); + kh_del(32, h, k); + for (k = kh_begin(h); k != kh_end(h); ++k) + if (kh_exist(h, k)) kh_value(h, k) = 1; + kh_destroy(32, h); + return 0; +} +*/ + +/* + 2013-05-02 (0.2.8): + + * Use quadratic probing. When the capacity is power of 2, stepping function + i*(i+1)/2 guarantees to traverse each bucket. It is better than double + hashing on cache performance and is more robust than linear probing. + + In theory, double hashing should be more robust than quadratic probing. + However, my implementation is probably not for large hash tables, because + the second hash function is closely tied to the first hash function, + which reduce the effectiveness of double hashing. + + Reference: http://research.cs.vt.edu/AVresearch/hashing/quadratic.php + + 2011-12-29 (0.2.7): + + * Minor code clean up; no actual effect. + + 2011-09-16 (0.2.6): + + * The capacity is a power of 2. This seems to dramatically improve the + speed for simple keys. Thank Zilong Tan for the suggestion. Reference: + + - http://code.google.com/p/ulib/ + - http://nothings.org/computer/judy/ + + * Allow to optionally use linear probing which usually has better + performance for random input. Double hashing is still the default as it + is more robust to certain non-random input. + + * Added Wang's integer hash function (not used by default). This hash + function is more robust to certain non-random input. + + 2011-02-14 (0.2.5): + + * Allow to declare global functions. + + 2009-09-26 (0.2.4): + + * Improve portability + + 2008-09-19 (0.2.3): + + * Corrected the example + * Improved interfaces + + 2008-09-11 (0.2.2): + + * Improved speed a little in kh_put() + + 2008-09-10 (0.2.1): + + * Added kh_clear() + * Fixed a compiling error + + 2008-09-02 (0.2.0): + + * Changed to token concatenation which increases flexibility. + + 2008-08-31 (0.1.2): + + * Fixed a bug in kh_get(), which has not been tested previously. + + 2008-08-31 (0.1.1): + + * Added destructor +*/ + + +#ifndef __AC_KHASH_H +#define __AC_KHASH_H + +/*! + @header + + Generic hash table library. + */ + +#define AC_VERSION_KHASH_H "0.2.8" + +#include +#include +#include + +/* compiler specific configuration */ + +#if UINT_MAX == 0xffffffffu +typedef unsigned int khint32_t; +#elif ULONG_MAX == 0xffffffffu +typedef unsigned long khint32_t; +#endif + +#if ULONG_MAX == ULLONG_MAX +typedef unsigned long khint64_t; +#else +typedef unsigned long long khint64_t; +#endif + +#ifndef kh_inline +#ifdef _MSC_VER +#define kh_inline __inline +#else +#define kh_inline inline +#endif +#endif /* kh_inline */ + +#ifndef klib_unused +#if (defined __clang__ && __clang_major__ >= 3) || (defined __GNUC__ && __GNUC__ >= 3) +#define klib_unused __attribute__ ((__unused__)) +#else +#define klib_unused +#endif +#endif /* klib_unused */ + +typedef khint32_t khint_t; +typedef khint_t khiter_t; + +#define __ac_isempty(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&2) +#define __ac_isdel(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&1) +#define __ac_iseither(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&3) +#define __ac_set_isdel_false(flag, i) (flag[i>>4]&=~(1ul<<((i&0xfU)<<1))) +#define __ac_set_isempty_false(flag, i) (flag[i>>4]&=~(2ul<<((i&0xfU)<<1))) +#define __ac_set_isboth_false(flag, i) (flag[i>>4]&=~(3ul<<((i&0xfU)<<1))) +#define __ac_set_isdel_true(flag, i) (flag[i>>4]|=1ul<<((i&0xfU)<<1)) + +#define __ac_fsize(m) ((m) < 16? 1 : (m)>>4) + +#ifndef kroundup32 +#define kroundup32(x) (--(x), (x)|=(x)>>1, (x)|=(x)>>2, (x)|=(x)>>4, (x)|=(x)>>8, (x)|=(x)>>16, ++(x)) +#endif + +#ifndef kcalloc +#define kcalloc(N,Z) calloc(N,Z) +#endif +#ifndef kmalloc +#define kmalloc(Z) malloc(Z) +#endif +#ifndef krealloc +#define krealloc(P,Z) realloc(P,Z) +#endif +#ifndef kfree +#define kfree(P) free(P) +#endif + +static const double __ac_HASH_UPPER = 0.77; + +#define __KHASH_TYPE(name, khkey_t, khval_t) \ + typedef struct kh_##name##_s { \ + khint_t n_buckets, size, n_occupied, upper_bound; \ + khint32_t *flags; \ + khkey_t *keys; \ + khval_t *vals; \ + } kh_##name##_t; + +#define __KHASH_PROTOTYPES(name, khkey_t, khval_t) \ + extern kh_##name##_t *kh_init_##name(void); \ + extern void kh_destroy_##name(kh_##name##_t *h); \ + extern void kh_clear_##name(kh_##name##_t *h); \ + extern khint_t kh_get_##name(const kh_##name##_t *h, khkey_t key); \ + extern int kh_resize_##name(kh_##name##_t *h, khint_t new_n_buckets); \ + extern khint_t kh_put_##name(kh_##name##_t *h, khkey_t key, int *ret); \ + extern void kh_del_##name(kh_##name##_t *h, khint_t x); + +#define __KHASH_IMPL(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \ + SCOPE kh_##name##_t *kh_init_##name(void) { \ + return (kh_##name##_t*)kcalloc(1, sizeof(kh_##name##_t)); \ + } \ + SCOPE void kh_destroy_##name(kh_##name##_t *h) \ + { \ + if (h) { \ + kfree((void *)h->keys); kfree(h->flags); \ + kfree((void *)h->vals); \ + kfree(h); \ + } \ + } \ + SCOPE void kh_clear_##name(kh_##name##_t *h) \ + { \ + if (h && h->flags) { \ + memset(h->flags, 0xaa, __ac_fsize(h->n_buckets) * sizeof(khint32_t)); \ + h->size = h->n_occupied = 0; \ + } \ + } \ + SCOPE khint_t kh_get_##name(const kh_##name##_t *h, khkey_t key) \ + { \ + if (h->n_buckets) { \ + khint_t k, i, last, mask, step = 0; \ + mask = h->n_buckets - 1; \ + k = __hash_func(key); i = k & mask; \ + last = i; \ + while (!__ac_isempty(h->flags, i) && (__ac_isdel(h->flags, i) || !__hash_equal(h->keys[i], key))) { \ + i = (i + (++step)) & mask; \ + if (i == last) return h->n_buckets; \ + } \ + return __ac_iseither(h->flags, i)? h->n_buckets : i; \ + } else return 0; \ + } \ + SCOPE int kh_resize_##name(kh_##name##_t *h, khint_t new_n_buckets) \ + { /* This function uses 0.25*n_buckets bytes of working space instead of [sizeof(key_t+val_t)+.25]*n_buckets. */ \ + khint32_t *new_flags = 0; \ + khint_t j = 1; \ + { \ + kroundup32(new_n_buckets); \ + if (new_n_buckets < 4) new_n_buckets = 4; \ + if (h->size >= (khint_t)(new_n_buckets * __ac_HASH_UPPER + 0.5)) j = 0; /* requested size is too small */ \ + else { /* hash table size to be changed (shrink or expand); rehash */ \ + new_flags = (khint32_t*)kmalloc(__ac_fsize(new_n_buckets) * sizeof(khint32_t)); \ + if (!new_flags) return -1; \ + memset(new_flags, 0xaa, __ac_fsize(new_n_buckets) * sizeof(khint32_t)); \ + if (h->n_buckets < new_n_buckets) { /* expand */ \ + khkey_t *new_keys = (khkey_t*)krealloc((void *)h->keys, new_n_buckets * sizeof(khkey_t)); \ + if (!new_keys) { kfree(new_flags); return -1; } \ + h->keys = new_keys; \ + if (kh_is_map) { \ + khval_t *new_vals = (khval_t*)krealloc((void *)h->vals, new_n_buckets * sizeof(khval_t)); \ + if (!new_vals) { kfree(new_flags); return -1; } \ + h->vals = new_vals; \ + } \ + } /* otherwise shrink */ \ + } \ + } \ + if (j) { /* rehashing is needed */ \ + for (j = 0; j != h->n_buckets; ++j) { \ + if (__ac_iseither(h->flags, j) == 0) { \ + khkey_t key = h->keys[j]; \ + khval_t val; \ + khint_t new_mask; \ + new_mask = new_n_buckets - 1; \ + if (kh_is_map) val = h->vals[j]; \ + __ac_set_isdel_true(h->flags, j); \ + while (1) { /* kick-out process; sort of like in Cuckoo hashing */ \ + khint_t k, i, step = 0; \ + k = __hash_func(key); \ + i = k & new_mask; \ + while (!