forked from duckdb/postgres_scanner
-
Notifications
You must be signed in to change notification settings - Fork 0
/
postgres_scanner.cpp
1159 lines (991 loc) · 38.4 KB
/
postgres_scanner.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#define DUCKDB_BUILD_LOADABLE_EXTENSION
#include "duckdb.hpp"
#include <libpq-fe.h>
#include <arpa/inet.h>
// htonll is not available on Linux it seems
#ifndef ntohll
#define ntohll(x) ((((uint64_t)ntohl(x & 0xFFFFFFFF)) << 32) + ntohl(x >> 32))
#endif
#include "duckdb/parser/parsed_data/create_table_function_info.hpp"
#include "duckdb/planner/table_filter.hpp"
#include "duckdb/planner/filter/conjunction_filter.hpp"
#include "duckdb/planner/filter/constant_filter.hpp"
using namespace duckdb;
struct PostgresTypeInfo {
string typname;
int64_t typlen;
string typtype;
string nspname;
};
struct PostgresColumnInfo {
string attname;
int atttypmod;
PostgresTypeInfo type_info;
int64_t typelem; // OID pointer for arrays
PostgresTypeInfo elem_info;
};
static constexpr uint32_t POSTGRES_TID_MAX = 4294967295;
struct PostgresBindData : public FunctionData {
~PostgresBindData() {
if (conn) {
PQfinish(conn);
conn = nullptr;
}
}
string schema_name;
string table_name;
idx_t pages_approx = 0;
vector<PostgresColumnInfo> columns;
vector<string> names;
vector<LogicalType> types;
vector<bool> needs_cast;
idx_t pages_per_task = 1000;
string dsn;
string snapshot;
bool in_recovery;
PGconn *conn = nullptr;
public:
unique_ptr<FunctionData> Copy() const override {
throw NotImplementedException("");
}
bool Equals(const FunctionData &other) const override {
throw NotImplementedException("");
}
};
struct PGQueryResult {
PGQueryResult(PGresult *res_p) : res(res_p) {
}
~PGQueryResult() {
if (res) {
PQclear(res);
}
}
PGresult *res = nullptr;
public:
string GetString(idx_t row, idx_t col) {
D_ASSERT(res);
return string(PQgetvalue(res, row, col));
}
int32_t GetInt32(idx_t row, idx_t col) {
return atoi(PQgetvalue(res, row, col));
}
int64_t GetInt64(idx_t row, idx_t col) {
return atoll(PQgetvalue(res, row, col));
}
bool GetBool(idx_t row, idx_t col) {
return strcmp(PQgetvalue(res, row, col), "t");
}
idx_t Count() {
D_ASSERT(res);
return PQntuples(res);
}
};
struct PostgresLocalState : public LocalTableFunctionState {
~PostgresLocalState() {
if (conn) {
PQfinish(conn);
conn = nullptr;
}
}
bool done = false;
bool exec = false;
string sql;
vector<column_t> column_ids;
TableFilterSet *filters;
string col_names;
PGconn *conn = nullptr;
};
struct PostgresGlobalState : public GlobalTableFunctionState {
PostgresGlobalState(idx_t max_threads) : page_idx(0), max_threads(max_threads) {
}
mutex lock;
idx_t page_idx;
idx_t max_threads;
idx_t MaxThreads() const override {
return max_threads;
}
};
static PGconn *PGConnect(string &dsn) {
PGconn *conn = PQconnectdb(dsn.c_str());
// both PQStatus and PQerrorMessage check for nullptr
if (PQstatus(conn) == CONNECTION_BAD) {
throw IOException("Unable to connect to Postgres at %s: %s", dsn, string(PQerrorMessage(conn)));
}
return conn;
}
static unique_ptr<PGQueryResult> PGQuery(PGconn *conn, string q, ExecStatusType response_code = PGRES_TUPLES_OK) {
auto res = make_uniq<PGQueryResult>(PQexec(conn, q.c_str()));
if (!res->res || PQresultStatus(res->res) != response_code) {
throw IOException("Unable to query Postgres: %s %s", string(PQerrorMessage(conn)),
string(PQresultErrorMessage(res->res)));
}
return res;
}
static void PGExec(PGconn *conn, string q) {
PGQuery(conn, q, PGRES_COMMAND_OK);
}
static LogicalType DuckDBType2(PostgresTypeInfo *type_info, int atttypmod, PostgresTypeInfo *ele_info, PGconn *conn,
ClientContext &context) {
auto &pgtypename = type_info->typname;
// TODO better check, does the typtyp say something here?
