Age Owner TLA Line data Source code
1 : /*
2 : * brin_bloom.c
3 : * Implementation of Bloom opclass for BRIN
4 : *
5 : * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
6 : * Portions Copyright (c) 1994, Regents of the University of California
7 : *
8 : *
9 : * A BRIN opclass summarizing page range into a bloom filter.
10 : *
11 : * Bloom filters allow efficient testing whether a given page range contains
12 : * a particular value. Therefore, if we summarize each page range into a small
13 : * bloom filter, we can easily (and cheaply) test whether it contains values
14 : * we get later.
15 : *
16 : * The index only supports equality operators, similarly to hash indexes.
17 : * Bloom indexes are however much smaller, and support only bitmap scans.
18 : *
19 : * Note: Don't confuse this with bloom indexes, implemented in a contrib
20 : * module. That extension implements an entirely new AM, building a bloom
21 : * filter on multiple columns in a single row. This opclass works with an
22 : * existing AM (BRIN) and builds bloom filter on a column.
23 : *
24 : *
25 : * values vs. hashes
26 : * -----------------
27 : *
28 : * The original column values are not used directly, but are first hashed
29 : * using the regular type-specific hash function, producing a uint32 hash.
30 : * And this hash value is then added to the summary - i.e. it's hashed
31 : * again and added to the bloom filter.
32 : *
33 : * This allows the code to treat all data types (byval/byref/...) the same
34 : * way, with only minimal space requirements, because we're working with
35 : * hashes and not the original values. Everything is uint32.
36 : *
37 : * Of course, this assumes the built-in hash function is reasonably good,
38 : * without too many collisions etc. But that does seem to be the case, at
39 : * least based on past experience. After all, the same hash functions are
40 : * used for hash indexes, hash partitioning and so on.
41 : *
42 : *
43 : * hashing scheme
44 : * --------------
45 : *
46 : * Bloom filters require a number of independent hash functions. There are
47 : * different schemes how to construct them - for example we might use
48 : * hash_uint32_extended with random seeds, but that seems fairly expensive.
49 : * We use a scheme requiring only two functions described in this paper:
50 : *
51 : * Less Hashing, Same Performance:Building a Better Bloom Filter
52 : * Adam Kirsch, Michael Mitzenmacher, Harvard School of Engineering and
53 : * Applied Sciences, Cambridge, Massachusetts [DOI 10.1002/rsa.20208]
54 : *
55 : * The two hash functions h1 and h2 are calculated using hard-coded seeds,
56 : * and then combined using (h1 + i * h2) to generate the hash functions.
57 : *
58 : *
59 : * sizing the bloom filter
60 : * -----------------------
61 : *
62 : * Size of a bloom filter depends on the number of distinct values we will
63 : * store in it, and the desired false positive rate. The higher the number
64 : * of distinct values and/or the lower the false positive rate, the larger
65 : * the bloom filter. On the other hand, we want to keep the index as small
66 : * as possible - that's one of the basic advantages of BRIN indexes.
67 : *
68 : * Although the number of distinct elements (in a page range) depends on
69 : * the data, we can consider it fixed. This simplifies the trade-off to
70 : * just false positive rate vs. size.
71 : *
72 : * At the page range level, false positive rate is a probability the bloom
73 : * filter matches a random value. For the whole index (with sufficiently
74 : * many page ranges) it represents the fraction of the index ranges (and
75 : * thus fraction of the table to be scanned) matching the random value.
76 : *
77 : * Furthermore, the size of the bloom filter is subject to implementation
78 : * limits - it has to fit onto a single index page (8kB by default). As
79 : * the bitmap is inherently random (when "full" about half the bits is set
80 : * to 1, randomly), compression can't help very much.
81 : *
82 : * To reduce the size of a filter (to fit to a page), we have to either
83 : * accept higher false positive rate (undesirable), or reduce the number
84 : * of distinct items to be stored in the filter. We can't alter the input
85 : * data, of course, but we may make the BRIN page ranges smaller - instead
86 : * of the default 128 pages (1MB) we may build index with 16-page ranges,
87 : * or something like that. This should reduce the number of distinct values
88 : * in the page range, making the filter smaller (with fixed false positive
89 : * rate). Even for random data sets this should help, as the number of rows
90 : * per heap page is limited (to ~290 with very narrow tables, likely ~20
91 : * in practice).
