1 /*------------------------------------------------------------------------- 2 * 3 * array.h 4 * Declarations for Postgres arrays. 5 * 6 * A standard varlena array has the following internal structure: 7 * <vl_len_> - standard varlena header word 8 * <ndim> - number of dimensions of the array 9 * <dataoffset> - offset to stored data, or 0 if no nulls bitmap 10 * <elemtype> - element type OID 11 * <dimensions> - length of each array axis (C array of int) 12 * <lower bnds> - lower boundary of each dimension (C array of int) 13 * <null bitmap> - bitmap showing locations of nulls (OPTIONAL) 14 * <actual data> - whatever is the stored data 15 * 16 * The <dimensions> and <lower bnds> arrays each have ndim elements. 17 * 18 * The <null bitmap> may be omitted if the array contains no NULL elements. 19 * If it is absent, the <dataoffset> field is zero and the offset to the 20 * stored data must be computed on-the-fly. If the bitmap is present, 21 * <dataoffset> is nonzero and is equal to the offset from the array start 22 * to the first data element (including any alignment padding). The bitmap 23 * follows the same conventions as tuple null bitmaps, ie, a 1 indicates 24 * a non-null entry and the LSB of each bitmap byte is used first. 25 * 26 * The actual data starts on a MAXALIGN boundary. Individual items in the 27 * array are aligned as specified by the array element type. They are 28 * stored in row-major order (last subscript varies most rapidly). 29 * 30 * NOTE: it is important that array elements of toastable datatypes NOT be 31 * toasted, since the tupletoaster won't know they are there. (We could 32 * support compressed toasted items; only out-of-line items are dangerous. 33 * However, it seems preferable to store such items uncompressed and allow 34 * the toaster to compress the whole array as one input.) 35 * 36 * 37 * The OIDVECTOR and INT2VECTOR datatypes are storage-compatible with 38 * generic arrays, but they support only one-dimensional arrays with no 39 * nulls (and no null bitmap). They don't support being toasted, either. 40 * 41 * There are also some "fixed-length array" datatypes, such as NAME and 42 * POINT. These are simply a sequence of a fixed number of items each 43 * of a fixed-length datatype, with no overhead; the item size must be 44 * a multiple of its alignment requirement, because we do no padding. 45 * We support subscripting on these types, but array_in() and array_out() 46 * only work with varlena arrays. 47 * 48 * In addition, arrays are a major user of the "expanded object" TOAST 49 * infrastructure. This allows a varlena array to be converted to a 50 * separate representation that may include "deconstructed" Datum/isnull 51 * arrays holding the elements. 52 * 53 * 54 * Portions Copyright (c) 1996-2018, PostgreSQL Global Development Group 55 * Portions Copyright (c) 1994, Regents of the University of California 56 * 57 * src/include/utils/array.h 58 * 59 *------------------------------------------------------------------------- 60 */ 61 #ifndef ARRAY_H 62 #define ARRAY_H 63 64 #include "fmgr.h" 65 #include "utils/expandeddatum.h" 66 67 /* avoid including execnodes.h here */ 68 struct ExprState; 69 struct ExprContext; 70 71 72 /* 73 * Arrays are varlena objects, so must meet the varlena convention that 74 * the first int32 of the object contains the total object size in bytes. 75 * Be sure to use VARSIZE() and SET_VARSIZE() to access it, though! 76 * 77 * CAUTION: if you change the header for ordinary arrays you will also 78 * need to change the headers for oidvector and int2vector! 