// postgres array types start with an _
if (StringUtil::StartsWith(pgtypename, "_")) {
return LogicalType::LIST(DuckDBType2(ele_info, atttypmod, nullptr, conn, context));
}
if (type_info->typtype == "e") { // ENUM
auto res = PGQuery(
conn, StringUtil::Format("SELECT unnest(enum_range(NULL::%s.%s))", type_info->nspname, type_info->typname));
Vector duckdb_levels(LogicalType::VARCHAR, res->Count());
for (idx_t row = 0; row < res->Count(); row++) {
duckdb_levels.SetValue(row, res->GetString(row, 0));
}
return LogicalType::ENUM("postgres_enum_" + pgtypename, duckdb_levels, res->Count());
}
if (pgtypename == "bool") {
return LogicalType::BOOLEAN;
} else if (pgtypename == "int2") {
return LogicalType::SMALLINT;
} else if (pgtypename == "int4") {
return LogicalType::INTEGER;
} else if (pgtypename == "int8") {
return LogicalType::BIGINT;
} else if (pgtypename == "oid") { // "The oid type is currently implemented as an unsigned four-byte integer."
return LogicalType::UINTEGER;
} else if (pgtypename == "float4") {
return LogicalType::FLOAT;
} else if (pgtypename == "float8") {
return LogicalType::DOUBLE;
} else if (pgtypename == "numeric") {
if (atttypmod == -1) { // unbounded decimal/numeric, will just return as double
return LogicalType::DOUBLE;
}
auto width = ((atttypmod - sizeof(int32_t)) >> 16) & 0xffff;
auto scale = (((atttypmod - sizeof(int32_t)) & 0x7ff) ^ 1024) - 1024;
return LogicalType::DECIMAL(width, scale);
} else if (pgtypename == "char" || pgtypename == "bpchar" || pgtypename == "varchar" || pgtypename == "text" ||
pgtypename == "jsonb" || pgtypename == "json") {
return LogicalType::VARCHAR;
} else if (pgtypename == "date") {
return LogicalType::DATE;
} else if (pgtypename == "bytea") {
return LogicalType::BLOB;
} else if (pgtypename == "time") {
return LogicalType::TIME;
} else if (pgtypename == "timetz") {
return LogicalType::TIME_TZ;
} else if (pgtypename == "timestamp") {
return LogicalType::TIMESTAMP;
} else if (pgtypename == "timestamptz") {
return LogicalType::TIMESTAMP_TZ;
} else if (pgtypename == "interval") {
return LogicalType::INTERVAL;
} else if (pgtypename == "uuid") {
return LogicalType::UUID;
} else {
return LogicalType::INVALID;
}
}
static LogicalType DuckDBType(PostgresColumnInfo &info, PGconn *conn, ClientContext &context) {
return DuckDBType2(&info.type_info, info.atttypmod, &info.elem_info, conn, context);
}
static unique_ptr<FunctionData> PostgresBind(ClientContext &context, TableFunctionBindInput &input,
vector<LogicalType> &return_types, vector<string> &names) {
auto bind_data = make_uniq<PostgresBindData>();
bind_data->dsn = input.inputs[0].GetValue<string>();
bind_data->schema_name = input.inputs[1].GetValue<string>();
bind_data->table_name = input.inputs[2].GetValue<string>();
bind_data->conn = PGConnect(bind_data->dsn);
// we create a transaction here, and get the snapshot id so the parallel
// reader threads can use the same snapshot
PGExec(bind_data->conn, "BEGIN TRANSACTION ISOLATION LEVEL REPEATABLE READ READ ONLY");
bind_data->in_recovery = (bool)PGQuery(bind_data->conn, "SELECT pg_is_in_recovery()")->GetBool(0, 0);
bind_data->snapshot = "";
if (!bind_data->in_recovery) {
bind_data->snapshot = PGQuery(bind_data->conn, "SELECT pg_export_snapshot()")->GetString(0, 0);
}
// find the id of the table in question to simplify below queries and avoid
// complex joins (ha)
auto res = PGQuery(bind_data->conn, StringUtil::Format(R"(
SELECT pg_class.oid, GREATEST(relpages, 1)
FROM pg_class JOIN pg_namespace ON relnamespace = pg_namespace.oid
WHERE nspname='%s' AND relname='%s'
)",
bind_data->schema_name, bind_data->table_name));
if (res->Count() != 1) {
throw InvalidInputException("Postgres table \"%s\".\"%s\" not found", bind_data->schema_name,
bind_data->table_name);
}
auto oid = res->GetInt64(0, 0);
bind_data->pages_approx = res->GetInt64(0, 1);
res.reset();
// query the table schema so we can interpret the bits in the pages
// fun fact: this query also works in DuckDB ^^
res = PGQuery(bind_data->conn, StringUtil::Format(
R"(
SELECT
attname, atttypmod, pg_namespace.nspname,
pg_type.typname, pg_type.typlen, pg_type.typtype, pg_type.typelem,