__ac_isempty(new_flags, i)) i = (i + (++step)) & new_mask; \ + __ac_set_isempty_false(new_flags, i); \ + if (i < h->n_buckets && __ac_iseither(h->flags, i) == 0) { /* kick out the existing element */ \ + { khkey_t tmp = h->keys[i]; h->keys[i] = key; key = tmp; } \ + if (kh_is_map) { khval_t tmp = h->vals[i]; h->vals[i] = val; val = tmp; } \ + __ac_set_isdel_true(h->flags, i); /* mark it as deleted in the old hash table */ \ + } else { /* write the element and jump out of the loop */ \ + h->keys[i] = key; \ + if (kh_is_map) h->vals[i] = val; \ + break; \ + } \ + } \ + } \ + } \ + if (h->n_buckets > new_n_buckets) { /* shrink the hash table */ \ + h->keys = (khkey_t*)krealloc((void *)h->keys, new_n_buckets * sizeof(khkey_t)); \ + if (kh_is_map) h->vals = (khval_t*)krealloc((void *)h->vals, new_n_buckets * sizeof(khval_t)); \ + } \ + kfree(h->flags); /* free the working space */ \ + h->flags = new_flags; \ + h->n_buckets = new_n_buckets; \ + h->n_occupied = h->size; \ + h->upper_bound = (khint_t)(h->n_buckets * __ac_HASH_UPPER + 0.5); \ + } \ + return 0; \ + } \ + SCOPE khint_t kh_put_##name(kh_##name##_t *h, khkey_t key, int *ret) \ + { \ + khint_t x; \ + if (h->n_occupied >= h->upper_bound) { /* update the hash table */ \ + if (h->n_buckets > (h->size<<1)) { \ + if (kh_resize_##name(h, h->n_buckets - 1) < 0) { /* clear "deleted" elements */ \ + *ret = -1; return h->n_buckets; \ + } \ + } else if (kh_resize_##name(h, h->n_buckets + 1) < 0) { /* expand the hash table */ \ + *ret = -1; return h->n_buckets; \ + } \ + } /* TODO: to implement automatically shrinking; resize() already support shrinking */ \ + { \ + khint_t k, i, site, last, mask = h->n_buckets - 1, step = 0; \ + x = site = h->n_buckets; k = __hash_func(key); i = k & mask; \ + if (__ac_isempty(h->flags, i)) x = i; /* for speed up */ \ + else { \ + last = i; \ + while (!__ac_isempty(h->flags, i) && (__ac_isdel(h->flags, i) || !__hash_equal(h->keys[i], key))) { \ + if (__ac_isdel(h->flags, i)) site = i; \ + i = (i + (++step)) & mask; \ + if (i == last) { x = site; break; } \ + } \ + if (x == h->n_buckets) { \ + if (__ac_isempty(h->flags, i) && site != h->n_buckets) x = site; \ + else x = i; \ + } \ + } \ + } \ + if (__ac_isempty(h->flags, x)) { /* not present at all */ \ + h->keys[x] = key; \ + __ac_set_isboth_false(h->flags, x); \ + ++h->size; ++h->n_occupied; \ + *ret = 1; \ + } else if (__ac_isdel(h->flags, x)) { /* deleted */ \ + h->keys[x] = key; \ + __ac_set_isboth_false(h->flags, x); \ + ++h->size; \ + *ret = 2; \ + } else *ret = 0; /* Don't touch h->keys[x] if present and not deleted */ \ + return x; \ + } \ + SCOPE void kh_del_##name(kh_##name##_t *h, khint_t x) \ + { \ + if (x != h->n_buckets && !__ac_iseither(h->flags, x)) { \ + __ac_set_isdel_true(h->flags, x); \ + --h->size; \ + } \ + } + +#define KHASH_DECLARE(name, khkey_t, khval_t) \ + __KHASH_TYPE(name, khkey_t, khval_t) \ + __KHASH_PROTOTYPES(name, khkey_t, khval_t) + +#define KHASH_INIT2(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \ + __KHASH_TYPE(name, khkey_t, khval_t) \ + __KHASH_IMPL(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) + +#define KHASH_INIT(name, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \ + KHASH_INIT2(name, static kh_inline klib_unused, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) + +/* --- BEGIN OF HASH FUNCTIONS --- */ + +/*! @function + @abstract Integer hash function + @param key The integer [khint32_t] + @return The hash value [khint_t] + */ +#define kh_int_hash_func(key) (khint32_t)(key) +/*! @function + @abstract Integer comparison function + */ +#define kh_int_hash_equal(a, b) ((a) == (b)) +/*! @function + @abstract 64-bit integer hash function + @param key The integer [khint64_t] + @return The hash value [khint_t] + */ +#define kh_int64_hash_func(key) (khint32_t)((key)>>33^(key)^(key)<<11) +/*! @function + @abstract 64-bit integer comparison function + */ +#define kh_int64_hash_equal(a, b) ((a) == (b)) +/*! @function + @abstract const char* hash function + @param s Pointer to a null terminated string + @return The hash value + */ +static kh_inline khint_t __ac_X31_hash_string(const char *s) +{ + khint_t h = (khint_t)*s; + if (h) for (++s ; *s; ++s) h = (h << 5) - h + (khint_t)*s; + return h; +} +/*! @function + @abstract Another interface to const char* hash function + @param key Pointer to a null terminated string [const char*] + @return The hash value [khint_t] + */ +#define kh_str_hash_func(key) __ac_X31_hash_string(key) +/*! @function + @abstract Const char* comparison function + */ +#define kh_str_hash_equal(a, b) (strcmp(a, b) == 0) + +static kh_inline khint_t __ac_Wang_hash(khint_t key) +{ + key += ~(key << 15); + key ^= (key >> 10); + key += (key << 3); + key ^= (key >> 6); + key += ~(key << 11); + key ^= (key >> 16); + return key; +} +#define kh_int_hash_func2(key) __ac_Wang_hash((khint_t)key) + +/* --- END OF HASH FUNCTIONS --- */ + +/* Other convenient macros... */ + +/*! + @abstract Type of the hash table. + @param name Name of the hash table [symbol] + */ +#define khash_t(name) kh_##name##_t + +/*! @function + @abstract Initiate a hash table. + @param name Name of the hash table [symbol] + @return Pointer to the hash table [khash_t(name)*] + */ +#define kh_init(name) kh_init_##name() + +/*! @function + @abstract Destroy a hash table. + @param name Name of the hash table [symbol] + @param h Pointer to the hash table [khash_t(name)*] + */ +#define kh_destroy(name, h) kh_destroy_##name(h) + +/*! @function + @abstract Reset a hash table without deallocating memory. + @param name Name of the hash table [symbol] + @param h Pointer to the hash table [khash_t(name)*] + */ +#define kh_clear(name, h) kh_clear_##name(h) + +/*! @function + @abstract Resize a hash table. + @param name Name of the hash table [symbol] + @param