79 */ 80 typedef struct 81 { 82 int32 vl_len_; /* varlena header (do not touch directly!) */ 83 int ndim; /* # of dimensions */ 84 int32 dataoffset; /* offset to data, or 0 if no bitmap */ 85 Oid elemtype; /* element type OID */ 86 } ArrayType; 87 88 /* 89 * An expanded array is contained within a private memory context (as 90 * all expanded objects must be) and has a control structure as below. 91 * 92 * The expanded array might contain a regular "flat" array if that was the 93 * original input and we've not modified it significantly. Otherwise, the 94 * contents are represented by Datum/isnull arrays plus dimensionality and 95 * type information. We could also have both forms, if we've deconstructed 96 * the original array for access purposes but not yet changed it. For pass- 97 * by-reference element types, the Datums would point into the flat array in 98 * this situation. Once we start modifying array elements, new pass-by-ref 99 * elements are separately palloc'd within the memory context. 100 */ 101 #define EA_MAGIC 689375833 /* ID for debugging crosschecks */ 102 103 typedef struct ExpandedArrayHeader 104 { 105 /* Standard header for expanded objects */ 106 ExpandedObjectHeader hdr; 107 108 /* Magic value identifying an expanded array (for debugging only) */ 109 int ea_magic; 110 111 /* Dimensionality info (always valid) */ 112 int ndims; /* # of dimensions */ 113 int *dims; /* array dimensions */ 114 int *lbound; /* index lower bounds for each dimension */ 115 116 /* Element type info (always valid) */ 117 Oid element_type; /* element type OID */ 118 int16 typlen; /* needed info about element datatype */ 119 bool typbyval; 120 char typalign; 121 122 /* 123 * If we have a Datum-array representation of the array, it's kept here; 124 * else dvalues/dnulls are NULL. The dvalues and dnulls arrays are always 125 * palloc'd within the object private context, but may change size from 126 * time to time. For pass-by-ref element types, dvalues entries might 127 * point either into the fstartptr..fendptr area, or to separately 128 * palloc'd chunks. Elements should always be fully detoasted, as they 129 * are in the standard flat representation. 130 * 131 * Even when dvalues is valid, dnulls can be NULL if there are no null 132 * elements. 133 */ 134 Datum *dvalues; /* array of Datums */ 135 bool *dnulls; /* array of is-null flags for Datums */ 136 int dvalueslen; /* allocated length of above arrays */ 137 int nelems; /* number of valid entries in above arrays */ 138 139 /* 140 * flat_size is the current space requirement for the flat equivalent of 141 * the expanded array, if known; otherwise it's 0. We store this to make 142 * consecutive calls of get_flat_size cheap. 143 */ 144 Size flat_size; 145 146 /* 147 * fvalue points to the flat representation if it is valid, else it is 148 * NULL. If we have or ever had a flat representation then 149 * fstartptr/fendptr point to the start and end+1 of its data area; this 150 * is so that we can tell which Datum pointers point into the flat 151 * representation rather than being pointers to separately palloc'd data. 152 */ 153 ArrayType *fvalue; /* must be a fully detoasted array */ 154 char *fstartptr; /* start of its data area */ 155 char *fendptr; /* end+1 of its data area */ 156 } ExpandedArrayHeader; 157 158 /* 159 * Functions that can handle either a "flat" varlena array or an expanded 160 * array use this union to work with their input. Don't refer to "flt"; 161 * instead, cast to ArrayType. This struct nominally requires 8-byte 162 * alignment on 64-bit, but it's often used for an ArrayType having 4-byte 163 * alignment. UBSan complains about referencing "flt" in such cases. 