pg_type_elem.typname elem_typname, pg_type_elem.typlen elem_typlen, pg_type_elem.typtype elem_typtype
FROM pg_attribute
JOIN pg_type ON atttypid=pg_type.oid
LEFT JOIN pg_type pg_type_elem ON pg_type.typelem=pg_type_elem.oid
LEFT JOIN pg_namespace ON pg_type.typnamespace = pg_namespace.oid
WHERE attrelid=%d AND attnum > 0
ORDER BY attnum;
)",
oid));
// can't scan a table without columns (yes those exist)
if (res->Count() == 0) {
throw InvalidInputException("Table %s does not contain any columns.", bind_data->table_name);
}
for (idx_t row = 0; row < res->Count(); row++) {
PostgresColumnInfo info;
info.attname = res->GetString(row, 0);
info.atttypmod = res->GetInt32(row, 1);
info.type_info.nspname = res->GetString(row, 2);
info.type_info.typname = res->GetString(row, 3);
info.type_info.typlen = res->GetInt64(row, 4);
info.type_info.typtype = res->GetString(row, 5);
info.typelem = res->GetInt64(row, 6);
info.elem_info.nspname = res->GetString(row, 2);
info.elem_info.typname = res->GetString(row, 7);
info.elem_info.typlen = res->GetInt64(row, 8);
info.elem_info.typtype = res->GetString(row, 9);
bind_data->names.push_back(info.attname);
auto duckdb_type = DuckDBType(info, bind_data->conn, context);
// we cast unsupported types to varchar on read
auto needs_cast = duckdb_type == LogicalType::INVALID;
bind_data->needs_cast.push_back(needs_cast);
if (!needs_cast) {
bind_data->types.push_back(std::move(duckdb_type));
} else {
bind_data->types.push_back(LogicalType::VARCHAR);
}
bind_data->columns.push_back(info);
}
res.reset();
return_types = bind_data->types;
names = bind_data->names;
return std::move(bind_data);
}
static string TransformFilter(string &column_name, TableFilter &filter);
static string CreateExpression(string &column_name, vector<unique_ptr<TableFilter>> &filters, string op) {
vector<string> filter_entries;
for (auto &filter : filters) {
filter_entries.push_back(TransformFilter(column_name, *filter));
}
return "(" + StringUtil::Join(filter_entries, " " + op + " ") + ")";
}
static string TransformComparision(ExpressionType type) {
switch (type) {
case ExpressionType::COMPARE_EQUAL:
return "=";
case ExpressionType::COMPARE_NOTEQUAL:
return "!=";
case ExpressionType::COMPARE_LESSTHAN:
return "<";
case ExpressionType::COMPARE_GREATERTHAN:
return ">";
case ExpressionType::COMPARE_LESSTHANOREQUALTO:
return "<=";
case ExpressionType::COMPARE_GREATERTHANOREQUALTO:
return ">=";
default:
throw NotImplementedException("Unsupported expression type");
}
}
static string TransformFilter(string &column_name, TableFilter &filter) {
switch (filter.filter_type) {
case TableFilterType::IS_NULL:
return column_name + " IS NULL";
case TableFilterType::IS_NOT_NULL:
return column_name + " IS NOT NULL";
case TableFilterType::CONJUNCTION_AND: {
auto &conjunction_filter = (ConjunctionAndFilter &)filter;
return CreateExpression(column_name, conjunction_filter.child_filters, "AND");
}
case TableFilterType::CONJUNCTION_OR: {
auto &conjunction_filter = (ConjunctionAndFilter &)filter;
return CreateExpression(column_name, conjunction_filter.child_filters, "OR");
}
case TableFilterType::CONSTANT_COMPARISON: {
auto &constant_filter = (ConstantFilter &)filter;
// TODO properly escape ' in constant value
auto constant_string = "'" + constant_filter.constant.ToString() + "'";
auto operator_string = TransformComparision(constant_filter.comparison_type);
return StringUtil::Format("%s %s %s", column_name, operator_string, constant_string);
}
default:
throw InternalException("Unsupported table filter type");
}
}
static void PostgresInitInternal(ClientContext &context, const PostgresBindData *bind_data_p,
PostgresLocalState &lstate, idx_t task_min, idx_t task_max) {
D_ASSERT(bind_data_p);
D_ASSERT(task_min <= task_max);
auto bind_data = (const PostgresBindData *)bind_data_p;
// Queries like SELECT count(*) do not require actually returning the columns from the
// Postgres table, but only counting the row. In this case, we will be asked to return
// the 'rowid' special column here. It must be the only selected column. The corresponding
// deparsed query will be 'SELECT NULL'..Note that the user is not allowed to explicitly
// request the 'rowid' special column from a Postgres table in a SQL query.