h Pointer to the hash table [khash_t(name)*] + @param s New size [khint_t] + */ +#define kh_resize(name, h, s) kh_resize_##name(h, s) + +/*! @function + @abstract Insert a key to the hash table. + @param name Name of the hash table [symbol] + @param h Pointer to the hash table [khash_t(name)*] + @param k Key [type of keys] + @param r Extra return code: -1 if the operation failed; + 0 if the key is present in the hash table; + 1 if the bucket is empty (never used); 2 if the element in + the bucket has been deleted [int*] + @return Iterator to the inserted element [khint_t] + */ +#define kh_put(name, h, k, r) kh_put_##name(h, k, r) + +/*! @function + @abstract Retrieve a key from the hash table. + @param name Name of the hash table [symbol] + @param h Pointer to the hash table [khash_t(name)*] + @param k Key [type of keys] + @return Iterator to the found element, or kh_end(h) if the element is absent [khint_t] + */ +#define kh_get(name, h, k) kh_get_##name(h, k) + +/*! @function + @abstract Remove a key from the hash table. + @param name Name of the hash table [symbol] + @param h Pointer to the hash table [khash_t(name)*] + @param k Iterator to the element to be deleted [khint_t] + */ +#define kh_del(name, h, k) kh_del_##name(h, k) + +/*! @function + @abstract Test whether a bucket contains data. + @param h Pointer to the hash table [khash_t(name)*] + @param x Iterator to the bucket [khint_t] + @return 1 if containing data; 0 otherwise [int] + */ +#define kh_exist(h, x) (!__ac_iseither((h)->flags, (x))) + +/*! @function + @abstract Get key given an iterator + @param h Pointer to the hash table [khash_t(name)*] + @param x Iterator to the bucket [khint_t] + @return Key [type of keys] + */ +#define kh_key(h, x) ((h)->keys[x]) + +/*! @function + @abstract Get value given an iterator + @param h Pointer to the hash table [khash_t(name)*] + @param x Iterator to the bucket [khint_t] + @return Value [type of values] + @discussion For hash sets, calling this results in segfault. + */ +#define kh_val(h, x) ((h)->vals[x]) + +/*! @function + @abstract Alias of kh_val() + */ +#define kh_value(h, x) ((h)->vals[x]) + +/*! @function + @abstract Get the start iterator + @param h Pointer to the hash table [khash_t(name)*] + @return The start iterator [khint_t] + */ +#define kh_begin(h) (khint_t)(0) + +/*! @function + @abstract Get the end iterator + @param h Pointer to the hash table [khash_t(name)*] + @return The end iterator [khint_t] + */ +#define kh_end(h) ((h)->n_buckets) + +/*! @function + @abstract Get the number of elements in the hash table + @param h Pointer to the hash table [khash_t(name)*] + @return Number of elements in the hash table [khint_t] + */ +#define kh_size(h) ((h)->size) + +/*! @function + @abstract Get the number of buckets in the hash table + @param h Pointer to the hash table [khash_t(name)*] + @return Number of buckets in the hash table [khint_t] + */ +#define kh_n_buckets(h) ((h)->n_buckets) + +/*! @function + @abstract Iterate over the entries in the hash table + @param h Pointer to the hash table [khash_t(name)*] + @param kvar Variable to which key will be assigned + @param vvar Variable to which value will be assigned + @param code Block of code to execute + */ +#define kh_foreach(h, kvar, vvar, code) { khint_t __i; \ + for (__i = kh_begin(h); __i != kh_end(h); ++__i) { \ + if (!kh_exist(h,__i)) continue; \ + (kvar) = kh_key(h,__i); \ + (vvar) = kh_val(h,__i); \ + code; \ + } } + +/*! @function + @abstract Iterate over the values in the hash table + @param h Pointer to the hash table [khash_t(name)*] + @param vvar Variable to which value will be assigned + @param code Block of code to execute + */ +#define kh_foreach_value(h, vvar, code) { khint_t __i; \ + for (__i = kh_begin(h); __i != kh_end(h); ++__i) { \ + if (!kh_exist(h,__i)) continue; \ + (vvar) = kh_val(h,__i); \ + code; \ + } } + +/* More conenient interfaces */ + +/*! @function + @abstract Instantiate a hash set containing integer keys + @param name Name of the hash table [symbol] + */ +#define KHASH_SET_INIT_INT(name) \ + KHASH_INIT(name, khint32_t, char, 0, kh_int_hash_func, kh_int_hash_equal) + +/*! @function + @abstract Instantiate a hash map containing integer keys + @param name Name of the hash table [symbol] + @param khval_t Type of values [type] + */ +#define KHASH_MAP_INIT_INT(name, khval_t) \ + KHASH_INIT(name, khint32_t, khval_t, 1, kh_int_hash_func, kh_int_hash_equal) + +/*! @function + @abstract Instantiate a hash map containing 64-bit integer keys + @param name Name of the hash table [symbol] + */ +#define KHASH_SET_INIT_INT64(name) \ + KHASH_INIT(name, khint64_t, char, 0, kh_int64_hash_func, kh_int64_hash_equal) + +/*! @function + @abstract Instantiate a hash map containing 64-bit integer keys + @param name Name of the hash table [symbol] + @param khval_t Type of values [type] + */ +#define KHASH_MAP_INIT_INT64(name, khval_t) \ + KHASH_INIT(name, khint64_t, khval_t, 1, kh_int64_hash_func, kh_int64_hash_equal) + +typedef const char *kh_cstr_t; +/*! @function + @abstract Instantiate a hash map containing const char* keys + @param name Name of the hash table [symbol] + */ +#define KHASH_SET_INIT_STR(name) \ + KHASH_INIT(name, kh_cstr_t, char, 0, kh_str_hash_func, kh_str_hash_equal) + +/*! @function + @abstract Instantiate a hash map containing const char* keys + @param name Name of the hash table [symbol] + @param khval_t Type of values [type] + */ +#define KHASH_MAP_INIT_STR(name, khval_t) \ + KHASH_INIT(name, kh_cstr_t, khval_t, 1, kh_str_hash_func, kh_str_hash_equal) + +#endif /* __AC_KHASH_H */ diff --git a/external/uthash/uthash.h b/external/uthash/uthash.h deleted file mode 100644 index 7205c67e..00000000 --- a/external/uthash/uthash.h +++ /dev/null @@ -1,958 +0,0 @@ -/* -Copyright (c) 2003-2014, Troy D. Hanson http://troydhanson.github.com/uthash/ -All rights reserved. - -Redistribution and use in source and binary forms, with or without -modification, are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright - notice, this list of conditions and the following disclaimer. - -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS -IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED -TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A -PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER -OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, -EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, -PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR -PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF -LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING -NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS -SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -*/ - -#ifndef UTHASH_H -#define UTHASH_H - -#include /* memcmp,strlen */ -#include /* ptrdiff_t */ -#include /* exit() */ - -/* These macros use decltype or the earlier __typeof GNU extension. - As decltype is only available in newer compilers (VS2010 or gcc 4.3+ - when compiling c++ source) this code uses whatever method is needed - or, for VS2008 where neither is available, uses casting workarounds. */ -#if defined(_MSC_VER) /* MS compiler */ -#if _MSC_VER >= 1600 && defined(__cplusplus) /* VS2010 or newer in C++ mode */ -#define DECLTYPE(x) (decltype(x)) -#else /* VS2008 or older (or VS2010 in C mode) */ -#define NO_DECLTYPE -#define DECLTYPE(x) -#endif -#elif defined(__BORLANDC__) || defined(__LCC__) || defined(__WATCOMC__) -#define NO_DECLTYPE -#define DECLTYPE(x) -#else /* GNU, Sun and other compilers */ -#define DECLTYPE(x) (__typeof(x)) -#endif - -#ifdef NO_DECLTYPE -#define DECLTYPE_ASSIGN(dst,src) \ -do { \ - char **_da_dst = (char**)(&(dst)); \ - *_da_dst = (char*)(src); \ -} while(0) -#else -#define DECLTYPE_ASSIGN(dst,src) \ -do { \ - (dst) = DECLTYPE(dst)(src); \ -} while(0) -#endif - -/* a number of the hash function use uint32_t which isn't defined on Pre VS2010 */ -#if defined (_WIN32) -#if defined(_MSC_VER) && _MSC_VER >= 1600 -#include -#elif defined(__WATCOMC__) -#include -#else -typedef unsigned int uint32_t; -typedef unsigned char uint8_t; -#endif -#else -#include -#endif - -#define UTHASH_VERSION 1.9.9 - -#ifndef uthash_fatal -#define uthash_fatal(msg) exit(-1) /* fatal error (out of memory,etc) */ -#endif -#ifndef uthash_malloc -#define uthash_malloc(sz) malloc(sz) /* malloc fcn */ -#endif -#ifndef uthash_free -#define uthash_free(ptr,sz) free(ptr) /* free fcn */ -#endif - -#ifndef uthash_noexpand_fyi -#define uthash_noexpand_fyi(tbl) /* can be defined to log noexpand */ -#endif -#ifndef uthash_expand_fyi -#define uthash_expand_fyi(tbl) /* can be defined to log expands */ -#endif - -/* initial number of buckets */ -#define HASH_INITIAL_NUM_BUCKETS 32 /* initial number of buckets */ -#define HASH_INITIAL_NUM_BUCKETS_LOG2 5 /* lg2 of initial number of buckets */ -#define HASH_BKT_CAPACITY_THRESH 10 /* expand when bucket count reaches */ - -/* calculate the element whose hash handle address is hhe */ -#define ELMT_FROM_HH(tbl,hhp) ((void*)(((char*)(hhp)) - ((tbl)->hho))) - -#define HASH_FIND(hh,head,keyptr,keylen,out) \ -do { \ - unsigned _hf_bkt,_hf_hashv; \ - out=NULL; \ - if (head) { \ - HASH_FCN(keyptr,keylen, (head)->hh.tbl->num_buckets, _hf_hashv, _hf_bkt); \ - if (HASH_BLOOM_TEST((head)->hh.tbl, _hf_hashv)) { \ - HASH_FIND_IN_BKT((head)->hh.tbl, hh, (head)->hh.tbl->buckets[ _hf_bkt ], \ - keyptr,keylen,out); \ - } \ - } \ -} while (0) - -#ifdef HASH_BLOOM -#define HASH_BLOOM_BITLEN (1ULL << HASH_BLOOM) -#define HASH_BLOOM_BYTELEN (HASH_BLOOM_BITLEN/8) + ((HASH_BLOOM_BITLEN%8) ? 1:0) -#define HASH_BLOOM_MAKE(tbl) \ -do { \ - (tbl)->bloom_nbits = HASH_BLOOM; \ - (tbl)->bloom_bv = (uint8_t*)uthash_malloc(HASH_BLOOM_BYTELEN); \ - if (!((tbl)->bloom_bv)) { uthash_fatal( "out of memory"); } \ - memset((tbl)->bloom_bv, 0, HASH_BLOOM_BYTELEN); \ - (tbl)->bloom_sig = HASH_BLOOM_SIGNATURE; \ -} while (0) - -#define HASH_BLOOM_FREE(tbl) \ -do { \ - uthash_free((tbl)->bloom_bv, HASH_BLOOM_BYTELEN); \ -} while (0) - -#define HASH_BLOOM_BITSET(bv,idx) (bv[(idx)/8] |= (1U << ((idx)%8))) -#define HASH_BLOOM_BITTEST(bv,idx) (bv[(idx)/8] & (1U << ((idx)%8))) - -#define HASH_BLOOM_ADD(tbl,hashv) \ - HASH_BLOOM_BITSET((tbl)->bloom_bv, (hashv & (uint32_t)((1ULL << (tbl)->bloom_nbits) - 1))) - -#define HASH_BLOOM_TEST(tbl,hashv) \ - HASH_BLOOM_BITTEST((tbl)->bloom_bv, (hashv & (uint32_t)((1ULL << (tbl)->bloom_nbits) - 1))) - -#else -#define HASH_BLOOM_MAKE(tbl) -#define HASH_BLOOM_FREE(tbl) -#define HASH_BLOOM_ADD(tbl,hashv) -#define HASH_BLOOM_TEST(tbl,hashv) (1) -#define HASH_BLOOM_BYTELEN 0 -#endif - -#define HASH_MAKE_TABLE(hh,head) \ -do { \ - (head)->hh.tbl = (UT_hash_table*)uthash_malloc( \ - sizeof(UT_hash_table)); \ - if (!((head)->hh.tbl)) { uthash_fatal( "out of memory"); } \ - memset((head)->hh.tbl, 0, sizeof(UT_hash_table)); \ - (head)->hh.tbl->tail = &((head)->hh); \ - (head)->hh.tbl->num_buckets = HASH_INITIAL_NUM_BUCKETS; \ - (head)->hh.tbl->log2_num_buckets = HASH_INITIAL_NUM_BUCKETS_LOG2; \ - (head)->hh.tbl->hho = (char*)(&(head)->hh) - (char*)(head); \ - (head)->hh.tbl->buckets = (UT_hash_bucket*)uthash_malloc( \ - HASH_INITIAL_NUM_BUCKETS*sizeof(struct UT_hash_bucket)); \ - if (! (head)->hh.tbl->buckets) { uthash_fatal( "out of memory"); } \ - memset((head)->hh.tbl->buckets, 0, \ - HASH_INITIAL_NUM_BUCKETS*sizeof(struct UT_hash_bucket)); \ - HASH_BLOOM_MAKE((head)->hh.tbl); \ - (head)->hh.tbl->signature = HASH_SIGNATURE; \ -} while(0) - -#define HASH_ADD(hh,head,fieldname,keylen_in,add) \ - HASH_ADD_KEYPTR(hh,head,&((add)->fieldname),keylen_in,add) - -#define HASH_REPLACE(hh,head,fieldname,keylen_in,add,replaced) \ -do { \ - replaced=NULL; \ - HASH_FIND(hh,head,&((add)->fieldname),keylen_in,replaced); \ - if (replaced!=NULL) { \ - HASH_DELETE(hh,head,replaced); \ - }; \ - HASH_ADD(hh,head,fieldname,keylen_in,add); \ -} while(0) - -#define HASH_ADD_KEYPTR(hh,head,keyptr,keylen_in,add) \ -do { \ - unsigned _ha_bkt; \ - (add)->hh.next = NULL; \ - (add)->hh.key = (char*)(keyptr); \ - (add)->hh.keylen = (unsigned)(keylen_in); \ - if (!