92 : *
93 : * Of course, good sizing decisions depend on having the necessary data,
94 : * i.e. number of distinct values in a page range (of a given size) and
95 : * table size (to estimate cost change due to change in false positive
96 : * rate due to having larger index vs. scanning larger indexes). We may
97 : * not have that data - for example when building an index on empty table
98 : * it's not really possible. And for some data we only have estimates for
99 : * the whole table and we can only estimate per-range values (ndistinct).
100 : *
101 : * Another challenge is that while the bloom filter is per-column, it's
102 : * the whole index tuple that has to fit into a page. And for multi-column
103 : * indexes that may include pieces we have no control over (not necessarily
104 : * bloom filters, the other columns may use other BRIN opclasses). So it's
105 : * not entirely clear how to distribute the space between those columns.
106 : *
107 : * The current logic, implemented in brin_bloom_get_ndistinct, attempts to
108 : * make some basic sizing decisions, based on the size of BRIN ranges, and
109 : * the maximum number of rows per range.
110 : *
111 : *
112 : * IDENTIFICATION
113 : * src/backend/access/brin/brin_bloom.c
114 : */
115 : #include "postgres.h"
116 :
117 : #include "access/genam.h"
118 : #include "access/brin.h"
119 : #include "access/brin_internal.h"
120 : #include "access/brin_page.h"
121 : #include "access/brin_tuple.h"
122 : #include "access/hash.h"
123 : #include "access/htup_details.h"
124 : #include "access/reloptions.h"
125 : #include "access/stratnum.h"
126 : #include "catalog/pg_type.h"
127 : #include "catalog/pg_amop.h"
128 : #include "utils/builtins.h"
129 : #include "utils/datum.h"
130 : #include "utils/lsyscache.h"
131 : #include "utils/rel.h"
132 : #include "utils/syscache.h"
133 :
134 : #include <math.h>
135 :
136 : #define BloomEqualStrategyNumber 1
137 :
138 : /*
139 : * Additional SQL level support functions. We only have one, which is
140 : * used to calculate hash of the input value.
141 : *
142 : * Procedure numbers must not use values reserved for BRIN itself; see
143 : * brin_internal.h.
144 : */
145 : #define BLOOM_MAX_PROCNUMS 1 /* maximum support procs we need */
146 : #define PROCNUM_HASH 11 /* required */
147 :
148 : /*
149 : * Subtract this from procnum to obtain index in BloomOpaque arrays
150 : * (Must be equal to minimum of private procnums).
151 : */
152 : #define PROCNUM_BASE 11
153 :
154 : /*
155 : * Storage type for BRIN's reloptions.
156 : */
157 : typedef struct BloomOptions
158 : {
159 : int32 vl_len_; /* varlena header (do not touch directly!) */
160 : double nDistinctPerRange; /* number of distinct values per range */
161 : double falsePositiveRate; /* false positive for bloom filter */
162 : } BloomOptions;
163 :
164 : /*
165 : * The current min value (16) is somewhat arbitrary, but it's based
166 : * on the fact that the filter header is ~20B alone, which is about
167 : * the same as the filter bitmap for 16 distinct items with 1% false
168 : * positive rate. So by allowing lower values we'd not gain much. In
169 : * any case, the min should not be larger than MaxHeapTuplesPerPage
170 : * (~290), which is the theoretical maximum for single-page ranges.
171 : */
172 : #define BLOOM_MIN_NDISTINCT_PER_RANGE 16
173 :
174 : /*
175 : * Used to determine number of distinct items, based on the number of rows
176 : * in a page range. The 10% is somewhat similar to what estimate_num_groups
177 : * does, so we use the same factor here.
178 : */
179 : #define BLOOM_DEFAULT_NDISTINCT_PER_RANGE -0.1 /* 10% of values */
180 :
181 : /*
182 : * Allowed range and default value for the false positive range. The exact
183 : * values are somewhat arbitrary, but were chosen considering the various
184 : * parameters (size of filter vs. page size, etc.).
185 : *
186 : * The lower the false-positive rate, the more accurate the filter is, but
187 : * it also gets larger - at some point this eliminates the main advantage
188 : * of BRIN indexes, which is the tiny size. At 0.01% the index is about
189 : * 10% of the table (assuming 290 distinct values per 8kB page).
190 : *
191 : * On the other hand, as the false-positive rate increases, larger part of
192 : * the table has to be scanned due to mismatches - at 25% we're probably
193 : * close to sequential scan being cheaper.