bool have_rowid = false;
for (idx_t i = 0; i < lstate.column_ids.size(); i++) {
if (lstate.column_ids[i] == (column_t)-1) {
have_rowid = true;
break;
}
}
if (have_rowid && lstate.column_ids.size() > 1) {
throw InternalException("Cannot return ROW_ID from Postgres table");
}
std::string col_names;
if (have_rowid) {
// We are only counting rows, not interested in the actual values of the columns.
col_names = "NULL";
} else {
col_names =
StringUtil::Join(lstate.column_ids.data(), lstate.column_ids.size(), ", ", [&](const idx_t column_id) {
return StringUtil::Format("\"%s\"%s", bind_data->names[column_id],
bind_data->needs_cast[column_id] ? "::VARCHAR" : "");
});
}
string filter_string;
if (lstate.filters && !lstate.filters->filters.empty()) {
vector<string> filter_entries;
for (auto &entry : lstate.filters->filters) {
// TODO properly escape " in column names
auto column_name = "\"" + bind_data->names[lstate.column_ids[entry.first]] + "\"";
auto &filter = *entry.second;
filter_entries.push_back(TransformFilter(column_name, filter));
}
filter_string = " AND " + StringUtil::Join(filter_entries, " AND ");
}
lstate.sql = StringUtil::Format(
R"(
COPY (SELECT %s FROM "%s"."%s" WHERE ctid BETWEEN '(%d,0)'::tid AND '(%d,0)'::tid %s) TO STDOUT (FORMAT binary);
)",
col_names, bind_data->schema_name, bind_data->table_name, task_min, task_max, filter_string);
lstate.exec = false;
lstate.done = false;
}
static PGconn *PostgresScanConnect(string dsn, bool in_recovery, string snapshot) {
auto conn = PGConnect(dsn);
PGExec(conn, "BEGIN TRANSACTION ISOLATION LEVEL REPEATABLE READ READ ONLY");
if (!in_recovery) {
PGExec(conn, StringUtil::Format("SET TRANSACTION SNAPSHOT '%s'", snapshot));
}
return conn;
}
#define POSTGRES_EPOCH_JDATE 2451545 /* == date2j(2000, 1, 1) */
#define NBASE 10000
#define DEC_DIGITS 4 /* decimal digits per NBASE digit */
/*
* Interpretation of high bits.
*/
#define NUMERIC_SIGN_MASK 0xC000
#define NUMERIC_POS 0x0000
#define NUMERIC_NEG 0x4000
#define NUMERIC_SHORT 0x8000
#define NUMERIC_SPECIAL 0xC000
/*
* Definitions for special values (NaN, positive infinity, negative infinity).
*
* The two bits after the NUMERIC_SPECIAL bits are 00 for NaN, 01 for positive
* infinity, 11 for negative infinity. (This makes the sign bit match where
* it is in a short-format value, though we make no use of that at present.)
* We could mask off the remaining bits before testing the active bits, but
* currently those bits must be zeroes, so masking would just add cycles.
*/
#define NUMERIC_EXT_SIGN_MASK 0xF000 /* high bits plus NaN/Inf flag bits */
#define NUMERIC_NAN 0xC000
#define NUMERIC_PINF 0xD000
#define NUMERIC_NINF 0xF000
#define NUMERIC_INF_SIGN_MASK 0x2000
#define NUMERIC_EXT_FLAGBITS(n) ((n)->choice.n_header & NUMERIC_EXT_SIGN_MASK)
#define NUMERIC_IS_NAN(n) ((n)->choice.n_header == NUMERIC_NAN)
#define NUMERIC_IS_PINF(n) ((n)->choice.n_header == NUMERIC_PINF)
#define NUMERIC_IS_NINF(n) ((n)->choice.n_header == NUMERIC_NINF)
#define NUMERIC_IS_INF(n) (((n)->choice.n_header & ~NUMERIC_INF_SIGN_MASK) == NUMERIC_PINF)
/*
* Short format definitions.