164 */ 165 typedef union AnyArrayType 166 { 167 ArrayType flt; 168 ExpandedArrayHeader xpn; 169 } AnyArrayType; 170 171 /* 172 * working state for accumArrayResult() and friends 173 * note that the input must be scalars (legal array elements) 174 */ 175 typedef struct ArrayBuildState 176 { 177 MemoryContext mcontext; /* where all the temp stuff is kept */ 178 Datum *dvalues; /* array of accumulated Datums */ 179 bool *dnulls; /* array of is-null flags for Datums */ 180 int alen; /* allocated length of above arrays */ 181 int nelems; /* number of valid entries in above arrays */ 182 Oid element_type; /* data type of the Datums */ 183 int16 typlen; /* needed info about datatype */ 184 bool typbyval; 185 char typalign; 186 bool private_cxt; /* use private memory context */ 187 } ArrayBuildState; 188 189 /* 190 * working state for accumArrayResultArr() and friends 191 * note that the input must be arrays, and the same array type is returned 192 */ 193 typedef struct ArrayBuildStateArr 194 { 195 MemoryContext mcontext; /* where all the temp stuff is kept */ 196 char *data; /* accumulated data */ 197 bits8 *nullbitmap; /* bitmap of is-null flags, or NULL if none */ 198 int abytes; /* allocated length of "data" */ 199 int nbytes; /* number of bytes used so far */ 200 int aitems; /* allocated length of bitmap (in elements) */ 201 int nitems; /* total number of elements in result */ 202 int ndims; /* current dimensions of result */ 203 int dims[MAXDIM]; 204 int lbs[MAXDIM]; 205 Oid array_type; /* data type of the arrays */ 206 Oid element_type; /* data type of the array elements */ 207 bool private_cxt; /* use private memory context */ 208 } ArrayBuildStateArr; 209 210 /* 211 * working state for accumArrayResultAny() and friends 212 * these functions handle both cases 213 */ 214 typedef struct ArrayBuildStateAny 215 { 216 /* Exactly one of these is not NULL: */ 217 ArrayBuildState *scalarstate; 218 ArrayBuildStateArr *arraystate; 219 } ArrayBuildStateAny; 220 221 /* 222 * structure to cache type metadata needed for array manipulation 223 */ 224 typedef struct ArrayMetaState 225 { 226 Oid element_type; 227 int16 typlen; 228 bool typbyval; 229 char typalign; 230 char typdelim; 231 Oid typioparam; 232 Oid typiofunc; 233 FmgrInfo proc; 234 } ArrayMetaState; 235 236 /* 237 * private state needed by array_map (here because caller must provide it) 238 */ 239 typedef struct ArrayMapState 240 { 241 ArrayMetaState inp_extra; 242 ArrayMetaState ret_extra; 243 } ArrayMapState; 244 245 /* ArrayIteratorData is private in arrayfuncs.c */ 246 typedef struct ArrayIteratorData *ArrayIterator; 247 248 /* fmgr macros for regular varlena array objects */ 249 #define DatumGetArrayTypeP(X) ((ArrayType *) PG_DETOAST_DATUM(X)) 250 #define DatumGetArrayTypePCopy(X) ((ArrayType *) PG_DETOAST_DATUM_COPY(X)) 251 #define PG_GETARG_ARRAYTYPE_P(n) DatumGetArrayTypeP(PG_GETARG_DATUM(n)) 252 #define PG_GETARG_ARRAYTYPE_P_COPY(n) DatumGetArrayTypePCopy(PG_GETARG_DATUM(n)) 253 #define PG_RETURN_ARRAYTYPE_P(x) PG_RETURN_POINTER(x) 254 255 /* fmgr macros for expanded array objects */ 256 #define PG_GETARG_EXPANDED_ARRAY(n) DatumGetExpandedArray(PG_GETARG_DATUM(n)) 257 #define PG_GETARG_EXPANDED_ARRAYX(n, metacache) \ 258 DatumGetExpandedArrayX(PG_GETARG_DATUM(n), metacache) 259 #define PG_RETURN_EXPANDED_ARRAY(x) PG_RETURN_DATUM(EOHPGetRWDatum(&(x)->hdr)) 260 261 /* fmgr macros for AnyArrayType (ie, get either varlena or expanded form) */ 262 #define PG_GETARG_ANY_ARRAY_P(n) DatumGetAnyArrayP(PG_GETARG_DATUM(n)) 263 264 /* 265 * Access macros for varlena array header fields. 266 * 267 * ARR_DIMS returns a pointer to an array of array dimensions (number of 268 * elements along the various array axes). 269 * 270 * ARR_LBOUND returns a pointer to an array of array lower bounds. 