194 : */
195 : #define BLOOM_MIN_FALSE_POSITIVE_RATE 0.0001 /* 0.01% fp rate */
196 : #define BLOOM_MAX_FALSE_POSITIVE_RATE 0.25 /* 25% fp rate */
197 : #define BLOOM_DEFAULT_FALSE_POSITIVE_RATE 0.01 /* 1% fp rate */
198 :
199 : #define BloomGetNDistinctPerRange(opts) \
200 : ((opts) && (((BloomOptions *) (opts))->nDistinctPerRange != 0) ? \
201 : (((BloomOptions *) (opts))->nDistinctPerRange) : \
202 : BLOOM_DEFAULT_NDISTINCT_PER_RANGE)
203 :
204 : #define BloomGetFalsePositiveRate(opts) \
205 : ((opts) && (((BloomOptions *) (opts))->falsePositiveRate != 0.0) ? \
206 : (((BloomOptions *) (opts))->falsePositiveRate) : \
207 : BLOOM_DEFAULT_FALSE_POSITIVE_RATE)
208 :
209 : /*
210 : * And estimate of the largest bloom we can fit onto a page. This is not
211 : * a perfect guarantee, for a couple of reasons. For example, the row may
212 : * be larger because the index has multiple columns.
213 : */
214 : #define BloomMaxFilterSize \
215 : MAXALIGN_DOWN(BLCKSZ - \
216 : (MAXALIGN(SizeOfPageHeaderData + \
217 : sizeof(ItemIdData)) + \
218 : MAXALIGN(sizeof(BrinSpecialSpace)) + \
219 : SizeOfBrinTuple))
220 :
221 : /*
222 : * Seeds used to calculate two hash functions h1 and h2, which are then used
223 : * to generate k hashes using the (h1 + i * h2) scheme.
224 : */
225 : #define BLOOM_SEED_1 0x71d924af
226 : #define BLOOM_SEED_2 0xba48b314
227 :
228 : /*
229 : * Bloom Filter
230 : *
231 : * Represents a bloom filter, built on hashes of the indexed values. That is,
232 : * we compute a uint32 hash of the value, and then store this hash into the
233 : * bloom filter (and compute additional hashes on it).
234 : *
235 : * XXX We could implement "sparse" bloom filters, keeping only the bytes that
236 : * are not entirely 0. But while indexes don't support TOAST, the varlena can
237 : * still be compressed. So this seems unnecessary, because the compression
238 : * should do the same job.
239 : *
240 : * XXX We can also watch the number of bits set in the bloom filter, and then
241 : * stop using it (and not store the bitmap, to save space) when the false
242 : * positive rate gets too high. But even if the false positive rate exceeds the
243 : * desired value, it still can eliminate some page ranges.
244 : */
245 : typedef struct BloomFilter
246 : {
247 : /* varlena header (do not touch directly!) */
248 : int32 vl_len_;
249 :
250 : /* space for various flags (unused for now) */
251 : uint16 flags;
252 :
253 : /* fields for the HASHED phase */
254 : uint8 nhashes; /* number of hash functions */
255 : uint32 nbits; /* number of bits in the bitmap (size) */
256 : uint32 nbits_set; /* number of bits set to 1 */
257 :
258 : /* data of the bloom filter */
259 : char data[FLEXIBLE_ARRAY_MEMBER];
260 : } BloomFilter;
261 :
262 :
263 : /*
264 : * bloom_init
265 : * Initialize the Bloom Filter, allocate all the memory.
266 : *
267 : * The filter is initialized with optimal size for ndistinct expected values
268 : * and the requested false positive rate. The filter is stored as varlena.
269 : */
270 : static BloomFilter *
744 tomas.vondra 271 CBC 3882 : bloom_init(int ndistinct, double false_positive_rate)
272 : {
273 : Size len;
274 : BloomFilter *filter;
275 :
276 : int nbits; /* size of filter / number of bits */
277 : int nbytes; /* size of filter / number of bytes */
278 :
279 : double k; /* number of hash functions */
280 :
281 3882 : Assert(ndistinct > 0);
282 3882 : Assert((false_positive_rate >= BLOOM_MIN_FALSE_POSITIVE_RATE) &&
283 : (false_positive_rate < BLOOM_MAX_FALSE_POSITIVE_RATE));
284 :
285 : /* sizing bloom filter: -(n * ln(p)) / (ln(2))^2 */
286 3882 : nbits = ceil(-(ndistinct * log(false_positive_rate)) / pow(log(2.0), 2));
287 :
288 : /* round m to whole bytes */
289 3882 : nbytes = ((nbits + 7) / 8);
290 3882 : nbits = nbytes * 8;
291 :
292 : /*
293 : * Reject filters that are obviously too large to store on a page.