*/
#define NUMERIC_DSCALE_MASK 0x3FFF
#define NUMERIC_SHORT_SIGN_MASK 0x2000
#define NUMERIC_SHORT_DSCALE_MASK 0x1F80
#define NUMERIC_SHORT_DSCALE_SHIFT 7
#define NUMERIC_SHORT_DSCALE_MAX (NUMERIC_SHORT_DSCALE_MASK >> NUMERIC_SHORT_DSCALE_SHIFT)
#define NUMERIC_SHORT_WEIGHT_SIGN_MASK 0x0040
#define NUMERIC_SHORT_WEIGHT_MASK 0x003F
#define NUMERIC_SHORT_WEIGHT_MAX NUMERIC_SHORT_WEIGHT_MASK
#define NUMERIC_SHORT_WEIGHT_MIN (-(NUMERIC_SHORT_WEIGHT_MASK + 1))
#define NUMERIC_SIGN(is_short, header1) \
(is_short ? ((header1 & NUMERIC_SHORT_SIGN_MASK) ? NUMERIC_NEG : NUMERIC_POS) : (header1 & NUMERIC_SIGN_MASK))
#define NUMERIC_DSCALE(is_short, header1) \
(is_short ? (header1 & NUMERIC_SHORT_DSCALE_MASK) >> NUMERIC_SHORT_DSCALE_SHIFT : (header1 & NUMERIC_DSCALE_MASK))
#define NUMERIC_WEIGHT(is_short, header1, header2) \
(is_short ? ((header1 & NUMERIC_SHORT_WEIGHT_SIGN_MASK ? ~NUMERIC_SHORT_WEIGHT_MASK : 0) | \
(header1 & NUMERIC_SHORT_WEIGHT_MASK)) \
: (header2))
// copied from cast_helpers.cpp because windows linking issues
static const int64_t POWERS_OF_TEN[] {1,
10,
100,
1000,
10000,
100000,
1000000,
10000000,
100000000,
1000000000,
10000000000,
100000000000,
1000000000000,
10000000000000,
100000000000000,
1000000000000000,
10000000000000000,
100000000000000000,
1000000000000000000};
template <class T>
T LoadEndIncrement(const_data_ptr_t &pspsptr) {
T val = Load<T>(pspsptr);
if (sizeof(T) == sizeof(uint16_t)) {
val = ntohs(val);
} else if (sizeof(T) == sizeof(uint32_t)) {
val = ntohl(val);
} else if (sizeof(T) == sizeof(uint64_t)) {
val = ntohll(val);
} else {
D_ASSERT(0);
}
pspsptr += sizeof(T);
return val;
}
struct PostgresDecimalConfig {
uint16_t scale;
uint16_t ndigits;
int16_t weight;
bool is_negative;
};
static PostgresDecimalConfig ReadDecimalConfig(const_data_ptr_t &value_ptr) {
PostgresDecimalConfig config;
config.ndigits = LoadEndIncrement<uint16_t>(value_ptr);
config.weight = LoadEndIncrement<int16_t>(value_ptr);
auto sign = LoadEndIncrement<uint16_t>(value_ptr);
if (!(sign == NUMERIC_POS || sign == NUMERIC_NAN || sign == NUMERIC_PINF || sign == NUMERIC_NINF ||
sign == NUMERIC_NEG)) {
throw NotImplementedException("Postgres numeric NA/Inf");
}
config.is_negative = sign == NUMERIC_NEG;
config.scale = LoadEndIncrement<uint16_t>(value_ptr);
return config;
};
template <class T>
static T ReadDecimal(PostgresDecimalConfig &config, const_data_ptr_t value_ptr) {
// this is wild
auto scale_POWER = POWERS_OF_TEN[config.scale];
if (config.ndigits == 0) {
return 0;
}
T integral_part = 0, fractional_part = 0;
if (config.weight >= 0) {
D_ASSERT(config.weight <= config.ndigits);
integral_part = LoadEndIncrement<uint16_t>(value_ptr);
for (auto i = 1; i <= config.weight; i++) {
integral_part *= NBASE;
if (i < config.ndigits) {
integral_part += LoadEndIncrement<uint16_t>(value_ptr);
}
}
integral_part *= scale_POWER;
}
if (config.ndigits > config.weight + 1) {
fractional_part = LoadEndIncrement<uint16_t>(value_ptr);
for (auto i = config.weight + 2; i < config.ndigits; i++) {
fractional_part *= NBASE;
if (i < config.ndigits) {
fractional_part += LoadEndIncrement<uint16_t>(value_ptr);
}
}
// we need to find out how large the fractional part is in terms of powers
// of ten this depends on how many times we multiplied with NBASE
// if that is different from scale, we need to divide the extra part away
// again
// similarly, if trailing zeroes have been suppressed, we have not been multiplying t
// the fractional part with NBASE often enough. If so, add additional powers
auto fractional_power = (config.ndigits - config.weight - 1) * DEC_DIGITS;
auto fractional_power_correction = fractional_power - config.scale;
D_ASSERT(fractional_power_correction < 20);