(head)) { \ - head = (add); \ - (head)->hh.prev = NULL; \ - HASH_MAKE_TABLE(hh,head); \ - } else { \ - (head)->hh.tbl->tail->next = (add); \ - (add)->hh.prev = ELMT_FROM_HH((head)->hh.tbl, (head)->hh.tbl->tail); \ - (head)->hh.tbl->tail = &((add)->hh); \ - } \ - (head)->hh.tbl->num_items++; \ - (add)->hh.tbl = (head)->hh.tbl; \ - HASH_FCN(keyptr,keylen_in, (head)->hh.tbl->num_buckets, \ - (add)->hh.hashv, _ha_bkt); \ - HASH_ADD_TO_BKT((head)->hh.tbl->buckets[_ha_bkt],&(add)->hh); \ - HASH_BLOOM_ADD((head)->hh.tbl,(add)->hh.hashv); \ - HASH_EMIT_KEY(hh,head,keyptr,keylen_in); \ - HASH_FSCK(hh,head); \ -} while(0) - -#define HASH_TO_BKT( hashv, num_bkts, bkt ) \ -do { \ - bkt = ((hashv) & ((num_bkts) - 1)); \ -} while(0) - -/* delete "delptr" from the hash table. - * "the usual" patch-up process for the app-order doubly-linked-list. - * The use of _hd_hh_del below deserves special explanation. - * These used to be expressed using (delptr) but that led to a bug - * if someone used the same symbol for the head and deletee, like - * HASH_DELETE(hh,users,users); - * We want that to work, but by changing the head (users) below - * we were forfeiting our ability to further refer to the deletee (users) - * in the patch-up process. Solution: use scratch space to - * copy the deletee pointer, then the latter references are via that - * scratch pointer rather than through the repointed (users) symbol. - */ -#define HASH_DELETE(hh,head,delptr) \ -do { \ - unsigned _hd_bkt; \ - struct UT_hash_handle *_hd_hh_del; \ - if ( ((delptr)->hh.prev == NULL) && ((delptr)->hh.next == NULL) ) { \ - uthash_free((head)->hh.tbl->buckets, \ - (head)->hh.tbl->num_buckets*sizeof(struct UT_hash_bucket) ); \ - HASH_BLOOM_FREE((head)->hh.tbl); \ - uthash_free((head)->hh.tbl, sizeof(UT_hash_table)); \ - head = NULL; \ - } else { \ - _hd_hh_del = &((delptr)->hh); \ - if ((delptr) == ELMT_FROM_HH((head)->hh.tbl,(head)->hh.tbl->tail)) { \ - (head)->hh.tbl->tail = \ - (UT_hash_handle*)((ptrdiff_t)((delptr)->hh.prev) + \ - (head)->hh.tbl->hho); \ - } \ - if ((delptr)->hh.prev) { \ - ((UT_hash_handle*)((ptrdiff_t)((delptr)->hh.prev) + \ - (head)->hh.tbl->hho))->next = (delptr)->hh.next; \ - } else { \ - DECLTYPE_ASSIGN(head,(delptr)->hh.next); \ - } \ - if (_hd_hh_del->next) { \ - ((UT_hash_handle*)((ptrdiff_t)_hd_hh_del->next + \ - (head)->hh.tbl->hho))->prev = \ - _hd_hh_del->prev; \ - } \ - HASH_TO_BKT( _hd_hh_del->hashv, (head)->hh.tbl->num_buckets, _hd_bkt); \ - HASH_DEL_IN_BKT(hh,(head)->hh.tbl->buckets[_hd_bkt], _hd_hh_del); \ - (head)->hh.tbl->num_items--; \ - } \ - HASH_FSCK(hh,head); \ -} while (0) - - -/* convenience forms of HASH_FIND/HASH_ADD/HASH_DEL */ -#define HASH_FIND_STR(head,findstr,out) \ - HASH_FIND(hh,head,findstr,strlen(findstr),out) -#define HASH_ADD_STR(head,strfield,add) \ - HASH_ADD(hh,head,strfield[0],strlen(add->strfield),add) -#define HASH_REPLACE_STR(head,strfield,add,replaced) \ - HASH_REPLACE(hh,head,strfield[0],strlen(add->strfield),add,replaced) -#define HASH_FIND_INT(head,findint,out) \ - HASH_FIND(hh,head,findint,sizeof(int),out) -#define HASH_ADD_INT(head,intfield,add) \ - HASH_ADD(hh,head,intfield,sizeof(int),add) -#define HASH_REPLACE_INT(head,intfield,add,replaced) \ - HASH_REPLACE(hh,head,intfield,sizeof(int),add,replaced) -#define HASH_FIND_PTR(head,findptr,out) \ - HASH_FIND(hh,head,findptr,sizeof(void *),out) -#define HASH_ADD_PTR(head,ptrfield,add) \ - HASH_ADD(hh,head,ptrfield,sizeof(void *),add) -#define HASH_REPLACE_PTR(head,ptrfield,add,replaced) \ - HASH_REPLACE(hh,head,ptrfield,sizeof(void *),add,replaced) -#define HASH_DEL(head,delptr) \ - HASH_DELETE(hh,head,delptr) - -/* HASH_FSCK checks hash integrity on every add/delete when HASH_DEBUG is defined. - * This is for uthash developer only; it compiles away if HASH_DEBUG isn't defined. - */ -#ifdef HASH_DEBUG -#define HASH_OOPS(...) do { fprintf(stderr,__VA_ARGS__); exit(-1); } while (0) -#define HASH_FSCK(hh,head) \ -do { \ - unsigned _bkt_i; \ - unsigned _count, _bkt_count; \ - char *_prev; \ - struct UT_hash_handle *_thh; \ - if (head) { \ - _count = 0; \ - for( _bkt_i = 0; _bkt_i < (head)->hh.tbl->num_buckets; _bkt_i++) { \ - _bkt_count = 0; \ - _thh = (head)->hh.tbl->buckets[_bkt_i].hh_head; \ - _prev = NULL; \ - while (_thh) { \ - if (_prev != (char*)(_thh->hh_prev)) { \ - HASH_OOPS("invalid hh_prev %p, actual %p\n", \ - _thh->hh_prev, _prev ); \ - } \ - _bkt_count++; \ - _prev = (char*)(_thh); \ - _thh = _thh->hh_next; \ - } \ - _count += _bkt_count; \ - if ((head)->hh.tbl->buckets[_bkt_i].count != _bkt_count) { \ - HASH_OOPS("invalid bucket count %d, actual %d\n", \ - (head)->hh.tbl->buckets[_bkt_i].count, _bkt_count); \ - } \ - } \ - if (_count != (head)->hh.tbl->num_items) { \ - HASH_OOPS("invalid hh item count %d, actual %d\n", \ - (head)->hh.tbl->num_items, _count ); \ - } \ - /* traverse hh in app order; check next/prev integrity, count */ \ - _count = 0; \ - _prev = NULL; \ - _thh = &(head)->hh; \ - while (_thh) { \ - _count++; \ - if (_prev !=(char*)(_thh->prev)) { \ - HASH_OOPS("invalid prev %p, actual %p\n", \ - _thh->prev, _prev ); \ - } \ - _prev = (char*)ELMT_FROM_HH((head)->hh.tbl, _thh); \ - _thh = ( _thh->next ? (UT_hash_handle*)((char*)(_thh->next) + \ - (head)->hh.tbl->hho) : NULL ); \ - } \ - if (_count != (head)->hh.tbl->num_items) { \ - HASH_OOPS("invalid app item count %d, actual %d\n", \ - (head)->hh.tbl->num_items, _count ); \ - } \ - } \ -} while (0) -#else -#define HASH_FSCK(hh,head) -#endif - -/* When compiled with -DHASH_EMIT_KEYS, length-prefixed keys are emitted to - * the descriptor to which this macro is defined for tuning the hash function. - * The app can #include to get the prototype for write(2). */ -#ifdef HASH_EMIT_KEYS -#define HASH_EMIT_KEY(hh,head,keyptr,fieldlen) \ -do { \ - unsigned _klen = fieldlen; \ - write(HASH_EMIT_KEYS, &_klen, sizeof(_klen)); \ - write(HASH_EMIT_KEYS, keyptr, fieldlen); \ -} while (0) -#else -#define HASH_EMIT_KEY(hh,head,keyptr,fieldlen) -#endif - -/* default to Jenkin's hash unless overridden e.g. DHASH_FUNCTION=HASH_SAX */ -#ifdef HASH_FUNCTION -#define HASH_FCN HASH_FUNCTION -#else -#define HASH_FCN HASH_JEN -#endif - -/* The Bernstein hash function, used in Perl prior to v5.6. Note (x<<5+x)=x*33. */ -#define HASH_BER(key,keylen,num_bkts,hashv,bkt) \ -do { \ - unsigned _hb_keylen=keylen; \ - char *_hb_key=(char*)(key); \ - (hashv) = 0; \ - while (_hb_keylen--) { (hashv) = (((hashv) << 5) + (hashv)) + *_hb_key++; } \ - bkt = (hashv) & (num_bkts-1); \ -} while (0) - - -/* SAX/FNV/OAT/JEN hash functions are macro variants of those listed at - * http://eternallyconfuzzled.com/tuts/algorithms/jsw_tut_hashing.aspx */ -#define HASH_SAX(key,keylen,num_bkts,hashv,bkt) \ -do { \ - unsigned _sx_i; \ - char *_hs_key=(char*)(key); \ - hashv = 0; \ - for(_sx_i=0; _sx_i < keylen; _sx_i++) \ - hashv ^= (hashv << 5) + (hashv >> 2) + _hs_key[_sx_i]; \ - bkt = hashv & (num_bkts-1); \ -} while (0) -/* FNV-1a variation */ -#define HASH_FNV(key,keylen,num_bkts,hashv,bkt) \ -do { \ - unsigned _fn_i; \ - char *_hf_key=(char*)(key); \ - hashv = 2166136261UL; \ - for(_fn_i=0; _fn_i < keylen; _fn_i++) \ - hashv = hashv ^ _hf_key[_fn_i]; \ - hashv = hashv * 16777619; \ - bkt = hashv & (num_bkts-1); \ -} while(0) - -#define HASH_OAT(key,keylen,num_bkts,hashv,bkt) \ -do { \ - unsigned _ho_i; \ - char *_ho_key=(char*)(key); \ - hashv = 0; \ - for(_ho_i=0; _ho_i < keylen; _ho_i++) { \ - hashv += _ho_key[_ho_i]; \ - hashv += (hashv << 10); \ - hashv ^= (hashv >> 6); \ - } \ - hashv += (hashv << 3); \ - hashv ^= (hashv >> 11); \ - hashv += (hashv << 15); \ - bkt = hashv & (num_bkts-1); \ -} while(0) - -#define HASH_JEN_MIX(a,b,c) \ -do { \ - a -= b; a -= c; a ^= ( c >> 13 ); \ - b -= c; b -= a; b ^= ( a << 8 ); \ - c -= a; c -= b; c ^= ( b >> 13 ); \ - a -= b; a -= c; a ^= ( c >> 12 ); \ - b -= c; b -= a; b ^= ( a << 16 ); \ - c -= a; c -= b; c ^= ( b >> 5 ); \ - a -= b; a -= c; a ^= ( c >> 3 ); \ - b -= c; b -= a; b ^= ( a << 10 ); \ - c -= a; c -= b; c ^= ( b >> 15 ); \ -} while (0) - -#define HASH_JEN(key,keylen,num_bkts,hashv,bkt) \ -do { \ - unsigned _hj_i,_hj_j,_hj_k; \ - unsigned char *_hj_key=(unsigned char*)(key); \ - hashv = 0xfeedbeef; \ - _hj_i = _hj_j = 0x9e3779b9; \ - _hj_k = (unsigned)(keylen); \ - while (_hj_k >= 12) { \ - _hj_i += (_hj_key[0] + ( (unsigned)_hj_key[1] << 8 ) \ - + ( (unsigned)_hj_key[2] << 16 ) \ - + ( (unsigned)_hj_key[3] << 24 ) ); \ - _hj_j += (_hj_key[4] + ( (unsigned)_hj_key[5] << 8 ) \ - + ( (unsigned)_hj_key[6] << 16 ) \ - + ( (unsigned)_hj_key[7] << 24 ) ); \ - hashv += (_hj_key[8] + ( (unsigned)_hj_key[9] << 8 ) \ - + ( (unsigned)_hj_key[10] << 16 ) \ - + ( (unsigned)_hj_key[11] << 24 ) ); \ - \ - HASH_JEN_MIX(_hj_i, _hj_j, hashv); \ - \ - _hj_key += 12; \ - _hj_k -= 12; \ - } \ - hashv += keylen; \ - switch ( _hj_k ) { \ - case 11: hashv += ( (unsigned)_hj_key[10] << 24 ); \ - case 10: hashv += ( (unsigned)_hj_key[9] << 16 ); \ - case 9: hashv += ( (unsigned)_hj_key[8] << 8 ); \ - case 8: _hj_j += ( (unsigned)_hj_key[7] << 24 ); \ - case 7: _hj_j += ( (unsigned)_hj_key[6] << 16 ); \ - case 6: _hj_j += ( (unsigned)_hj_key[5] << 8 ); \ - case 5: _hj_j += _hj_key[4]; \ - case 4: _hj_i += ( (unsigned)_hj_key[3] << 24 ); \ - case 3: _hj_i += ( (unsigned)_hj_key[2] << 16 ); \ - case 2: _hj_i += ( (unsigned)_hj_key[1] << 8 ); \ - case 1: _hj_i += _hj_key[0]; \ - } \ - HASH_JEN_MIX(_hj_i, _hj_j, hashv); \ - bkt = hashv & (num_bkts-1); \ -} while(0) - -/* The Paul Hsieh hash function */ -#undef get16bits -#if (defined(__GNUC__) && defined(__i386__)) || defined(__WATCOMC__) \ - || defined(_MSC_VER) || defined (__BORLANDC__) || defined (__TURBOC__) -#define get16bits(d) (*((const uint16_t *) (d))) -#endif - -#if !defined (get16bits) -#define get16bits(d) ((((uint32_t)(((const uint8_t *)(d))[1])) << 8) \ - +(uint32_t)(((const uint8_t *)(d))[0]) ) -#endif -#define HASH_SFH(key,keylen,num_bkts,hashv,bkt) \ -do { \ - unsigned char *_sfh_key=(unsigned char*)(key); \ - uint32_t _sfh_tmp, _sfh_len = keylen; \ - \ - int _sfh_rem = _sfh_len & 3; \ - _sfh_len >>= 2; \ - hashv = 0xcafebabe; \ - \ - /* Main loop */ \ - for (;_sfh_len > 0; _sfh_len--) { \ - hashv += get16bits (_sfh_key); \ - _sfh_tmp = (uint32_t)(get16bits (_sfh_key+2)) << 11 ^ hashv; \ - hashv = (hashv << 16) ^ _sfh_tmp; \ - _sfh_key += 2*sizeof (uint16_t); \ - hashv += hashv >> 11; \ - } \ - \ - /* Handle end cases */ \ - switch (_sfh_rem) { \ - case 3: hashv += get16bits (_sfh_key); \ - hashv ^= hashv << 16; \ - hashv ^= (uint32_t)(_sfh_key[sizeof (uint16_t)] << 18); \ - hashv += hashv >> 11; \ - break; \ - case 2: hashv += get16bits (_sfh_key); \ - hashv ^= hashv << 11; \ - hashv += hashv >> 17; \ - break; \ - case 1: hashv += *_sfh_key; \ - hashv ^= hashv << 10; \ - hashv += hashv >> 1; \ - } \ - \ - /* Force "avalanching" of final 127 bits */ \ - hashv ^= hashv << 3; \ - hashv += hashv >> 5; \ - hashv ^= hashv << 4; \ - hashv += hashv >> 17; \ - hashv ^= hashv << 25; \ - hashv += hashv >> 6; \ - bkt = hashv & (num_bkts-1); \ -} while(0) - -#ifdef HASH_USING_NO_STRICT_ALIASING -/* The MurmurHash exploits some CPU's (x86,x86_64) tolerance for unaligned reads. - * For other types of CPU's (e.g. Sparc) an unaligned read causes a bus error. - * MurmurHash uses the faster approach only on CPU's where we know it's safe. - * - * Note the preprocessor built-in defines can be emitted using: - * - * gcc -m64 -dM -E - < /dev/null (on gcc) - * cc -## a.c (where a.c is a simple test file) (Sun Studio) - */ -#if (defined(__i386__) || defined(__x86_64__) || defined(_M_IX86)) -#define MUR_GETBLOCK(p,i) p[i] -#else /* non intel */ -#define MUR_PLUS0_ALIGNED(p) (((unsigned long)p & 0x3) == 0) -#define MUR_PLUS1_ALIGNED(p) (((unsigned long)p & 0x3) == 1) -#define MUR_PLUS2_ALIGNED(p) (((unsigned long)p & 0x3) == 2) -#define MUR_PLUS3_ALIGNED(p) (((unsigned long)p & 0x3) == 3) -#define WP(p) ((uint32_t*)((unsigned long)(p) & ~3UL)) -#if (defined(__BIG_ENDIAN__) || defined(SPARC) || defined(__ppc__) || defined(__ppc64__)) -#define MUR_THREE_ONE(p) ((((*WP(p))&0x00ffffff) << 8) | (((*(WP(p)+1))&0xff000000) >> 24)) -#define MUR_TWO_TWO(p) ((((*WP(p))&0x0000ffff) <<16) | (((*(WP(p)+1))&0xffff0000) >> 16)) -#define MUR_ONE_THREE(p) ((((*WP(p))&0x000000ff) <<24) | (((*(WP(p)+1))&0xffffff00) >> 8)) -#else /* assume little endian non-intel */ -#define MUR_THREE_ONE(p) ((((*WP(p))&0xffffff00) >> 8) | (((*(WP(p)+1))&0x000000ff) << 24)) -#define MUR_TWO_TWO(p) ((((*WP(p))&0xffff0000) >>16) | (((*(WP(p)+1))&0x0000ffff) << 16)) -#define MUR_ONE_THREE(p) ((((*WP(p))&0xff000000) >>24) | (((*(WP(p)+1))&0x00ffffff) << 8)) -#endif -#define MUR_GETBLOCK(p,i) (MUR_PLUS0_ALIGNED(p) ? ((p)[i]) : \ - (MUR_PLUS1_ALIGNED(p) ? MUR_THREE_ONE(p) : \ - (MUR_PLUS2_ALIGNED(p) ? MUR_TWO_TWO(p) : \ - MUR_ONE_THREE(p)))) -#endif -#define MUR_ROTL32(x,r) (((x) << (r)) | ((x) >> (32 - (r)))) -#define MUR_FMIX(_h) \ -do { \ - _h ^= _h >> 16; \ - _h *= 0x85ebca6b; \ - _h ^= _h >> 13; \ - _h *= 0xc2b2ae35l; \ - _h ^= _h >> 16; \ -} while(0) - -#define HASH_MUR(key,keylen,num_bkts,hashv,bkt) \ -do { \ - const uint8_t *_mur_data = (const uint8_t*)(key); \ - const int _mur_nblocks = (keylen) / 4; \ - uint32_t _mur_h1 = 0xf88D5353; \ - uint32_t _mur_c1 = 0xcc9e2d51; \ - uint32_t _mur_c2 = 0x1b873593; \ - uint32_t _mur_k1 = 0; \ - const uint8_t *_mur_tail; \ - const uint32_t *_mur_blocks = (const uint32_t*)(_mur_data+_mur_nblocks*4); \ - int _mur_i; \ - for(_mur_i = -_mur_nblocks; _mur_i; _mur_i++) { \ - _mur_k1 = MUR_GETBLOCK(_mur_blocks,_mur_i); \ - _mur_k1 *= _mur_c1; \ - _mur_k1 = MUR_ROTL32(_mur_k1,15); \ - _mur_k1 *= _mur_c2; \ - \ - _mur_h1 ^= _mur_k1; \ - _mur_h1 = MUR_ROTL32(_mur_h1,13); \ - _mur_h1 = _mur_h1*5+0xe6546b64; \ - } \ - _mur_tail = (const uint8_t*)(_mur_data + _mur_nblocks*4); \ - _mur_k1=0; \ - switch((keylen) & 3) { \ - case 3: _mur_k1 ^= _mur_tail[2] << 16; \ - case 2: _mur_k1 ^= _mur_tail[1] << 8; \ - case 1: _mur_k1 ^= _mur_tail[0]; \ - _mur_k1 *= _mur_c1; \ - _mur_k1 = MUR_ROTL32(_mur_k1,15); \ - _mur_k1 *= _mur_c2; \ - _mur_h1 ^= _mur_k1; \ - } \ - _mur_h1 ^= (keylen); \ - MUR_FMIX(_mur_h1); \ - hashv = _mur_h1; \ - bkt = hashv & (num_bkts-1); \ -} while(0) -#endif /* HASH_USING_NO_STRICT_ALIASING */ - -/* key comparison function; return 0 if keys equal */ -#define HASH_KEYCMP(a,b,len) memcmp(a,b,len) - -/* iterate over items in a known bucket to find desired item */ -#define HASH_FIND_IN_BKT(tbl,hh,head,keyptr,keylen_in,out) \ -do { \ - if (head.hh_head) DECLTYPE_ASSIGN(out,ELMT_FROM_HH(tbl,head.hh_head)); \ - else out=NULL; \ - while (out) { \ - if ((out)->hh.keylen == keylen_in) { \ - if ((HASH_KEYCMP((out)->hh.key,keyptr,keylen_in)) == 0) break; \ - } \ - if ((out)->hh.hh_next) DECLTYPE_ASSIGN(out,ELMT_FROM_HH(tbl,(out)->hh.hh_next)); \ - else out = NULL; \ - } \ -} while(0) - -/* add an item to a bucket */ -#define HASH_ADD_TO_BKT(head,addhh) \ -do { \ - head.count++; \ - (addhh)->hh_next = head.hh_head; \ - (addhh)->hh_prev = NULL; \ - if (head.hh_head) { (head).hh_head->hh_prev = (addhh); } \ - (head).hh_head=addhh; \ - if (head.count >= ((head.expand_mult+1) * HASH_BKT_CAPACITY_THRESH) \ - && (addhh)->tbl->noexpand != 1) { \ - HASH_EXPAND_BUCKETS((addhh)->tbl); \ - } \ -} while(0) - -/* remove an item from a given bucket */ -#define HASH_DEL_IN_BKT(hh,head,hh_del) \ - (head).count--; \ - if ((head).hh_head == hh_del) { \ - (head).hh_head = hh_del->hh_next; \ - } \ - if (hh_del->hh_prev) { \ - hh_del->hh_prev->hh_next = hh_del->hh_next; \ - } \ - if (hh_del->hh_next) { \ - hh_del->hh_next->hh_prev = hh_del->hh_prev; \ - } - -/* Bucket expansion has the effect of doubling the number of buckets - * and redistributing the items into the new buckets. Ideally the - * items will distribute more or less evenly into the new buckets - * (the extent to which this is true is a measure of the quality of - * the hash function as it applies to the key domain). - * - * With the items distributed into more buckets, the chain length - * (item count) in each bucket is reduced. Thus by expanding buckets - * the hash keeps a bound on the chain length. This bounded chain - * length is the essence of how a hash provides constant time lookup. - * - * The calculation of tbl->ideal_chain_maxlen below deserves some - * explanation. First, keep in mind that we're calculating the ideal - * maximum chain length based on the *new* (doubled) bucket count. - * In fractions this is just n/b (n=number of items,b=new num buckets). - * Since the ideal chain length is an integer, we want to calculate - * ceil(n/b). We don't depend on floating point arithmetic in this - * hash, so to calculate ceil(n/b) with integers we could write - * - * ceil(n/b) = (n/b) + ((n%b)?1:0) - * - * and in fact a previous version of this hash did just that. - * But now we have improved things a bit by recognizing that b is - * always a power of two. We keep its base 2 log handy (call it lb), - * so now we can write this with a bit shift and logical AND: - * - * ceil(n/b) = (n>>lb) + ( (n & (b-1)) ? 1:0) - * - */ -#define HASH_EXPAND_BUCKETS(tbl) \ -do { \ - unsigned _he_bkt; \ - unsigned _he_bkt_i; \ - struct UT_hash_handle *_he_thh, *_he_hh_nxt; \ - UT_hash_bucket *_he_new_buckets, *_he_newbkt; \ - _he_new_buckets = (UT_hash_bucket*)uthash_malloc( \ - 2 * tbl->num_buckets * sizeof(struct UT_hash_bucket)); \ - if (!_he_new_buckets) { uthash_fatal( "out of memory"); } \ - memset(_he_new_buckets, 0, \ - 2 * tbl->num_buckets * sizeof(struct UT_hash_bucket)); \ - tbl->ideal_chain_maxlen = \ - (tbl->num_items >> (tbl->log2_num_buckets+1)) + \ - ((tbl->num_items & ((tbl->num_buckets*2)-1)) ? 1 : 0); \ - tbl->nonideal_items = 0; \ - for(_he_bkt_i = 0; _he_bkt_i < tbl->num_buckets; _he_bkt_i++) \ - { \ - _he_thh = tbl->buckets[ _he_bkt_i ].hh_head; \ - while (_he_thh) { \ - _he_hh_nxt = _he_thh->hh_next; \ - HASH_TO_BKT( _he_thh->hashv, tbl->num_buckets*2, _he_bkt); \ - _he_newbkt = &(_he_new_buckets[ _he_bkt ]); \ - if (++(_he_newbkt->count) > tbl->ideal_chain_maxlen) { \ - tbl->nonideal_items++; \ - _he_newbkt->expand_mult = _he_newbkt->count / \ - tbl->ideal_chain_maxlen; \ - } \ - _he_thh->hh_prev = NULL; \ - _he_thh->hh_next = _he_newbkt->hh_head; \ - if (_he_newbkt->hh_head) _he_newbkt->hh_head->hh_prev = \ - _he_thh; \ - _he_newbkt->hh_head = _he_thh; \ - _he_thh = _he_hh_nxt; \ - } \ - } \ - uthash_free( tbl->buckets, tbl->num_buckets*sizeof(struct UT_hash_bucket) ); \ - tbl->num_buckets *= 2; \ - tbl->log2_num_buckets++; \ - tbl->buckets = _he_new_buckets; \ - tbl->ineff_expands = (tbl->nonideal_items > (tbl->num_items >> 1)) ? \ - (tbl->ineff_expands+1) : 0; \ - if (tbl->ineff_expands > 1) { \ - tbl->noexpand=1; \ - uthash_noexpand_fyi(tbl); \ - } \ - uthash_expand_fyi(tbl); \ -} while(0) - - -/* This is an adaptation of Simon Tatham's O(n log(n)) mergesort */ -/* Note that HASH_SORT assumes the hash handle name to be hh. - * HASH_SRT was added to allow the hash handle name to be passed in. */ -#define HASH_SORT(head,cmpfcn) HASH_SRT(hh,head,cmpfcn) -#define HASH_SRT(hh,head,cmpfcn) \ -do { \ - unsigned _hs_i; \ - unsigned _hs_looping,_hs_nmerges,_hs_insize,_hs_psize,_hs_qsize; \ - struct UT_hash_handle *_hs_p, *_hs_q, *_hs_e, *_hs_list, *_hs_tail; \ - if (head) { \ - _hs_insize = 1; \ - _hs_looping = 1; \ - _hs_list = &((head)->hh); \ - while (_hs_looping) { \ - _hs_p = _hs_list; \ - _hs_list = NULL; \ - _hs_tail = NULL; \ - _hs_nmerges = 0; \ - while (_hs_p) { \ - _hs_nmerges++; \ - _hs_q = _hs_p; \ - _hs_psize = 0; \ - for ( _hs_i = 0; _hs_i < _hs_insize; _hs_i++ ) { \ - _hs_psize++; \ - _hs_q = (UT_hash_handle*)((_hs_q->next) ? \ - ((void*)((char*)(_hs_q->next) + \ - (head)->hh.tbl->hho)) : NULL); \ - if (! (_hs_q) ) break; \ - } \ - _hs_qsize = _hs_insize; \ - while ((_hs_psize > 0) || ((_hs_qsize > 0) && _hs_q )) { \ - if (_hs_psize == 0) { \ - _hs_e = _hs_q; \ - _hs_q = (UT_hash_handle*)((_hs_q->next) ? \ - ((void*)((char*)(_hs_q->next) + \ - (head)->hh.tbl->hho)) : NULL); \ - _hs_qsize--; \ - } else if ( (_hs_qsize == 0) || !