271 * 272 * That is: if the third axis of an array has elements 5 through 8, then 273 * ARR_DIMS(a)[2] == 4 and ARR_LBOUND(a)[2] == 5. 274 * 275 * Unlike C, the default lower bound is 1. 276 */ 277 #define ARR_SIZE(a) VARSIZE(a) 278 #define ARR_NDIM(a) ((a)->ndim) 279 #define ARR_HASNULL(a) ((a)->dataoffset != 0) 280 #define ARR_ELEMTYPE(a) ((a)->elemtype) 281 282 #define ARR_DIMS(a) \ 283 ((int *) (((char *) (a)) + sizeof(ArrayType))) 284 #define ARR_LBOUND(a) \ 285 ((int *) (((char *) (a)) + sizeof(ArrayType) + \ 286 sizeof(int) * ARR_NDIM(a))) 287 288 #define ARR_NULLBITMAP(a) \ 289 (ARR_HASNULL(a) ? \ 290 (bits8 *) (((char *) (a)) + sizeof(ArrayType) + \ 291 2 * sizeof(int) * ARR_NDIM(a)) \ 292 : (bits8 *) NULL) 293 294 /* 295 * The total array header size (in bytes) for an array with the specified 296 * number of dimensions and total number of items. 297 */ 298 #define ARR_OVERHEAD_NONULLS(ndims) \ 299 MAXALIGN(sizeof(ArrayType) + 2 * sizeof(int) * (ndims)) 300 #define ARR_OVERHEAD_WITHNULLS(ndims, nitems) \ 301 MAXALIGN(sizeof(ArrayType) + 2 * sizeof(int) * (ndims) + \ 302 ((nitems) + 7) / 8) 303 304 #define ARR_DATA_OFFSET(a) \ 305 (ARR_HASNULL(a) ? (a)->dataoffset : ARR_OVERHEAD_NONULLS(ARR_NDIM(a))) 306 307 /* 308 * Returns a pointer to the actual array data. 309 */ 310 #define ARR_DATA_PTR(a) \ 311 (((char *) (a)) + ARR_DATA_OFFSET(a)) 312 313 /* 314 * Macros for working with AnyArrayType inputs. Beware multiple references! 315 */ 316 #define AARR_NDIM(a) \ 317 (VARATT_IS_EXPANDED_HEADER(a) ? \ 318 (a)->xpn.ndims : ARR_NDIM((ArrayType *) (a))) 319 #define AARR_HASNULL(a) \ 320 (VARATT_IS_EXPANDED_HEADER(a) ? \ 321 ((a)->xpn.dvalues != NULL ? (a)->xpn.dnulls != NULL : ARR_HASNULL((a)->xpn.fvalue)) : \ 322 ARR_HASNULL((ArrayType *) (a))) 323 #define AARR_ELEMTYPE(a) \ 324 (VARATT_IS_EXPANDED_HEADER(a) ? \ 325 (a)->xpn.element_type : ARR_ELEMTYPE((ArrayType *) (a))) 326 #define AARR_DIMS(a) \ 327 (VARATT_IS_EXPANDED_HEADER(a) ? \ 328 (a)->xpn.dims : ARR_DIMS((ArrayType *) (a))) 329 #define AARR_LBOUND(a) \ 330 (VARATT_IS_EXPANDED_HEADER(a) ? \ 331 (a)->xpn.lbound : ARR_LBOUND((ArrayType *) (a))) 332 333 334 /* 335 * GUC parameter 336 */ 337 extern bool Array_nulls; 338 339 /* 340 * prototypes for functions defined in arrayfuncs.c 341 */ 342 extern void CopyArrayEls(ArrayType *array, 343 Datum *values, 344 bool *nulls, 345 int nitems, 346 int typlen, 347 bool typbyval, 348 char typalign, 349 bool freedata); 350 351 extern Datum array_get_element(Datum arraydatum, int nSubscripts, int *indx, 352 int arraytyplen, int elmlen, bool elmbyval, char elmalign, 353 bool *isNull); 354 extern Datum array_set_element(Datum arraydatum, int nSubscripts, int *indx, 355 Datum dataValue, bool isNull, 356 int arraytyplen, int elmlen, bool elmbyval, char elmalign); 357 extern Datum array_get_slice(Datum arraydatum, int nSubscripts, 358 int *upperIndx, int *lowerIndx, 359 bool *upperProvided, bool *lowerProvided, 360 int arraytyplen, int elmlen, bool elmbyval, char elmalign); 361 extern Datum array_set_slice(Datum arraydatum, int nSubscripts, 362 int *upperIndx, int *lowerIndx, 363 bool *upperProvided, bool *lowerProvided, 364 Datum srcArrayDatum, bool isNull, 365 int arraytyplen, int elmlen, bool elmbyval, char elmalign); 366 367 extern Datum array_ref(ArrayType *array, int nSubscripts, int *indx, 368 int arraytyplen, int elmlen, bool elmbyval, char elmalign, 369 bool *isNull); 370 extern ArrayType *array_set(ArrayType *array, int nSubscripts, int *indx, 371 Datum dataValue, bool isNull, 372 int arraytyplen, int elmlen, bool elmbyval, char elmalign); 373 374 extern Datum array_map(Datum arrayd, 375 struct ExprState *exprstate, struct