if (fractional_power_correction >= 0) {
fractional_part /= POWERS_OF_TEN[fractional_power_correction];
} else {
fractional_part *= POWERS_OF_TEN[-fractional_power_correction];
}
}
// finally
auto base_res = (integral_part + fractional_part);
return (config.is_negative ? -base_res : base_res);
}
static void ProcessValue(const LogicalType &type, const PostgresTypeInfo *type_info, int atttypmod, int64_t typelem,
const PostgresTypeInfo *elem_info, const_data_ptr_t value_ptr, idx_t value_len,
Vector &out_vec, idx_t output_offset) {
switch (type.id()) {
case LogicalTypeId::SMALLINT:
D_ASSERT(value_len == sizeof(int16_t));
FlatVector::GetData<int16_t>(out_vec)[output_offset] = LoadEndIncrement<int16_t>(value_ptr);
break;
case LogicalTypeId::INTEGER:
D_ASSERT(value_len == sizeof(int32_t));
FlatVector::GetData<int32_t>(out_vec)[output_offset] = LoadEndIncrement<int32_t>(value_ptr);
break;
case LogicalTypeId::UINTEGER:
D_ASSERT(value_len == sizeof(uint32_t));
FlatVector::GetData<uint32_t>(out_vec)[output_offset] = LoadEndIncrement<uint32_t>(value_ptr);
break;
case LogicalTypeId::BIGINT:
D_ASSERT(value_len == sizeof(int64_t));
FlatVector::GetData<int64_t>(out_vec)[output_offset] = LoadEndIncrement<int64_t>(value_ptr);
break;
case LogicalTypeId::FLOAT: {
D_ASSERT(value_len == sizeof(float));
auto i = LoadEndIncrement<uint32_t>(value_ptr);
FlatVector::GetData<float>(out_vec)[output_offset] = *((float *)&i);
break;
}
case LogicalTypeId::DOUBLE: {
if (type_info->typname ==
"numeric") { // this was an unbounded decimal, read params from value and cast to double
auto config = ReadDecimalConfig(value_ptr);
auto val = ReadDecimal<int64_t>(config, value_ptr);
FlatVector::GetData<double>(out_vec)[output_offset] = (double)val / POWERS_OF_TEN[config.scale];
break;
}
D_ASSERT(value_len == sizeof(double));
auto i = LoadEndIncrement<uint64_t>(value_ptr);
FlatVector::GetData<double>(out_vec)[output_offset] = *((double *)&i);
break;
}
case LogicalTypeId::BLOB:
case LogicalTypeId::VARCHAR: {
if (type_info->typname == "jsonb") {
auto version = Load<uint8_t>(value_ptr);
value_ptr++;
value_len--;
if (version != 1) {
throw NotImplementedException("JSONB version number mismatch, expected 1, got %d", version);
}
}
FlatVector::GetData<string_t>(out_vec)[output_offset] =
StringVector::AddStringOrBlob(out_vec, (char *)value_ptr, value_len);
break;
}
case LogicalTypeId::BOOLEAN:
D_ASSERT(value_len == sizeof(bool));
FlatVector::GetData<bool>(out_vec)[output_offset] = *value_ptr > 0;
break;
case LogicalTypeId::DECIMAL: {
if (value_len < sizeof(uint16_t) * 4) {
throw InvalidInputException("Need at least 8 bytes to read a Postgres decimal. Got %d", value_len);
}
auto decimal_config = ReadDecimalConfig(value_ptr);
D_ASSERT(decimal_config.scale == DecimalType::GetScale(type));
switch (type.InternalType()) {
case PhysicalType::INT16:
FlatVector::GetData<int16_t>(out_vec)[output_offset] = ReadDecimal<int16_t>(decimal_config, value_ptr);
break;
case PhysicalType::INT32:
FlatVector::GetData<int32_t>(out_vec)[output_offset] = ReadDecimal<int32_t>(decimal_config, value_ptr);
break;
case PhysicalType::INT64:
FlatVector::GetData<int64_t>(out_vec)[output_offset] = ReadDecimal<int64_t>(decimal_config, value_ptr);
break;
case PhysicalType::INT128:
FlatVector::GetData<hugeint_t>(out_vec)[output_offset] = ReadDecimal<hugeint_t>(decimal_config, value_ptr);
break;
default:
throw InvalidInputException("Unsupported decimal storage type");
}
break;
}
case LogicalTypeId::DATE: {
D_ASSERT(value_len == sizeof(int32_t));
auto jd = LoadEndIncrement<uint32_t>(value_ptr);
auto out_ptr = FlatVector::GetData<date_t>(out_vec);
out_ptr[output_offset].days = jd + POSTGRES_EPOCH_JDATE - 2440588; // magic!