294 : *
295 : * Initially the bloom filter is just zeroes and so very compressible, but
296 : * as we add values it gets more and more random, and so less and less
297 : * compressible. So initially everything fits on the page, but we might
298 : * get surprising failures later - we want to prevent that, so we reject
299 : * bloom filter that are obviously too large.
300 : *
301 : * XXX It's not uncommon to oversize the bloom filter a bit, to defend
302 : * against unexpected data anomalies (parts of table with more distinct
303 : * values per range etc.). But we still need to make sure even the
304 : * oversized filter fits on page, if such need arises.
305 : *
306 : * XXX This check is not perfect, because the index may have multiple
307 : * filters that are small individually, but too large when combined.
308 : */
309 3882 : if (nbytes > BloomMaxFilterSize)
744 tomas.vondra 310 UBC 0 : elog(ERROR, "the bloom filter is too large (%d > %zu)", nbytes,
311 : BloomMaxFilterSize);
312 :
313 : /*
314 : * round(log(2.0) * m / ndistinct), but assume round() may not be
315 : * available on Windows
316 : */
744 tomas.vondra 317 CBC 3882 : k = log(2.0) * nbits / ndistinct;
318 3882 : k = (k - floor(k) >= 0.5) ? ceil(k) : floor(k);
319 :
320 : /*
321 : * We allocate the whole filter. Most of it is going to be 0 bits, so the
322 : * varlena is easy to compress.
323 : */
324 3882 : len = offsetof(BloomFilter, data) + nbytes;
325 :
326 3882 : filter = (BloomFilter *) palloc0(len);
327 :
328 3882 : filter->flags = 0;
329 3882 : filter->nhashes = (int) k;
330 3882 : filter->nbits = nbits;
331 :
332 3882 : SET_VARSIZE(filter, len);
333 :
334 3882 : return filter;
335 : }
336 :
337 :
338 : /*
339 : * bloom_add_value
340 : * Add value to the bloom filter.
341 : */
342 : static BloomFilter *
697 tgl 343 22023 : bloom_add_value(BloomFilter *filter, uint32 value, bool *updated)
344 : {
345 : int i;
346 : uint64 h1,
347 : h2;
348 :
349 : /* compute the hashes, used for the bloom filter */
744 tomas.vondra 350 22023 : h1 = hash_bytes_uint32_extended(value, BLOOM_SEED_1) % filter->nbits;
351 22023 : h2 = hash_bytes_uint32_extended(value, BLOOM_SEED_2) % filter->nbits;
352 :
353 : /* compute the requested number of hashes */
354 176184 : for (i = 0; i < filter->nhashes; i++)
355 : {
356 : /* h1 + h2 + f(i) */
357 154161 : uint32 h = (h1 + i * h2) % filter->nbits;
358 154161 : uint32 byte = (h / 8);
359 154161 : uint32 bit = (h % 8);
360 :
361 : /* if the bit is not set, set it and remember we did that */
362 154161 : if (!(filter->data[byte] & (0x01 << bit)))
363 : {
364 54402 : filter->data[byte] |= (0x01 << bit);
365 54402 : filter->nbits_set++;
366 54402 : if (updated)
367 54402 : *updated = true;
368 : }
369 : }
370 :
371 22023 : return filter;
372 : }
373 :
374 :
375 : /*
376 : * bloom_contains_value
377 : * Check if the bloom filter contains a particular value.
378 : */
379 : static bool
697 tgl 380 4104 : bloom_contains_value(BloomFilter *filter, uint32 value)
381 : {
382 : int i;
383 : uint64 h1,
384 : h2;
385 :
386 : /* calculate the two hashes */
744 tomas.vondra 387 4104 : h1 = hash_bytes_uint32_extended(value, BLOOM_SEED_1) % filter->nbits;
388 4104 : h2 = hash_bytes_uint32_extended(value, BLOOM_SEED_2) % filter->nbits;
389 :
390 : /* compute the requested number of hashes */
391 5217 : for (i = 0; i < filter->nhashes; i++)
392 : {
393 : /* h1 + h2 + f(i) */
394 5082 : uint32 h = (h1 + i * h2) % filter->nbits;
395 5082 : uint32 byte = (h / 8);
396 5082 : uint32 bit = (h % 8);
397 :
398 : /* if the bit is not set, the value is not there */
399 5082 : if (!(filter->data[byte] & (0x01 << bit)))
400 3969 : return false;
401 : }
402 :
403 : /* all hashes found in bloom filter */
404 135 : return true;
405 : }
406 :
407 : typedef struct BloomOpaque
408 : {
409 : /*
410 : * XXX At this point we only need a single proc (to compute the hash), but
411 : * let's keep the array just like inclusion and minmax opclasses, for
412 : * consistency. We may need additional procs in the future.
