]>
git.saurik.com Git - redis.git/blob - src/multi.c
3 /* ================================ MULTI/EXEC ============================== */
5 /* Client state initialization for MULTI/EXEC */
6 void initClientMultiState(redisClient
*c
) {
7 c
->mstate
.commands
= NULL
;
11 /* Release all the resources associated with MULTI/EXEC state */
12 void freeClientMultiState(redisClient
*c
) {
15 for (j
= 0; j
< c
->mstate
.count
; j
++) {
17 multiCmd
*mc
= c
->mstate
.commands
+j
;
19 for (i
= 0; i
< mc
->argc
; i
++)
20 decrRefCount(mc
->argv
[i
]);
23 zfree(c
->mstate
.commands
);
26 /* Add a new command into the MULTI commands queue */
27 void queueMultiCommand(redisClient
*c
) {
31 c
->mstate
.commands
= zrealloc(c
->mstate
.commands
,
32 sizeof(multiCmd
)*(c
->mstate
.count
+1));
33 mc
= c
->mstate
.commands
+c
->mstate
.count
;
36 mc
->argv
= zmalloc(sizeof(robj
*)*c
->argc
);
37 memcpy(mc
->argv
,c
->argv
,sizeof(robj
*)*c
->argc
);
38 for (j
= 0; j
< c
->argc
; j
++)
39 incrRefCount(mc
->argv
[j
]);
43 void discardTransaction(redisClient
*c
) {
44 freeClientMultiState(c
);
45 initClientMultiState(c
);
46 c
->flags
&= ~(REDIS_MULTI
|REDIS_DIRTY_CAS
);;
50 void multiCommand(redisClient
*c
) {
51 if (c
->flags
& REDIS_MULTI
) {
52 addReplyError(c
,"MULTI calls can not be nested");
55 c
->flags
|= REDIS_MULTI
;
56 addReply(c
,shared
.ok
);
59 void discardCommand(redisClient
*c
) {
60 if (!(c
->flags
& REDIS_MULTI
)) {
61 addReplyError(c
,"DISCARD without MULTI");
64 discardTransaction(c
);
65 addReply(c
,shared
.ok
);
68 /* Send a MULTI command to all the slaves and AOF file. Check the execCommand
69 * implememntation for more information. */
70 void execCommandReplicateMulti(redisClient
*c
) {
71 robj
*multistring
= createStringObject("MULTI",5);
73 if (server
.aof_state
!= REDIS_AOF_OFF
)
74 feedAppendOnlyFile(server
.multiCommand
,c
->db
->id
,&multistring
,1);
75 if (listLength(server
.slaves
))
76 replicationFeedSlaves(server
.slaves
,c
->db
->id
,&multistring
,1);
77 decrRefCount(multistring
);
80 void execCommand(redisClient
*c
) {
84 struct redisCommand
*orig_cmd
;
86 if (!(c
->flags
& REDIS_MULTI
)) {
87 addReplyError(c
,"EXEC without MULTI");
91 /* Check if we need to abort the EXEC if some WATCHed key was touched.
92 * A failed EXEC will return a multi bulk nil object. */
93 if (c
->flags
& REDIS_DIRTY_CAS
) {
94 freeClientMultiState(c
);
95 initClientMultiState(c
);
96 c
->flags
&= ~(REDIS_MULTI
|REDIS_DIRTY_CAS
);
98 addReply(c
,shared
.nullmultibulk
);
102 /* Replicate a MULTI request now that we are sure the block is executed.
103 * This way we'll deliver the MULTI/..../EXEC block as a whole and
104 * both the AOF and the replication link will have the same consistency
105 * and atomicity guarantees. */
106 execCommandReplicateMulti(c
);
108 /* Exec all the queued commands */
109 unwatchAllKeys(c
); /* Unwatch ASAP otherwise we'll waste CPU cycles */
113 addReplyMultiBulkLen(c
,c
->mstate
.count
);
114 for (j
= 0; j
< c
->mstate
.count
; j
++) {
115 c
->argc
= c
->mstate
.commands
[j
].argc
;
116 c
->argv
= c
->mstate
.commands
[j
].argv
;
117 c
->cmd
= c
->mstate
.commands
[j
].cmd
;
118 call(c
,REDIS_CALL_FULL
);
120 /* Commands may alter argc/argv, restore mstate. */
121 c
->mstate
.commands
[j
].argc
= c
->argc
;
122 c
->mstate
.commands
[j
].argv
= c
->argv
;
123 c
->mstate
.commands
[j
].cmd
= c
->cmd
;
128 freeClientMultiState(c
);
129 initClientMultiState(c
);
130 c
->flags
&= ~(REDIS_MULTI
|REDIS_DIRTY_CAS
);
131 /* Make sure the EXEC command is always replicated / AOF, since we
132 * always send the MULTI command (we can't know beforehand if the
133 * next operations will contain at least a modification to the DB). */
137 /* ===================== WATCH (CAS alike for MULTI/EXEC) ===================
139 * The implementation uses a per-DB hash table mapping keys to list of clients
140 * WATCHing those keys, so that given a key that is going to be modified
141 * we can mark all the associated clients as dirty.