break;
}
case LogicalTypeId::TIME: {
D_ASSERT(value_len == sizeof(int64_t));
D_ASSERT(atttypmod == -1);
FlatVector::GetData<dtime_t>(out_vec)[output_offset].micros = LoadEndIncrement<uint64_t>(value_ptr);
break;
}
case LogicalTypeId::TIME_TZ: {
D_ASSERT(value_len == sizeof(int64_t) + sizeof(int32_t));
D_ASSERT(atttypmod == -1);
auto usec = LoadEndIncrement<uint64_t>(value_ptr);
auto tzoffset = LoadEndIncrement<int32_t>(value_ptr);
FlatVector::GetData<dtime_t>(out_vec)[output_offset].micros = usec + tzoffset * Interval::MICROS_PER_SEC;
break;
}
case LogicalTypeId::TIMESTAMP_TZ:
case LogicalTypeId::TIMESTAMP: {
D_ASSERT(value_len == sizeof(int64_t));
D_ASSERT(atttypmod == -1);
auto usec = ntohll(Load<uint64_t>(value_ptr));
auto time = usec % Interval::MICROS_PER_DAY;
// adjust date
auto date = (usec / Interval::MICROS_PER_DAY) + POSTGRES_EPOCH_JDATE - 2440588;
// glue it back together
FlatVector::GetData<timestamp_t>(out_vec)[output_offset].value = date * Interval::MICROS_PER_DAY + time;
break;
}
case LogicalTypeId::ENUM: {
auto enum_val = string((const char *)value_ptr, value_len);
auto offset = EnumType::GetPos(type, enum_val);
if (offset < 0) {
throw IOException("Could not map ENUM value %s", enum_val);
}
switch (type.InternalType()) {
case PhysicalType::UINT8:
FlatVector::GetData<uint8_t>(out_vec)[output_offset] = (uint8_t)offset;
break;
case PhysicalType::UINT16:
FlatVector::GetData<uint16_t>(out_vec)[output_offset] = (uint16_t)offset;
break;
case PhysicalType::UINT32:
FlatVector::GetData<uint32_t>(out_vec)[output_offset] = (uint32_t)offset;
break;
default:
throw InternalException("ENUM can only have unsigned integers (except "
"UINT64) as physical types, got %s",
TypeIdToString(type.InternalType()));
}
break;
}
case LogicalTypeId::INTERVAL: {
if (atttypmod != -1) {
throw IOException("Interval with unsupported typmod %d", atttypmod);
}
interval_t res;
res.micros = LoadEndIncrement<uint64_t>(value_ptr);
res.days = LoadEndIncrement<uint32_t>(value_ptr);
res.months = LoadEndIncrement<uint32_t>(value_ptr);
FlatVector::GetData<interval_t>(out_vec)[output_offset] = res;
break;
}
case LogicalTypeId::UUID: {
D_ASSERT(value_len == 2 * sizeof(int64_t));
D_ASSERT(atttypmod == -1);
hugeint_t res;
auto upper = LoadEndIncrement<uint64_t>(value_ptr);
res.upper = upper ^ (int64_t(1) << 63);
res.lower = LoadEndIncrement<uint64_t>(value_ptr);
FlatVector::GetData<hugeint_t>(out_vec)[output_offset] = res;
break;
}
case LogicalTypeId::LIST: {
D_ASSERT(elem_info);
auto &list_entry = FlatVector::GetData<list_entry_t>(out_vec)[output_offset];
auto child_offset = ListVector::GetListSize(out_vec);
if (value_len < 1) {
list_entry.offset = ListVector::GetListSize(out_vec);
list_entry.length = 0;
break;
}
D_ASSERT(value_len >= 3 * sizeof(uint32_t));
auto flag_one = LoadEndIncrement<uint32_t>(value_ptr);
auto flag_two = LoadEndIncrement<uint32_t>(value_ptr);
if (flag_one == 0) {
list_entry.offset = child_offset;
list_entry.length = 0;
return;
}
// D_ASSERT(flag_two == 1); // TODO what is this?!
auto value_oid = LoadEndIncrement<uint32_t>(value_ptr);
D_ASSERT(value_oid == typelem);
auto array_length = LoadEndIncrement<uint32_t>(value_ptr);
auto array_dim = LoadEndIncrement<uint32_t>(value_ptr);
if (array_dim != 1) {
throw NotImplementedException("Only one-dimensional Postgres arrays are supported %u %u ", array_length,
array_dim);
}
auto &child_vec = ListVector::GetEntry(out_vec);
ListVector::Reserve(out_vec, child_offset + array_length);
for (idx_t child_idx = 0; child_idx < array_length; child_idx++) {
// handle NULLs again (TODO: unify this with scan)
auto ele_len = LoadEndIncrement<int32_t>(value_ptr);
if (ele_len == -1) { // NULL
FlatVector::Validity(child_vec).Set(child_offset + child_idx, false);
continue;
}
ProcessValue(ListType::GetChildType(type), elem_info, atttypmod, 0, nullptr, value_ptr, ele_len, child_vec,
child_offset + child_idx);
value_ptr += ele_len;
}
ListVector::SetListSize(out_vec, child_offset + array_length);
list_entry.offset = child_offset;
list_entry.length = array_length;
break;
}
default:
throw InternalException("Unsupported Type %s", type.ToString());
}
}
struct PostgresBinaryBuffer {