413 : */
414 : FmgrInfo extra_procinfos[BLOOM_MAX_PROCNUMS];
415 : bool extra_proc_missing[BLOOM_MAX_PROCNUMS];
416 : } BloomOpaque;
417 :
418 : static FmgrInfo *bloom_get_procinfo(BrinDesc *bdesc, uint16 attno,
419 : uint16 procnum);
420 :
421 :
422 : Datum
423 2300 : brin_bloom_opcinfo(PG_FUNCTION_ARGS)
424 : {
425 : BrinOpcInfo *result;
426 :
427 : /*
428 : * opaque->strategy_procinfos is initialized lazily; here it is set to
429 : * all-uninitialized by palloc0 which sets fn_oid to InvalidOid.
430 : *
431 : * bloom indexes only store the filter as a single BYTEA column
432 : */
433 :
434 2300 : result = palloc0(MAXALIGN(SizeofBrinOpcInfo(1)) +
435 : sizeof(BloomOpaque));
436 2300 : result->oi_nstored = 1;
437 2300 : result->oi_regular_nulls = true;
438 2300 : result->oi_opaque = (BloomOpaque *)
439 2300 : MAXALIGN((char *) result + SizeofBrinOpcInfo(1));
440 2300 : result->oi_typcache[0] = lookup_type_cache(PG_BRIN_BLOOM_SUMMARYOID, 0);
441 :
442 2300 : PG_RETURN_POINTER(result);
443 : }
444 :
445 : /*
446 : * brin_bloom_get_ndistinct
447 : * Determine the ndistinct value used to size bloom filter.
448 : *
449 : * Adjust the ndistinct value based on the pagesPerRange value. First,
450 : * if it's negative, it's assumed to be relative to maximum number of
451 : * tuples in the range (assuming each page gets MaxHeapTuplesPerPage
452 : * tuples, which is likely a significant over-estimate). We also clamp
453 : * the value, not to over-size the bloom filter unnecessarily.
454 : *
455 : * XXX We can only do this when the pagesPerRange value was supplied.
456 : * If it wasn't, it has to be a read-only access to the index, in which
457 : * case we don't really care. But perhaps we should fall-back to the
458 : * default pagesPerRange value?
459 : *
460 : * XXX We might also fetch info about ndistinct estimate for the column,
461 : * and compute the expected number of distinct values in a range. But
462 : * that may be tricky due to data being sorted in various ways, so it
463 : * seems better to rely on the upper estimate.
464 : *
465 : * XXX We might also calculate a better estimate of rows per BRIN range,
466 : * instead of using MaxHeapTuplesPerPage (which probably produces values
467 : * much higher than reality).
468 : */
469 : static int
470 3882 : brin_bloom_get_ndistinct(BrinDesc *bdesc, BloomOptions *opts)
471 : {
472 : double ndistinct;
473 : double maxtuples;
474 : BlockNumber pagesPerRange;
475 :
476 3882 : pagesPerRange = BrinGetPagesPerRange(bdesc->bd_index);
477 3882 : ndistinct = BloomGetNDistinctPerRange(opts);
478 :
479 3882 : Assert(BlockNumberIsValid(pagesPerRange));
480 :
481 3882 : maxtuples = MaxHeapTuplesPerPage * pagesPerRange;
482 :
483 : /*
484 : * Similarly to n_distinct, negative values are relative - in this case to
485 : * maximum number of tuples in the page range (maxtuples).
486 : */
487 3882 : if (ndistinct < 0)
488 3882 : ndistinct = (-ndistinct) * maxtuples;
489 :
490 : /*
491 : * Positive values are to be used directly, but we still apply a couple of
492 : * safeties to avoid using unreasonably small bloom filters.
