Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 1 | /* |
| 2 | * name-hash.c |
| 3 | * |
| 4 | * Hashing names in the index state |
| 5 | * |
| 6 | * Copyright (C) 2008 Linus Torvalds |
| 7 | */ |
| 8 | #define NO_THE_INDEX_COMPATIBILITY_MACROS |
| 9 | #include "cache.h" |
| 10 | |
Linus Torvalds | cd2fef5 | 2008-03-21 15:55:19 -0700 | [diff] [blame] | 11 | /* |
| 12 | * This removes bit 5 if bit 6 is set. |
| 13 | * |
| 14 | * That will make US-ASCII characters hash to their upper-case |
| 15 | * equivalent. We could easily do this one whole word at a time, |
| 16 | * but that's for future worries. |
| 17 | */ |
| 18 | static inline unsigned char icase_hash(unsigned char c) |
| 19 | { |
| 20 | return c & ~((c & 0x40) >> 1); |
| 21 | } |
| 22 | |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 23 | static unsigned int hash_name(const char *name, int namelen) |
| 24 | { |
| 25 | unsigned int hash = 0x123; |
| 26 | |
| 27 | do { |
| 28 | unsigned char c = *name++; |
Linus Torvalds | cd2fef5 | 2008-03-21 15:55:19 -0700 | [diff] [blame] | 29 | c = icase_hash(c); |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 30 | hash = hash*101 + c; |
| 31 | } while (--namelen); |
| 32 | return hash; |
| 33 | } |
| 34 | |
Joshua Jensen | 5102c61 | 2010-10-03 09:56:43 +0000 | [diff] [blame] | 35 | static void hash_index_entry_directories(struct index_state *istate, struct cache_entry *ce) |
| 36 | { |
| 37 | /* |
| 38 | * Throw each directory component in the hash for quick lookup |
| 39 | * during a git status. Directory components are stored with their |
| 40 | * closing slash. Despite submodules being a directory, they never |
| 41 | * reach this point, because they are stored without a closing slash |
| 42 | * in the cache. |
| 43 | * |
| 44 | * Note that the cache_entry stored with the directory does not |
| 45 | * represent the directory itself. It is a pointer to an existing |
| 46 | * filename, and its only purpose is to represent existence of the |
| 47 | * directory in the cache. It is very possible multiple directory |
| 48 | * hash entries may point to the same cache_entry. |
| 49 | */ |
| 50 | unsigned int hash; |
| 51 | void **pos; |
| 52 | |
| 53 | const char *ptr = ce->name; |
| 54 | while (*ptr) { |
| 55 | while (*ptr && *ptr != '/') |
| 56 | ++ptr; |
| 57 | if (*ptr == '/') { |
| 58 | ++ptr; |
| 59 | hash = hash_name(ce->name, ptr - ce->name); |
Jeff King | 2548183 | 2011-10-06 12:06:09 -0400 | [diff] [blame] | 60 | pos = insert_hash(hash, ce, &istate->name_hash); |
| 61 | if (pos) { |
| 62 | ce->dir_next = *pos; |
| 63 | *pos = ce; |
Joshua Jensen | 5102c61 | 2010-10-03 09:56:43 +0000 | [diff] [blame] | 64 | } |
| 65 | } |
| 66 | } |
| 67 | } |
| 68 | |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 69 | static void hash_index_entry(struct index_state *istate, struct cache_entry *ce) |
| 70 | { |
| 71 | void **pos; |
| 72 | unsigned int hash; |
| 73 | |
| 74 | if (ce->ce_flags & CE_HASHED) |
| 75 | return; |
| 76 | ce->ce_flags |= CE_HASHED; |
| 77 | ce->next = NULL; |
| 78 | hash = hash_name(ce->name, ce_namelen(ce)); |
| 79 | pos = insert_hash(hash, ce, &istate->name_hash); |
| 80 | if (pos) { |
| 81 | ce->next = *pos; |
| 82 | *pos = ce; |
| 83 | } |
Joshua Jensen | 5102c61 | 2010-10-03 09:56:43 +0000 | [diff] [blame] | 84 | |
| 85 | if (ignore_case) |
| 86 | hash_index_entry_directories(istate, ce); |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 87 | } |
| 88 | |
| 89 | static void lazy_init_name_hash(struct index_state *istate) |
| 90 | { |
| 91 | int nr; |
| 92 | |
| 93 | if (istate->name_hash_initialized) |
| 94 | return; |
| 95 | for (nr = 0; nr < istate->cache_nr; nr++) |
| 96 | hash_index_entry(istate, istate->cache[nr]); |
| 97 | istate->name_hash_initialized = 1; |
| 98 | } |
| 99 | |
| 100 | void add_name_hash(struct index_state *istate, struct cache_entry *ce) |
