1CREATE INDEX(7) PostgreSQL 15.4 Documentation CREATE INDEX(7)
2
6 CREATE_INDEX - define a new index
7
9 CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] [ [ IF NOT EXISTS ] name ] ON [ ONLY ] table_name [ USING method ]
10 ( { column_name | ( expression ) } [ COLLATE collation ] [ opclass [ ( opclass_parameter = value [, ... ] ) ] ] [ ASC | DESC ] [ NULLS { FIRST | LAST } ] [, ...] )
11 [ INCLUDE ( column_name [, ...] ) ]
12 [ NULLS [ NOT ] DISTINCT ]
13 [ WITH ( storage_parameter [= value] [, ... ] ) ]
14 [ TABLESPACE tablespace_name ]
15 [ WHERE predicate ]
16
18 CREATE INDEX constructs an index on the specified column(s) of the
19 specified relation, which can be a table or a materialized view.
20 Indexes are primarily used to enhance database performance (though
21 inappropriate use can result in slower performance).
22 The key field(s) for the index are specified as column names, or
24 alternatively as expressions written in parentheses. Multiple fields
25 can be specified if the index method supports multicolumn indexes.
26 An index field can be an expression computed from the values of one or
28 more columns of the table row. This feature can be used to obtain fast
29 access to data based on some transformation of the basic data. For
30 example, an index computed on upper(col) would allow the clause WHERE
31 upper(col) = 'JIM' to use an index.
32 PostgreSQL provides the index methods B-tree, hash, GiST, SP-GiST, GIN,
34 and BRIN. Users can also define their own index methods, but that is
35 fairly complicated.
36 When the WHERE clause is present, a partial index is created. A partial
38 index is an index that contains entries for only a portion of a table,
39 usually a portion that is more useful for indexing than the rest of the
40 table. For example, if you have a table that contains both billed and
41 unbilled orders where the unbilled orders take up a small fraction of
42 the total table and yet that is an often used section, you can improve
43 performance by creating an index on just that portion. Another possible
44 application is to use WHERE with UNIQUE to enforce uniqueness over a
45 subset of a table. See Section 11.8 for more discussion.
46 The expression used in the WHERE clause can refer only to columns of
48 the underlying table, but it can use all columns, not just the ones
49 being indexed. Presently, subqueries and aggregate expressions are also
50 forbidden in WHERE. The same restrictions apply to index fields that
51 are expressions.
52 All functions and operators used in an index definition must be
54 “immutable”, that is, their results must depend only on their arguments
55 and never on any outside influence (such as the contents of another
56 table or the current time). This restriction ensures that the behavior
57 of the index is well-defined. To use a user-defined function in an
58 index expression or WHERE clause, remember to mark the function
59 immutable when you create it.
60
62 UNIQUE
63 Causes the system to check for duplicate values in the table when
64 the index is created (if data already exist) and each time data is
65 added. Attempts to insert or update data which would result in
66 duplicate entries will generate an error.
67 Additional restrictions apply when unique indexes are applied to
69 partitioned tables; see CREATE TABLE (CREATE_TABLE(7)).
70 CONCURRENTLY
72 When this option is used, PostgreSQL will build the index without
73 taking any locks that prevent concurrent inserts, updates, or
74 deletes on the table; whereas a standard index build locks out
75 writes (but not reads) on the table until it's done. There are
76 several caveats to be aware of when using this option — see
77 Building Indexes Concurrently below.
78 For temporary tables, CREATE INDEX is always non-concurrent, as no
80 other session can access them, and non-concurrent index creation is
81 cheaper.
82 IF NOT EXISTS
84 Do not throw an error if a relation with the same name already
85 exists. A notice is issued in this case. Note that there is no
86 guarantee that the existing index is anything like the one that
87 would have been created. Index name is required when IF NOT EXISTS
88 is specified.
89 INCLUDE
91 The optional INCLUDE clause specifies a list of columns which will
92 be included in the index as non-key columns. A non-key column
93 cannot be used in an index scan search qualification, and it is
94 disregarded for purposes of any uniqueness or exclusion constraint
95 enforced by the index. However, an index-only scan can return the
96 contents of non-key columns without having to visit the index's
97 table, since they are available directly from the index entry.
98 Thus, addition of non-key columns allows index-only scans to be
99 used for queries that otherwise could not use them.
100 It's wise to be conservative about adding non-key columns to an
102 index, especially wide columns. If an index tuple exceeds the
103 maximum size allowed for the index type, data insertion will fail.
104 In any case, non-key columns duplicate data from the index's table
105 and bloat the size of the index, thus potentially slowing searches.
106 Furthermore, B-tree deduplication is never used with indexes that
107 have a non-key column.
108 Columns listed in the INCLUDE clause don't need appropriate
110 operator classes; the clause can include columns whose data types
111 don't have operator classes defined for a given access method.
112 Expressions are not supported as included columns since they cannot
114 be used in index-only scans.
115 Currently, the B-tree, GiST and SP-GiST index access methods
117 support this feature. In these indexes, the values of columns
118 listed in the INCLUDE clause are included in leaf tuples which
119 correspond to heap tuples, but are not included in upper-level
120 index entries used for tree navigation.
121 name
123 The name of the index to be created. No schema name can be included
124 here; the index is always created in the same schema as its parent
125 table. The name of the index must be distinct from the name of any
126 other relation (table, sequence, index, view, materialized view, or
127 foreign table) in that schema. If the name is omitted, PostgreSQL
128 chooses a suitable name based on the parent table's name and the
129 indexed column name(s).
130 ONLY
132 Indicates not to recurse creating indexes on partitions, if the
133 table is partitioned. The default is to recurse.