493 : */
494 3882 : ndistinct = Max(ndistinct, BLOOM_MIN_NDISTINCT_PER_RANGE);
495 :
496 : /*
497 : * And don't use more than the maximum possible number of tuples, in the
498 : * range, which would be entirely wasteful.
499 : */
500 3882 : ndistinct = Min(ndistinct, maxtuples);
501 :
502 3882 : return (int) ndistinct;
503 : }
504 :
505 : /*
506 : * Examine the given index tuple (which contains partial status of a certain
507 : * page range) by comparing it to the given value that comes from another heap
508 : * tuple. If the new value is outside the bloom filter specified by the
509 : * existing tuple values, update the index tuple and return true. Otherwise,
510 : * return false and do not modify in this case.
511 : */
512 : Datum
513 22023 : brin_bloom_add_value(PG_FUNCTION_ARGS)
514 : {
515 22023 : BrinDesc *bdesc = (BrinDesc *) PG_GETARG_POINTER(0);
516 22023 : BrinValues *column = (BrinValues *) PG_GETARG_POINTER(1);
517 22023 : Datum newval = PG_GETARG_DATUM(2);
518 22023 : bool isnull PG_USED_FOR_ASSERTS_ONLY = PG_GETARG_DATUM(3);
519 22023 : BloomOptions *opts = (BloomOptions *) PG_GET_OPCLASS_OPTIONS();
520 22023 : Oid colloid = PG_GET_COLLATION();
521 : FmgrInfo *hashFn;
522 : uint32 hashValue;
523 22023 : bool updated = false;
524 : AttrNumber attno;
525 : BloomFilter *filter;
526 :
527 22023 : Assert(!isnull);
528 :
529 22023 : attno = column->bv_attno;
530 :
531 : /*
532 : * If this is the first non-null value, we need to initialize the bloom
533 : * filter. Otherwise just extract the existing bloom filter from
534 : * BrinValues.
535 : */
536 22023 : if (column->bv_allnulls)
537 : {
538 7764 : filter = bloom_init(brin_bloom_get_ndistinct(bdesc, opts),
539 3882 : BloomGetFalsePositiveRate(opts));
540 3882 : column->bv_values[0] = PointerGetDatum(filter);
541 3882 : column->bv_allnulls = false;
542 3882 : updated = true;
543 : }
544 : else
545 18141 : filter = (BloomFilter *) PG_DETOAST_DATUM(column->bv_values[0]);
546 :
547 : /*
548 : * Compute the hash of the new value, using the supplied hash function,
549 : * and then add the hash value to the bloom filter.
550 : */
551 22023 : hashFn = bloom_get_procinfo(bdesc, attno, PROCNUM_HASH);
552 :
553 22023 : hashValue = DatumGetUInt32(FunctionCall1Coll(hashFn, colloid, newval));
554 :
555 22023 : filter = bloom_add_value(filter, hashValue, &updated);
556 :
557 22023 : column->bv_values[0] = PointerGetDatum(filter);
558 :
559 22023 : PG_RETURN_BOOL(updated);
560 : }
561 :
562 : /*
563 : * Given an index tuple corresponding to a certain page range and a scan key,
564 : * return whether the scan key is consistent with the index tuple's bloom
565 : * filter. Return true if so, false otherwise.
566 : */
567 : Datum
568 4104 : brin_bloom_consistent(PG_FUNCTION_ARGS)
569 : {
570 4104 : BrinDesc *bdesc = (BrinDesc *) PG_GETARG_POINTER(0);
571 4104 : BrinValues *column = (BrinValues *) PG_GETARG_POINTER(1);
572 4104 : ScanKey *keys = (ScanKey *) PG_GETARG_POINTER(2);
573 4104 : int nkeys = PG_GETARG_INT32(3);
574 4104 : Oid colloid = PG_GET_COLLATION();
575 : AttrNumber attno;
576 : Datum value;
577 : Datum matches;
578 : FmgrInfo *finfo;
579 : uint32 hashValue;
580 : BloomFilter *filter;
581 : int keyno;
582 :
583 4104 : filter = (BloomFilter *) PG_DETOAST_DATUM(column->bv_values[0]);
584 :
585 4104 : Assert(filter);
586 :
587 4104 : matches = true;
588 :
589 4239 : for (keyno = 0; keyno < nkeys; keyno++)
590 : {
591 4104 : ScanKey key = keys[keyno];
592 :
593 : /* NULL keys are handled and filtered-out in bringetbitmap */
594 4104 : Assert(!(key->sk_flags & SK_ISNULL));
595 :
596 4104 : attno = key->sk_attno;
597 4104 : value = key->sk_argument;
598 :
599 4104 : switch (key->sk_strategy)
600 : {
601 4104 : case BloomEqualStrategyNumber:
602 :
603 : /*
604 : * In the equality case (WHERE col = someval), we want to
605 : * return the current page range if the minimum value in the
606 : * range <= scan key, and the maximum value >= scan key.