| 101 | { |
| 102 | ce->ce_flags &= ~CE_UNHASHED; |
| 103 | if (istate->name_hash_initialized) |
| 104 | hash_index_entry(istate, ce); |
| 105 | } |
| 106 | |
Linus Torvalds | cd2fef5 | 2008-03-21 15:55:19 -0700 | [diff] [blame] | 107 | static int slow_same_name(const char *name1, int len1, const char *name2, int len2) |
| 108 | { |
| 109 | if (len1 != len2) |
| 110 | return 0; |
| 111 | |
| 112 | while (len1) { |
| 113 | unsigned char c1 = *name1++; |
| 114 | unsigned char c2 = *name2++; |
| 115 | len1--; |
| 116 | if (c1 != c2) { |
| 117 | c1 = toupper(c1); |
| 118 | c2 = toupper(c2); |
| 119 | if (c1 != c2) |
| 120 | return 0; |
| 121 | } |
| 122 | } |
| 123 | return 1; |
| 124 | } |
| 125 | |
| 126 | static int same_name(const struct cache_entry *ce, const char *name, int namelen, int icase) |
| 127 | { |
| 128 | int len = ce_namelen(ce); |
| 129 | |
| 130 | /* |
| 131 | * Always do exact compare, even if we want a case-ignoring comparison; |
| 132 | * we do the quick exact one first, because it will be the common case. |
| 133 | */ |
| 134 | if (len == namelen && !cache_name_compare(name, namelen, ce->name, len)) |
| 135 | return 1; |
| 136 | |
Joshua Jensen | 5102c61 | 2010-10-03 09:56:43 +0000 | [diff] [blame] | 137 | if (!icase) |
| 138 | return 0; |
| 139 | |
| 140 | /* |
| 141 | * If the entry we're comparing is a filename (no trailing slash), then compare |
| 142 | * the lengths exactly. |
| 143 | */ |
| 144 | if (name[namelen - 1] != '/') |
| 145 | return slow_same_name(name, namelen, ce->name, len); |
| 146 | |
| 147 | /* |
| 148 | * For a directory, we point to an arbitrary cache_entry filename. Just |
| 149 | * make sure the directory portion matches. |
| 150 | */ |
| 151 | return slow_same_name(name, namelen, ce->name, namelen < len ? namelen : len); |
Linus Torvalds | cd2fef5 | 2008-03-21 15:55:19 -0700 | [diff] [blame] | 152 | } |
| 153 | |
| 154 | struct cache_entry *index_name_exists(struct index_state *istate, const char *name, int namelen, int icase) |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 155 | { |
| 156 | unsigned int hash = hash_name(name, namelen); |
| 157 | struct cache_entry *ce; |
| 158 | |
| 159 | lazy_init_name_hash(istate); |
| 160 | ce = lookup_hash(hash, &istate->name_hash); |
| 161 | |
| 162 | while (ce) { |
| 163 | if (!(ce->ce_flags & CE_UNHASHED)) { |
Linus Torvalds | cd2fef5 | 2008-03-21 15:55:19 -0700 | [diff] [blame] | 164 | if (same_name(ce, name, namelen, icase)) |
Linus Torvalds | df292c7 | 2008-03-21 15:53:00 -0700 | [diff] [blame] | 165 | return ce; |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 166 | } |
Jeff King | 2548183 | 2011-10-06 12:06:09 -0400 | [diff] [blame] | 167 | if (icase && name[namelen - 1] == '/') |
| 168 | ce = ce->dir_next; |
| 169 | else |
| 170 | ce = ce->next; |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 171 | } |
Joshua Jensen | 5102c61 | 2010-10-03 09:56:43 +0000 | [diff] [blame] | 172 | |
| 173 | /* |
| 174 | * Might be a submodule. Despite submodules being directories, |
| 175 | * they are stored in the name hash without a closing slash. |
| 176 | * When ignore_case is 1, directories are stored in the name hash |
| 177 | * with their closing slash. |
| 178 | * |
| 179 | * The side effect of this storage technique is we have need to |
| 180 | * remove the slash from name and perform the lookup again without |
| 181 | * the slash. If a match is made, S_ISGITLINK(ce->mode) will be |
| 182 | * true. |
| 183 | */ |
| 184 | if (icase && name[namelen - 1] == '/') { |
| 185 | ce = index_name_exists(istate, name, namelen - 1, icase); |
| 186 | if (ce && S_ISGITLINK(ce->ce_mode)) |
| 187 | return ce; |
| 188 | } |
Linus Torvalds | df292c7 | 2008-03-21 15:53:00 -0700 | [diff] [blame] | 189 | return NULL; |
Linus Torvalds | 96872bc | 2008-03-21 13:16:24 -0700 | [diff] [blame] | 190 | } |