ExprContext *econtext, 376 Oid retType, ArrayMapState *amstate); 377 378 extern void array_bitmap_copy(bits8 *destbitmap, int destoffset, 379 const bits8 *srcbitmap, int srcoffset, 380 int nitems); 381 382 extern ArrayType *construct_array(Datum *elems, int nelems, 383 Oid elmtype, 384 int elmlen, bool elmbyval, char elmalign); 385 extern ArrayType *construct_md_array(Datum *elems, 386 bool *nulls, 387 int ndims, 388 int *dims, 389 int *lbs, 390 Oid elmtype, int elmlen, bool elmbyval, char elmalign); 391 extern ArrayType *construct_empty_array(Oid elmtype); 392 extern ExpandedArrayHeader *construct_empty_expanded_array(Oid element_type, 393 MemoryContext parentcontext, 394 ArrayMetaState *metacache); 395 extern void deconstruct_array(ArrayType *array, 396 Oid elmtype, 397 int elmlen, bool elmbyval, char elmalign, 398 Datum **elemsp, bool **nullsp, int *nelemsp); 399 extern bool array_contains_nulls(ArrayType *array); 400 401 extern ArrayBuildState *initArrayResult(Oid element_type, 402 MemoryContext rcontext, bool subcontext); 403 extern ArrayBuildState *accumArrayResult(ArrayBuildState *astate, 404 Datum dvalue, bool disnull, 405 Oid element_type, 406 MemoryContext rcontext); 407 extern Datum makeArrayResult(ArrayBuildState *astate, 408 MemoryContext rcontext); 409 extern Datum makeMdArrayResult(ArrayBuildState *astate, int ndims, 410 int *dims, int *lbs, MemoryContext rcontext, bool release); 411 412 extern ArrayBuildStateArr *initArrayResultArr(Oid array_type, Oid element_type, 413 MemoryContext rcontext, bool subcontext); 414 extern ArrayBuildStateArr *accumArrayResultArr(ArrayBuildStateArr *astate, 415 Datum dvalue, bool disnull, 416 Oid array_type, 417 MemoryContext rcontext); 418 extern Datum makeArrayResultArr(ArrayBuildStateArr *astate, 419 MemoryContext rcontext, bool release); 420 421 extern ArrayBuildStateAny *initArrayResultAny(Oid input_type, 422 MemoryContext rcontext, bool subcontext); 423 extern ArrayBuildStateAny *accumArrayResultAny(ArrayBuildStateAny *astate, 424 Datum dvalue, bool disnull, 425 Oid input_type, 426 MemoryContext rcontext); 427 extern Datum makeArrayResultAny(ArrayBuildStateAny *astate, 428 MemoryContext rcontext, bool release); 429 430 extern ArrayIterator array_create_iterator(ArrayType *arr, int slice_ndim, ArrayMetaState *mstate); 431 extern bool array_iterate(ArrayIterator iterator, Datum *value, bool *isnull); 432 extern void array_free_iterator(ArrayIterator iterator); 433 434 /* 435 * prototypes for functions defined in arrayutils.c 436 */ 437 438 extern int ArrayGetOffset(int n, const int *dim, const int *lb, const int *indx); 439 extern int ArrayGetOffset0(int n, const int *tup, const int *scale); 440 extern int ArrayGetNItems(int ndim, const int *dims); 441 extern void ArrayCheckBounds(int ndim, const int *dims, const int *lb); 442 extern void mda_get_range(int n, int *span, const int *st, const int *endp); 443 extern void mda_get_prod(int n, const int *range, int *prod); 444 extern void mda_get_offset_values(int n, int *dist, const int *prod, const int *span); 445 extern int mda_next_tuple(int n, int *curr, const int *span); 446 extern int32 *ArrayGetIntegerTypmods(ArrayType *arr, int *n); 447 448 /* 449 * prototypes for functions defined in array_expanded.c 450 */ 451 extern Datum expand_array(Datum arraydatum, MemoryContext parentcontext, 452 ArrayMetaState *metacache); 453 extern ExpandedArrayHeader *DatumGetExpandedArray(Datum d); 454 extern ExpandedArrayHeader *DatumGetExpandedArrayX(Datum d, 455 ArrayMetaState *metacache); 456 extern AnyArrayType *DatumGetAnyArrayP(Datum d); 457 extern void deconstruct_expanded_array(ExpandedArrayHeader *eah); 458 459 #endif /* ARRAY_H */ 460