(_hs_q) ) { \ - _hs_e = _hs_p; \ - if (_hs_p){ \ - _hs_p = (UT_hash_handle*)((_hs_p->next) ? \ - ((void*)((char*)(_hs_p->next) + \ - (head)->hh.tbl->hho)) : NULL); \ - } \ - _hs_psize--; \ - } else if (( \ - cmpfcn(DECLTYPE(head)(ELMT_FROM_HH((head)->hh.tbl,_hs_p)), \ - DECLTYPE(head)(ELMT_FROM_HH((head)->hh.tbl,_hs_q))) \ - ) <= 0) { \ - _hs_e = _hs_p; \ - if (_hs_p){ \ - _hs_p = (UT_hash_handle*)((_hs_p->next) ? \ - ((void*)((char*)(_hs_p->next) + \ - (head)->hh.tbl->hho)) : NULL); \ - } \ - _hs_psize--; \ - } else { \ - _hs_e = _hs_q; \ - _hs_q = (UT_hash_handle*)((_hs_q->next) ? \ - ((void*)((char*)(_hs_q->next) + \ - (head)->hh.tbl->hho)) : NULL); \ - _hs_qsize--; \ - } \ - if ( _hs_tail ) { \ - _hs_tail->next = ((_hs_e) ? \ - ELMT_FROM_HH((head)->hh.tbl,_hs_e) : NULL); \ - } else { \ - _hs_list = _hs_e; \ - } \ - if (_hs_e) { \ - _hs_e->prev = ((_hs_tail) ? \ - ELMT_FROM_HH((head)->hh.tbl,_hs_tail) : NULL); \ - } \ - _hs_tail = _hs_e; \ - } \ - _hs_p = _hs_q; \ - } \ - if (_hs_tail){ \ - _hs_tail->next = NULL; \ - } \ - if ( _hs_nmerges <= 1 ) { \ - _hs_looping=0; \ - (head)->hh.tbl->tail = _hs_tail; \ - DECLTYPE_ASSIGN(head,ELMT_FROM_HH((head)->hh.tbl, _hs_list)); \ - } \ - _hs_insize *= 2; \ - } \ - HASH_FSCK(hh,head); \ - } \ -} while (0) - -/* This function selects items from one hash into another hash. - * The end result is that the selected items have dual presence - * in both hashes. There is no copy of the items made; rather - * they are added into the new hash through a secondary hash - * hash handle that must be present in the structure. */ -#define HASH_SELECT(hh_dst, dst, hh_src, src, cond) \ -do { \ - unsigned _src_bkt, _dst_bkt; \ - void *_last_elt=NULL, *_elt; \ - UT_hash_handle *_src_hh, *_dst_hh, *_last_elt_hh=NULL; \ - ptrdiff_t _dst_hho = ((char*)(&(dst)->hh_dst) - (char*)(dst)); \ - if (src) { \ - for(_src_bkt=0; _src_bkt < (src)->hh_src.tbl->num_buckets; _src_bkt++) { \ - for(_src_hh = (src)->hh_src.tbl->buckets[_src_bkt].hh_head; \ - _src_hh; \ - _src_hh = _src_hh->hh_next) { \ - _elt = ELMT_FROM_HH((src)->hh_src.tbl, _src_hh); \ - if (cond(_elt)) { \ - _dst_hh = (UT_hash_handle*)(((char*)_elt) + _dst_hho); \ - _dst_hh->key = _src_hh->key; \ - _dst_hh->keylen = _src_hh->keylen; \ - _dst_hh->hashv = _src_hh->hashv; \ - _dst_hh->prev = _last_elt; \ - _dst_hh->next = NULL; \ - if (_last_elt_hh) { _last_elt_hh->next = _elt; } \ - if (!dst) { \ - DECLTYPE_ASSIGN(dst,_elt); \ - HASH_MAKE_TABLE(hh_dst,dst); \ - } else { \ - _dst_hh->tbl = (dst)->hh_dst.tbl; \ - } \ - HASH_TO_BKT(_dst_hh->hashv, _dst_hh->tbl->num_buckets, _dst_bkt); \ - HASH_ADD_TO_BKT(_dst_hh->tbl->buckets[_dst_bkt],_dst_hh); \ - (dst)->hh_dst.tbl->num_items++; \ - _last_elt = _elt; \ - _last_elt_hh = _dst_hh; \ - } \ - } \ - } \ - } \ - HASH_FSCK(hh_dst,dst); \ -} while (0) - -#define HASH_CLEAR(hh,head) \ -do { \ - if (head) { \ - uthash_free((head)->hh.tbl->buckets, \ - (head)->hh.tbl->num_buckets*sizeof(struct UT_hash_bucket)); \ - HASH_BLOOM_FREE((head)->hh.tbl); \ - uthash_free((head)->hh.tbl, sizeof(UT_hash_table)); \ - (head)=NULL; \ - } \ -} while(0) - -#define HASH_OVERHEAD(hh,head) \ - (size_t)((((head)->hh.tbl->num_items * sizeof(UT_hash_handle)) + \ - ((head)->hh.tbl->num_buckets * sizeof(UT_hash_bucket)) + \ - (sizeof(UT_hash_table)) + \ - (HASH_BLOOM_BYTELEN))) - -#ifdef NO_DECLTYPE -#define HASH_ITER(hh,head,el,tmp) \ -for((el)=(head), (*(char**)(&(tmp)))=(char*)((head)?(head)->hh.next:NULL); \ - el; (el)=(tmp),(*(char**)(&(tmp)))=(char*)((tmp)?(tmp)->hh.next:NULL)) -#else -#define HASH_ITER(hh,head,el,tmp) \ -for((el)=(head),(tmp)=DECLTYPE(el)((head)?(head)->hh.next:NULL); \ - el; (el)=(tmp),(tmp)=DECLTYPE(el)((tmp)?(tmp)->hh.next:NULL)) -#endif - -/* obtain a count of items in the hash */ -#define HASH_COUNT(head) HASH_CNT(hh,head) -#define HASH_CNT(hh,head) ((head)?((head)->hh.tbl->num_items):0) - -typedef struct UT_hash_bucket { - struct UT_hash_handle *hh_head; - unsigned count; - - /* expand_mult is normally set to 0. In this situation, the max chain length - * threshold is enforced at its default value, HASH_BKT_CAPACITY_THRESH. (If - * the bucket's chain exceeds this length, bucket expansion is triggered). - * However, setting expand_mult to a non-zero value delays bucket expansion - * (that would be triggered by additions to this particular bucket) - * until its chain length reaches a *multiple* of HASH_BKT_CAPACITY_THRESH. - * (The multiplier is simply expand_mult+1). The whole idea of this - * multiplier is to reduce bucket expansions, since they are expensive, in - * situations where we know that a particular bucket tends to be overused. - * It is better to let its chain length grow to a longer yet-still-bounded - * value, than to do an O(n) bucket expansion too often. - */ - unsigned expand_mult; - -} UT_hash_bucket; - -/* random signature used only to find hash tables in external analysis */ -#define HASH_SIGNATURE 0xa0111fe1 -#define HASH_BLOOM_SIGNATURE 0xb12220f2 - -typedef struct UT_hash_table { - UT_hash_bucket *buckets; - unsigned num_buckets, log2_num_buckets; - unsigned num_items; - struct UT_hash_handle *tail; /* tail hh in app order, for fast append */ - ptrdiff_t hho; /* hash handle offset (byte pos of hash handle in element */ - - /* in an ideal situation (all buckets used equally), no bucket would have - * more than ceil(#items/#buckets) items. that's the ideal chain length. */ - unsigned ideal_chain_maxlen; - - /* nonideal_items is the number of items in the hash whose chain position - * exceeds the ideal chain maxlen. these items pay the penalty for an uneven - * hash distribution; reaching them in a chain traversal takes >ideal steps */ - unsigned nonideal_items; - - /* ineffective expands occur when a bucket doubling was performed, but - * afterward, more than half the items in the hash had nonideal chain - * positions. If this happens on two consecutive expansions we inhibit any - * further expansion, as it's not helping; this happens when the hash - * function isn't a good fit for the key domain. When expansion is inhibited - * the hash will still work, albeit no longer in constant time. */ - unsigned ineff_expands, noexpand; - - uint32_t signature; /* used only to find hash tables in external analysis */ -#ifdef HASH_BLOOM - uint32_t bloom_sig; /* used only to test bloom exists in external analysis */ - uint8_t *bloom_bv; - char bloom_nbits; -#endif - -} UT_hash_table; - -typedef struct UT_hash_handle { - struct UT_hash_table *tbl; - void *prev; /* prev element in app order */ - void *next; /* next element in app order */ - struct UT_hash_handle *hh_prev; /* previous hh in bucket order */ - struct UT_hash_handle *hh_next; /* next hh in bucket order */ - void *key; /* ptr to enclosing struct's key */ - unsigned keylen; /* enclosing struct's key len */ - unsigned hashv; /* result of hash-fcn(key) */ -} UT_hash_handle; - -#endif /* UTHASH_H */ -- cgit v1.2.3