607 : */
608 4104 : finfo = bloom_get_procinfo(bdesc, attno, PROCNUM_HASH);
609 :
610 4104 : hashValue = DatumGetUInt32(FunctionCall1Coll(finfo, colloid, value));
611 4104 : matches &= bloom_contains_value(filter, hashValue);
612 :
613 4104 : break;
744 tomas.vondra 614 UBC 0 : default:
615 : /* shouldn't happen */
616 0 : elog(ERROR, "invalid strategy number %d", key->sk_strategy);
617 : matches = 0;
618 : break;
619 : }
620 :
744 tomas.vondra 621 CBC 4104 : if (!matches)
622 3969 : break;
623 : }
624 :
625 4104 : PG_RETURN_DATUM(matches);
626 : }
627 :
628 : /*
629 : * Given two BrinValues, update the first of them as a union of the summary
630 : * values contained in both. The second one is untouched.
631 : *
632 : * XXX We assume the bloom filters have the same parameters for now. In the
633 : * future we should have 'can union' function, to decide if we can combine
634 : * two particular bloom filters.
635 : */
636 : Datum
744 tomas.vondra 637 UBC 0 : brin_bloom_union(PG_FUNCTION_ARGS)
638 : {
639 : int i;
640 : int nbytes;
641 0 : BrinValues *col_a = (BrinValues *) PG_GETARG_POINTER(1);
642 0 : BrinValues *col_b = (BrinValues *) PG_GETARG_POINTER(2);
643 : BloomFilter *filter_a;
644 : BloomFilter *filter_b;
645 :
646 0 : Assert(col_a->bv_attno == col_b->bv_attno);
647 0 : Assert(!col_a->bv_allnulls && !col_b->bv_allnulls);
648 :
649 0 : filter_a = (BloomFilter *) PG_DETOAST_DATUM(col_a->bv_values[0]);
650 0 : filter_b = (BloomFilter *) PG_DETOAST_DATUM(col_b->bv_values[0]);
651 :
652 : /* make sure the filters use the same parameters */
653 0 : Assert(filter_a && filter_b);
654 0 : Assert(filter_a->nbits == filter_b->nbits);
655 0 : Assert(filter_a->nhashes == filter_b->nhashes);
656 0 : Assert((filter_a->nbits > 0) && (filter_a->nbits % 8 == 0));
657 :
658 0 : nbytes = (filter_a->nbits) / 8;
659 :
660 : /* simply OR the bitmaps */
661 0 : for (i = 0; i < nbytes; i++)
662 0 : filter_a->data[i] |= filter_b->data[i];
663 :
664 0 : PG_RETURN_VOID();
665 : }
666 :
667 : /*
668 : * Cache and return inclusion opclass support procedure
669 : *
670 : * Return the procedure corresponding to the given function support number
671 : * or null if it does not exist.
672 : */
673 : static FmgrInfo *
744 tomas.vondra 674 CBC 26127 : bloom_get_procinfo(BrinDesc *bdesc, uint16 attno, uint16 procnum)
675 : {
676 : BloomOpaque *opaque;
677 26127 : uint16 basenum = procnum - PROCNUM_BASE;
678 :
679 : /*
680 : * We cache these in the opaque struct, to avoid repetitive syscache
681 : * lookups.
682 : */
683 26127 : opaque = (BloomOpaque *) bdesc->bd_info[attno - 1]->oi_opaque;
684 :
685 : /*
686 : * If we already searched for this proc and didn't find it, don't bother
687 : * searching again.