PostgresBinaryBuffer(PGconn *conn_p) : conn(conn_p) {
D_ASSERT(conn);
}
void Next() {
Reset();
len = PQgetCopyData(conn, &buffer, 0);
// len -2 is error
// len -1 is supposed to signal end but does not actually happen in practise
// we expect at least 2 bytes in each message for the tuple count
if (!buffer || len < sizeof(int16_t)) {
throw IOException("Unable to read binary COPY data from Postgres: %s", string(PQerrorMessage(conn)));
}
buffer_ptr = buffer;
}
void Reset() {
if (buffer) {
PQfreemem(buffer);
}
buffer = nullptr;
buffer_ptr = nullptr;
len = 0;
}
bool Ready() {
return buffer_ptr != nullptr;
}
~PostgresBinaryBuffer() {
Reset();
}
void CheckHeader() {
auto magic_len = 11;
auto flags_len = 8;
auto header_len = magic_len + flags_len;
if (len < header_len) {
throw IOException("Unable to read binary COPY data from Postgres, invalid header");
}
if (!memcmp(buffer_ptr, "PGCOPY\\n\\377\\r\\n\\0", magic_len)) {
throw IOException("Expected Postgres binary COPY header, got something else");
}
buffer_ptr += header_len;
// as far as i can tell the "Flags field" and the "Header
// extension area length" do not contain anything interesting
}
template <typename T>
const T Read() {
T ret;
D_ASSERT(len > 0);
D_ASSERT(buffer);
D_ASSERT(buffer_ptr);
memcpy(&ret, buffer_ptr, sizeof(ret));
buffer_ptr += sizeof(T);
return ret;
}
char *buffer = nullptr, *buffer_ptr = nullptr;
int len = 0;
PGconn *conn = nullptr;
};
static idx_t PostgresMaxThreads(ClientContext &context, const FunctionData *bind_data_p) {
D_ASSERT(bind_data_p);
auto bind_data = (const PostgresBindData *)bind_data_p;
return bind_data->pages_approx / bind_data->pages_per_task;
}
static unique_ptr<GlobalTableFunctionState> PostgresInitGlobalState(ClientContext &context,
TableFunctionInitInput &input) {
return make_uniq<PostgresGlobalState>(PostgresMaxThreads(context, input.bind_data.get()));
}
static bool PostgresParallelStateNext(ClientContext &context, const FunctionData *bind_data_p,
PostgresLocalState &lstate, PostgresGlobalState &gstate) {
D_ASSERT(bind_data_p);
auto bind_data = (const PostgresBindData *)bind_data_p;
lock_guard<mutex> parallel_lock(gstate.lock);
if (gstate.page_idx < bind_data->pages_approx) {
auto page_max = gstate.page_idx + bind_data->pages_per_task;
if (page_max >= bind_data->pages_approx) {
// the relpages entry is not the real max, so make the last task bigger
page_max = POSTGRES_TID_MAX;
}
PostgresInitInternal(context, bind_data, lstate, gstate.page_idx, page_max);
gstate.page_idx += bind_data->pages_per_task;
return true;
}
lstate.done = true;
return false;
}
static unique_ptr<LocalTableFunctionState> PostgresInitLocalState(ExecutionContext &context,
TableFunctionInitInput &input,
GlobalTableFunctionState *global_state) {
auto &bind_data = input.bind_data->Cast<PostgresBindData>();
auto &gstate = global_state->Cast<PostgresGlobalState>();
auto local_state = make_uniq<PostgresLocalState>();
local_state->column_ids = input.column_ids;
local_state->conn = PostgresScanConnect(bind_data.dsn, bind_data.in_recovery, bind_data.snapshot);
local_state->filters = input.filters.get();
if (!PostgresParallelStateNext(context.client, input.bind_data.get(), *local_state, gstate)) {
local_state->done = true;
}
return std::move(local_state);
}
static void PostgresScan(ClientContext &context, TableFunctionInput &data, DataChunk &output) {
auto &bind_data = data.bind_data->Cast<PostgresBindData>();
auto &local_state = data.local_state->Cast<PostgresLocalState>();
auto &gstate = data.global_state->Cast<PostgresGlobalState>();
idx_t output_offset = 0;
PostgresBinaryBuffer buf(local_state.conn);
while (true) {
if (local_state.done && !PostgresParallelStateNext(context, data.bind_data.get(), local_state, gstate)) {
return;
}
if (!local_state.exec) {
PGQuery(local_state.conn, local_state.sql, PGRES_COPY_OUT);
local_state.exec = true;
buf.Next();
buf.CheckHeader();
// the first tuple immediately follows the header in the first message, so
// we have to keep the buffer alive for now.
}
output.SetCardinality(output_offset);
if (output_offset == STANDARD_VECTOR_SIZE) {
return;
}
if (!buf.Ready()) {
buf.Next();
}
auto tuple_count = (int16_t)ntohs(buf.Read<uint16_t>());