143 * Also every client contains a list of WATCHed keys so that's possible to
144 * un-watch such keys when the client is freed or when UNWATCH is called. */
146 /* In the client->watched_keys list we need to use watchedKey structures
147 * as in order to identify a key in Redis we need both the key name and the
149 typedef struct watchedKey
{
154 /* Watch for the specified key */
155 void watchForKey(redisClient
*c
, robj
*key
) {
156 list
*clients
= NULL
;
161 /* Check if we are already watching for this key */
162 listRewind(c
->watched_keys
,&li
);
163 while((ln
= listNext(&li
))) {
164 wk
= listNodeValue(ln
);
165 if (wk
->db
== c
->db
&& equalStringObjects(key
,wk
->key
))
166 return; /* Key already watched */
168 /* This key is not already watched in this DB. Let's add it */
169 clients
= dictFetchValue(c
->db
->watched_keys
,key
);
171 clients
= listCreate();
172 dictAdd(c
->db
->watched_keys
,key
,clients
);
175 listAddNodeTail(clients
,c
);
176 /* Add the new key to the lits of keys watched by this client */
177 wk
= zmalloc(sizeof(*wk
));
181 listAddNodeTail(c
->watched_keys
,wk
);
184 /* Unwatch all the keys watched by this client. To clean the EXEC dirty
185 * flag is up to the caller. */
186 void unwatchAllKeys(redisClient
*c
) {
190 if (listLength(c
->watched_keys
) == 0) return;
191 listRewind(c
->watched_keys
,&li
);
192 while((ln
= listNext(&li
))) {
196 /* Lookup the watched key -> clients list and remove the client
198 wk
= listNodeValue(ln
);
199 clients
= dictFetchValue(wk
->db
->watched_keys
, wk
->key
);
200 redisAssertWithInfo(c
,NULL
,clients
!= NULL
);
201 listDelNode(clients
,listSearchKey(clients
,c
));
202 /* Kill the entry at all if this was the only client */
203 if (listLength(clients
) == 0)
204 dictDelete(wk
->db
->watched_keys
, wk
->key
);
205 /* Remove this watched key from the client->watched list */
206 listDelNode(c
->watched_keys
,ln
);
207 decrRefCount(wk
->key
);
212 /* "Touch" a key, so that if this key is being WATCHed by some client the
213 * next EXEC will fail. */
214 void touchWatchedKey(redisDb
*db
, robj
*key
) {
219 if (dictSize(db
->watched_keys
) == 0) return;
220 clients
= dictFetchValue(db
->watched_keys
, key
);
221 if (!clients
) return;
223 /* Mark all the clients watching this key as REDIS_DIRTY_CAS */
224 /* Check if we are already watching for this key */
225 listRewind(clients
,&li
);
226 while((ln
= listNext(&li
))) {
227 redisClient
*c
= listNodeValue(ln
);
229 c
->flags
|= REDIS_DIRTY_CAS
;
233 /* On FLUSHDB or FLUSHALL all the watched keys that are present before the
234 * flush but will be deleted as effect of the flushing operation should
235 * be touched. "dbid" is the DB that's getting the flush. -1 if it is
236 * a FLUSHALL operation (all the DBs flushed). */
237 void touchWatchedKeysOnFlush(int dbid
) {
241 /* For every client, check all the waited keys */
242 listRewind(server
.clients
,&li1
);
243 while((ln
= listNext(&li1
))) {
244 redisClient
*c
= listNodeValue(ln
);
245 listRewind(c
->watched_keys
,&li2
);
246 while((ln
= listNext(&li2
))) {
247 watchedKey
*wk
= listNodeValue(ln
);
249 /* For every watched key matching the specified DB, if the
250 * key exists, mark the client as dirty, as the key will be
252 if (dbid
== -1 || wk
->db
->id
== dbid
) {
253 if (dictFind(wk
->db
->dict
, wk
->key
->ptr
) != NULL
)
254 c
->flags
|= REDIS_DIRTY_CAS
;
260 void watchCommand(redisClient
*c
) {
263 if (c
->flags
& REDIS_MULTI
) {
264 addReplyError(c
,"WATCH inside MULTI is not allowed");
267 for (j
= 1; j
< c
->argc
; j
++)
268 watchForKey(c
,c
->argv
[j
]);
269 addReply(c
,shared
.ok
);
272 void unwatchCommand(redisClient
*c
) {
274 c
->flags
&= (~REDIS_DIRTY_CAS
);
275 addReply(c
,shared
.ok
);