X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/381920793346413bc34dc87a31bc4ddf501fcdf8..8632fb30401ccc6824b84499602fba7052cc3346:/zipmap.c?ds=sidebyside diff --git a/zipmap.c b/zipmap.c index 7d28c202..35faeabe 100644 --- a/zipmap.c +++ b/zipmap.c @@ -42,10 +42,11 @@ /* Memory layout of a zipmap, for the map "foo" => "bar", "hello" => "world": * - * "foo""bar""hello""world" + * "foo""bar""hello""world" * - * is 1 byte status. Currently only 1 bit is used: if the least - * significant bit is set, it means the zipmap needs to be defragmented. + * is 1 byte length that holds the current size of the zipmap. + * When the zipmap length is greater than or equal to 254, this value + * is not used and the zipmap needs to be traversed to find out the length. * * is the length of the following string (key or value). * lengths are encoded in a single value or in a 5 bytes value. @@ -62,22 +63,15 @@ * or even in order to add a key/value pair if it fits. * * is always an unsigned 8 bit number, because if after an - * update operation there are more than a few free bytes, they'll be converted - * into empty space prefixed by the special value 254. + * update operation there are more than a few free bytes, the zipmap will be + * reallocated to make sure it is as small as possible. * * The most compact representation of the above two elements hash is actually: * - * "\x00\x03foo\x03\x00bar\x05hello\x05\x00world\xff" + * "\x02\x03foo\x03\x00bar\x05hello\x05\x00world\xff" * - * Empty space is marked using a 254 bytes + a (coded as already - * specified). The length includes the 254 bytes in the count and the - * space taken by the field. So for instance removing the "foo" key - * from the zipmap above will lead to the following representation: - * - * "\x00\xfd\x10........\x05hello\x05\x00world\xff" - * - * Note that because empty space, keys, values, are all prefixed length - * "objects", the lookup will take O(N) where N is the numeber of elements + * Note that because keys and values are prefixed length "objects", + * the lookup will take O(N) where N is the number of elements * in the zipmap and *not* the number of bytes needed to represent the zipmap. * This lowers the constant times considerably. */ @@ -92,7 +86,7 @@ /* The following defines the max value for the field described in the * comments above, that is, the max number of trailing bytes in a value. */ -#define ZIPMAP_VALUE_MAX_FREE 5 +#define ZIPMAP_VALUE_MAX_FREE 4 /* The following macro returns the number of bytes needed to encode the length * for the integer value _l, that is, 1 byte for lengths < ZIPMAP_BIGLEN and @@ -142,14 +136,15 @@ static unsigned int zipmapEncodeLength(unsigned char *p, unsigned int len) { * reallocate the original zipmap to make room for more entries. */ static unsigned char *zipmapLookupRaw(unsigned char *zm, unsigned char *key, unsigned int klen, unsigned int *totlen) { unsigned char *p = zm+1, *k = NULL; - unsigned int l; + unsigned int l,llen; while(*p != ZIPMAP_END) { unsigned char free; /* Match or skip the key */ l = zipmapDecodeLength(p); - if (k == NULL && l == klen && !memcmp(p+1,key,l)) { + llen = zipmapEncodeLength(NULL,l); + if (k == NULL && l == klen && !memcmp(p+llen,key,l)) { /* Only return when the user doesn't care * for the total length of the zipmap. */ if (totlen != NULL) { @@ -158,7 +153,7 @@ static unsigned char *zipmapLookupRaw(unsigned char *zm, unsigned char *key, uns return p; } } - p += zipmapEncodeLength(NULL,l) + l; + p += llen+l; /* Skip the value as well */ l = zipmapDecodeLength(p); p += zipmapEncodeLength(NULL,l); @@ -181,7 +176,6 @@ static unsigned long zipmapRequiredLength(unsigned int klen, unsigned int vlen) /* Return the total amount used by a key (encoded length + payload) */ static unsigned int zipmapRawKeyLength(unsigned char *p) { unsigned int l = zipmapDecodeLength(p); - return zipmapEncodeLength(NULL,l) + l; } @@ -201,7 +195,6 @@ static unsigned int zipmapRawValueLength(unsigned char *p) { * free space if any). */ static unsigned int zipmapRawEntryLength(unsigned char *p) { unsigned int l = zipmapRawKeyLength(p); - return l + zipmapRawValueLength(p+l); } @@ -215,7 +208,7 @@ static inline unsigned char *zipmapResize(unsigned char *zm, unsigned int len) { * If 'update' is not NULL, *update is set to 1 if the key was * already preset, otherwise to 0. */ unsigned char *zipmapSet(unsigned char *zm, unsigned char *key, unsigned int klen, unsigned char *val, unsigned int vlen, int *update) { - unsigned int zmlen; + unsigned int zmlen, offset; unsigned int freelen, reqlen = zipmapRequiredLength(klen,vlen); unsigned int empty, vempty; unsigned char *p; @@ -232,36 +225,39 @@ unsigned char *zipmapSet(unsigned char *zm, unsigned char *key, unsigned int kle /* Increase zipmap length (this is an insert) */ if (zm[0] < ZIPMAP_BIGLEN) zm[0]++; } else { - unsigned char *b = p; - /* Key found. Is there enough space for the new value? */ /* Compute the total length: */ if (update) *update = 1; - freelen = zipmapRawKeyLength(b); - b += freelen; - freelen += zipmapRawValueLength(b); + freelen = zipmapRawEntryLength(p); if (freelen < reqlen) { - /* Move remaining entries to the current position, so this - * pair can be appended. Note: the +1 in memmove is caused - * by the end-of-zipmap byte. */ - memmove(p, p+freelen, zmlen-((p-zm)+freelen+1)); + /* Store the offset of this key within the current zipmap, so + * it can be resized. Then, move the tail backwards so this + * pair fits at the current position. */ + offset = p-zm; zm = zipmapResize(zm, zmlen-freelen+reqlen); - p = zm+zmlen-1-freelen; - zmlen = zmlen-1-freelen+reqlen; + p = zm+offset; + + /* The +1 in the number of bytes to be moved is caused by the + * end-of-zipmap byte. Note: the *original* zmlen is used. */ + memmove(p+reqlen, p+freelen, zmlen-(offset+freelen+1)); + zmlen = zmlen-freelen+reqlen; freelen = reqlen; } } - /* Ok we have a suitable block where to write the new key/value - * entry. */ + /* We now have a suitable block where the key/value entry can + * be written. If there is too much free space, move the tail + * of the zipmap a few bytes to the front and shrink the zipmap, + * as we want zipmaps to be very space efficient. */ empty = freelen-reqlen; - /* If there is too much free space mark it as a free block instead - * of adding it as trailing empty space for the value, as we want - * zipmaps to be very space efficient. */ if (empty >= ZIPMAP_VALUE_MAX_FREE) { - memmove(p+reqlen, p+freelen, zmlen-((p-zm)+freelen+1)); + /* First, move the tail bytes to the front, then resize + * the zipmap to be bytes smaller. */ + offset = p-zm; + memmove(p+reqlen, p+freelen, zmlen-(offset+freelen+1)); zmlen -= empty; zm = zipmapResize(zm, zmlen); + p = zm+offset; vempty = 0; } else { vempty = empty; @@ -407,7 +403,6 @@ int main(void) { zm = zipmapSet(zm,(unsigned char*) "surname",7, (unsigned char*) "foo",3,NULL); zm = zipmapSet(zm,(unsigned char*) "age",3, (unsigned char*) "foo",3,NULL); zipmapRepr(zm); - exit(1); zm = zipmapSet(zm,(unsigned char*) "hello",5, (unsigned char*) "world!",6,NULL); zm = zipmapSet(zm,(unsigned char*) "foo",3, (unsigned char*) "bar",3,NULL); @@ -420,6 +415,21 @@ int main(void) { zipmapRepr(zm); zm = zipmapDel(zm,(unsigned char*) "new",3,NULL); zipmapRepr(zm); + + printf("\nLook up large key:\n"); + { + unsigned char buf[512]; + unsigned char *value; + unsigned int vlen, i; + for (i = 0; i < 512; i++) buf[i] = 'a'; + + zm = zipmapSet(zm,buf,512,(unsigned char*) "long",4,NULL); + if (zipmapGet(zm,buf,512,&value,&vlen)) { + printf(" is associated to the %d bytes value: %.*s\n", + vlen, vlen, value); + } + } + printf("\nPerform a direct lookup:\n"); { unsigned char *value;