688 : */
689 26127 : if (opaque->extra_proc_missing[basenum])
744 tomas.vondra 690 UBC 0 : return NULL;
691 :
744 tomas.vondra 692 CBC 26127 : if (opaque->extra_procinfos[basenum].fn_oid == InvalidOid)
693 : {
694 357 : if (RegProcedureIsValid(index_getprocid(bdesc->bd_index, attno,
695 : procnum)))
696 : {
697 357 : fmgr_info_copy(&opaque->extra_procinfos[basenum],
698 : index_getprocinfo(bdesc->bd_index, attno, procnum),
699 : bdesc->bd_context);
700 : }
701 : else
702 : {
744 tomas.vondra 703 UBC 0 : opaque->extra_proc_missing[basenum] = true;
704 0 : return NULL;
705 : }
706 : }
707 :
744 tomas.vondra 708 CBC 26127 : return &opaque->extra_procinfos[basenum];
709 : }
710 :
711 : Datum
712 248 : brin_bloom_options(PG_FUNCTION_ARGS)
713 : {
714 248 : local_relopts *relopts = (local_relopts *) PG_GETARG_POINTER(0);
715 :
716 248 : init_local_reloptions(relopts, sizeof(BloomOptions));
717 :
718 248 : add_local_real_reloption(relopts, "n_distinct_per_range",
719 : "number of distinct items expected in a BRIN page range",
720 : BLOOM_DEFAULT_NDISTINCT_PER_RANGE,
721 : -1.0, INT_MAX, offsetof(BloomOptions, nDistinctPerRange));
722 :
723 248 : add_local_real_reloption(relopts, "false_positive_rate",
724 : "desired false-positive rate for the bloom filters",
725 : BLOOM_DEFAULT_FALSE_POSITIVE_RATE,
726 : BLOOM_MIN_FALSE_POSITIVE_RATE,
727 : BLOOM_MAX_FALSE_POSITIVE_RATE,
728 : offsetof(BloomOptions, falsePositiveRate));
729 :
730 248 : PG_RETURN_VOID();
731 : }
732 :
733 : /*
734 : * brin_bloom_summary_in
735 : * - input routine for type brin_bloom_summary.
736 : *
737 : * brin_bloom_summary is only used internally to represent summaries
738 : * in BRIN bloom indexes, so it has no operations of its own, and we
739 : * disallow input too.
740 : */
741 : Datum
744 tomas.vondra 742 UBC 0 : brin_bloom_summary_in(PG_FUNCTION_ARGS)
743 : {
744 : /*
745 : * brin_bloom_summary stores the data in binary form and parsing text
746 : * input is not needed, so disallow this.
747 : */
748 0 : ereport(ERROR,
749 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
750 : errmsg("cannot accept a value of type %s", "pg_brin_bloom_summary")));
751 :
752 : PG_RETURN_VOID(); /* keep compiler quiet */
753 : }
754 :
755 :
756 : /*
757 : * brin_bloom_summary_out
758 : * - output routine for type brin_bloom_summary.
759 : *
760 : * BRIN bloom summaries are serialized into a bytea value, but we want
761 : * to output something nicer humans can understand.
762 : */
763 : Datum
764 0 : brin_bloom_summary_out(PG_FUNCTION_ARGS)
765 : {
766 : BloomFilter *filter;
767 : StringInfoData str;
768 :
769 : /* detoast the data to get value with a full 4B header */
224 peter 770 UNC 0 : filter = (BloomFilter *) PG_DETOAST_DATUM_PACKED(PG_GETARG_DATUM(0));
771 :
744 tomas.vondra 772 UBC 0 : initStringInfo(&str);
773 0 : appendStringInfoChar(&str, '{');
774 :
775 0 : appendStringInfo(&str, "mode: hashed nhashes: %u nbits: %u nbits_set: %u",
776 0 : filter->nhashes, filter->nbits, filter->nbits_set);
777 :
778 0 : appendStringInfoChar(&str, '}');
779 :
780 0 : PG_RETURN_CSTRING(str.data);
781 : }
782 :
783 : /*
784 : * brin_bloom_summary_recv
785 : * - binary input routine for type brin_bloom_summary.
786 : */
787 : Datum
788 0 : brin_bloom_summary_recv(PG_FUNCTION_ARGS)
789 : {
790 0 : ereport(ERROR,
791 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
792 : errmsg("cannot accept a value of type %s", "pg_brin_bloom_summary")));
793 :
794 : PG_RETURN_VOID(); /* keep compiler quiet */
795 : }
796 :
797 : /*
798 : * brin_bloom_summary_send
799 : * - binary output routine for type brin_bloom_summary.
800 : *
801 : * BRIN bloom summaries are serialized in a bytea value (although the
802 : * type is named differently), so let's just send that.
803 : */
804 : Datum
805 0 : brin_bloom_summary_send(PG_FUNCTION_ARGS)
806 : {
807 0 : return byteasend(fcinfo);
808 : }
|
