| 1 |
/*------------------------------------------------------------------------- |
| 2 |
* |
| 3 |
* hashsort.c |
| 4 |
* Sort tuples for insertion into a new hash index. |
| 5 |
* |
| 6 |
* When building a very large hash index, we pre-sort the tuples by bucket |
| 7 |
* number to improve locality of access to the index, and thereby avoid |
| 8 |
* thrashing. We use tuplesort.c to sort the given index tuples into order. |
| 9 |
* |
| 10 |
* Note: if the number of rows in the table has been underestimated, |
| 11 |
* bucket splits may occur during the index build. In that case we'd |
| 12 |
* be inserting into two or more buckets for each possible masked-off |
| 13 |
* hash code value. That's no big problem though, since we'll still have |
| 14 |
* plenty of locality of access. |
| 15 |
* |
| 16 |
* |
| 17 |
* Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group |
| 18 |
* Portions Copyright (c) 1994, Regents of the University of California |
| 19 |
* |
| 20 |
* IDENTIFICATION |
| 21 |
* src/backend/access/hash/hashsort.c |
| 22 |
* |
| 23 |
*------------------------------------------------------------------------- |
| 24 |
*/ |
| 25 |
|
| 26 |
#include "postgres.h" |
| 27 |
|
| 28 |
#include "access/hash.h" |
| 29 |
#include "commands/progress.h" |
| 30 |
#include "miscadmin.h" |
| 31 |
#include "pgstat.h" |
| 32 |
#include "port/pg_bitutils.h" |
| 33 |
#include "utils/tuplesort.h" |
| 34 |
|
| 35 |
|
| 36 |
/* |
| 37 |
* Status record for spooling/sorting phase. |
| 38 |
*/ |
| 39 |
struct HSpool |
| 40 |
{ |
| 41 |
Tuplesortstate *sortstate; /* state data for tuplesort.c */ |
| 42 |
Relation index; |
| 43 |
|
| 44 |
/* |
| 45 |
* We sort the hash keys based on the buckets they belong to, then by the |
| 46 |
* hash values themselves, to optimize insertions onto hash pages. The |
| 47 |
* masks below are used in _hash_hashkey2bucket to determine the bucket of |
| 48 |
* a given hash key. |
| 49 |
*/ |
| 50 |
uint32 high_mask; |
| 51 |
uint32 low_mask; |
| 52 |
uint32 max_buckets; |
| 53 |
}; |
| 54 |
|
| 55 |
|
| 56 |
/* |
| 57 |
* create and initialize a spool structure |
| 58 |
*/ |
| 59 |
HSpool * |
| 60 |
_h_spoolinit(Relation heap, Relation index, uint32 num_buckets) |
| 61 |
{ |
| 62 |
HSpool *hspool = (HSpool *) palloc0(sizeof(HSpool)); |
| 63 |
|
| 64 |
hspool->index = index; |
| 65 |
|
| 66 |
/* |
| 67 |
* Determine the bitmask for hash code values. Since there are currently |
| 68 |
* num_buckets buckets in the index, the appropriate mask can be computed |
| 69 |
* as follows. |
| 70 |
* |
| 71 |
* NOTE : This hash mask calculation should be in sync with similar |
| 72 |
* calculation in _hash_init_metabuffer. |
| 73 |
*/ |
| 74 |
hspool->high_mask = pg_nextpower2_32(num_buckets + 1) - 1; |
| 75 |
hspool->low_mask = (hspool->high_mask >> 1); |
| 76 |
hspool->max_buckets = num_buckets - 1; |
| 77 |
|
| 78 |
/* |
| 79 |
* We size the sort area as maintenance_work_mem rather than work_mem to |
| 80 |
* speed index creation. This should be OK since a single backend can't |
| 81 |
* run multiple index creations in parallel. |
| 82 |
*/ |
| 83 |
hspool->sortstate = tuplesort_begin_index_hash(heap, |
| 84 |
index, |
| 85 |
hspool->high_mask, |
| 86 |
hspool->low_mask, |
| 87 |
hspool->max_buckets, |
| 88 |
maintenance_work_mem, |
| 89 |
NULL, |
| 90 |
TUPLESORT_NONE); |
| 91 |
|
| 92 |
return hspool; |
| 93 |
} |
| 94 |
|
| 95 |
/* |
| 96 |
* clean up a spool structure and its substructures. |
| 97 |
*/ |
| 98 |
void |
| 99 |
_h_spooldestroy(HSpool *hspool) |
| 100 |
{ |
| 101 |
tuplesort_end(hspool->sortstate); |
| 102 |
pfree(hspool); |
| 103 |
} |
| 104 |
|
| 105 |
/* |
| 106 |
* spool an index entry into the sort file. |
| 107 |
*/ |
| 108 |
void |
| 109 |
_h_spool(HSpool *hspool, ItemPointer self, const Datum *values, const bool *isnull) |
| 110 |
{ |
| 111 |
tuplesort_putindextuplevalues(hspool->sortstate, hspool->index, |
| 112 |
self, values, isnull); |
| 113 |
} |
| 114 |
|
| 115 |
/* |
| 116 |
* given a spool loaded by successive calls to _h_spool, |
| 117 |
* create an entire index. |
| 118 |
*/ |
| 119 |
void |
| 120 |
_h_indexbuild(HSpool *hspool, Relation heapRel) |
| 121 |
{ |
| 122 |
IndexTuple itup; |
| 123 |
int64 tups_done = 0; |
| 124 |
#ifdef USE_ASSERT_CHECKING |
| 125 |
uint32 hashkey = 0; |
| 126 |
#endif |
| 127 |
|
| 128 |
tuplesort_performsort(hspool->sortstate); |
| 129 |
|
| 130 |
while ((itup = tuplesort_getindextuple(hspool->sortstate, true)) != NULL) |
| 131 |
{ |
| 132 |
/* |
| 133 |
* Technically, it isn't critical that hash keys be found in sorted |
| 134 |
* order, since this sorting is only used to increase locality of |
| 135 |
* access as a performance optimization. It still seems like a good |
| 136 |
* idea to test tuplesort.c's handling of hash index tuple sorts |
| 137 |
* through an assertion, though. |
| 138 |
*/ |
| 139 |
#ifdef USE_ASSERT_CHECKING |
| 140 |
uint32 lasthashkey = hashkey; |
| 141 |
|
| 142 |
hashkey = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup), |
| 143 |
hspool->max_buckets, hspool->high_mask, |
| 144 |
hspool->low_mask); |
| 145 |
Assert(hashkey >= lasthashkey); |
| 146 |
#endif |
| 147 |
|
| 148 |
/* the tuples are sorted by hashkey, so pass 'sorted' as true */ |
| 149 |
_hash_doinsert(hspool->index, itup, heapRel, true); |
| 150 |
|
| 151 |
/* allow insertion phase to be interrupted, and track progress */ |
| 152 |
CHECK_FOR_INTERRUPTS(); |
| 153 |
|
| 154 |
pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE, |
| 155 |
++tups_done); |
| 156 |
} |
| 157 |
} |