134 table_name
136 The name (possibly schema-qualified) of the table to be indexed.
137 method
139 The name of the index method to be used. Choices are btree, hash,
140 gist, spgist, gin, brin, or user-installed access methods like
141 bloom. The default method is btree.
142 column_name
144 The name of a column of the table.
145 expression
147 An expression based on one or more columns of the table. The
148 expression usually must be written with surrounding parentheses, as
149 shown in the syntax. However, the parentheses can be omitted if the
150 expression has the form of a function call.
151 collation
153 The name of the collation to use for the index. By default, the
154 index uses the collation declared for the column to be indexed or
155 the result collation of the expression to be indexed. Indexes with
156 non-default collations can be useful for queries that involve
157 expressions using non-default collations.
158 opclass
160 The name of an operator class. See below for details.
161 opclass_parameter
163 The name of an operator class parameter. See below for details.
164 ASC
166 Specifies ascending sort order (which is the default).
167 DESC
169 Specifies descending sort order.
170 NULLS FIRST
172 Specifies that nulls sort before non-nulls. This is the default
173 when DESC is specified.
174 NULLS LAST
176 Specifies that nulls sort after non-nulls. This is the default when
177 DESC is not specified.
178 NULLS DISTINCT
180 NULLS NOT DISTINCT
181 Specifies whether for a unique index, null values should be
182 considered distinct (not equal). The default is that they are
183 distinct, so that a unique index could contain multiple null values
184 in a column.
185 storage_parameter
187 The name of an index-method-specific storage parameter. See Index
188 Storage Parameters below for details.
189 tablespace_name
191 The tablespace in which to create the index. If not specified,
192 default_tablespace is consulted, or temp_tablespaces for indexes on
193 temporary tables.
194 predicate
196 The constraint expression for a partial index.
197 Index Storage Parameters
199 The optional WITH clause specifies storage parameters for the index.
200 Each index method has its own set of allowed storage parameters. The
201 B-tree, hash, GiST and SP-GiST index methods all accept this parameter:
202 fillfactor (integer)
204 The fillfactor for an index is a percentage that determines how
205 full the index method will try to pack index pages. For B-trees,
206 leaf pages are filled to this percentage during initial index
207 builds, and also when extending the index at the right (adding new
208 largest key values). If pages subsequently become completely full,
209 they will be split, leading to fragmentation of the on-disk index
210 structure. B-trees use a default fillfactor of 90, but any integer
211 value from 10 to 100 can be selected.
212 B-tree indexes on tables where many inserts and/or updates are
214 anticipated can benefit from lower fillfactor settings at CREATE
215 INDEX time (following bulk loading into the table). Values in the
216 range of 50 - 90 can usefully “smooth out” the rate of page splits
217 during the early life of the B-tree index (lowering fillfactor like
218 this may even lower the absolute number of page splits, though this
219 effect is highly workload dependent). The B-tree bottom-up index
220 deletion technique described in Section 67.4.2 is dependent on
221 having some “extra” space on pages to store “extra” tuple versions,
222 and so can be affected by fillfactor (though the effect is usually
223 not significant).
224 In other specific cases it might be useful to increase fillfactor
226 to 100 at CREATE INDEX time as a way of maximizing space
227 utilization. You should only consider this when you are completely
228 sure that the table is static (i.e. that it will never be affected
229 by either inserts or updates). A fillfactor setting of 100
230 otherwise risks harming performance: even a few updates or inserts
231 will cause a sudden flood of page splits.
232 The other index methods use fillfactor in different but roughly
234 analogous ways; the default fillfactor varies between methods.
235 B-tree indexes additionally accept this parameter:
237 deduplicate_items (boolean)
239 Controls usage of the B-tree deduplication technique described in
240 Section 67.4.3. Set to ON or OFF to enable or disable the
241 optimization. (Alternative spellings of ON and OFF are allowed as
242 described in Section 20.1.) The default is ON.
243 Note
245 Turning deduplicate_items off via ALTER INDEX prevents future
246 insertions from triggering deduplication, but does not in
247 itself make existing posting list tuples use the standard tuple
248 representation.
249 GiST indexes additionally accept this parameter:
251 buffering (enum)
253 Determines whether the buffered build technique described in
254 Section 68.4.1 is used to build the index. With OFF buffering is
255 disabled, with ON it is enabled, and with AUTO it is initially
256 disabled, but is turned on on-the-fly once the index size reaches
257 effective_cache_size. The default is AUTO. Note that if sorted
258 build is possible, it will be used instead of buffered build unless
259 buffering=ON is specified.
260 GIN indexes accept different parameters:
262 fastupdate (boolean)
264 This setting controls usage of the fast update technique described
265 in Section 70.4.1. It is a Boolean parameter: ON enables fast
266 update, OFF disables it. The default is ON.
267 Note
269 Turning fastupdate off via ALTER INDEX prevents future
270 insertions from going into the list of pending index entries,
271 but does not in itself flush previous entries. You might want
272 to VACUUM the table or call gin_clean_pending_list function
273 afterward to ensure the pending list is emptied.
274 gin_pending_list_limit (integer)
276 Custom gin_pending_list_limit parameter. This value is specified in
277 kilobytes.
278 BRIN indexes accept different parameters:
280 pages_per_range (integer)
282 Defines the number of table blocks that make up one block range for
283 each entry of a BRIN index (see Section 71.1 for more details). The
284 default is 128.
285 autosummarize (boolean)
287 Defines whether a summarization run is queued for the previous page
288 range whenever an insertion is detected on the next one. See
289 Section 71.1.1 for more details. The default is off.
290 Building Indexes Concurrently
292 Creating an index can interfere with regular operation of a database.
293 Normally PostgreSQL locks the table to be indexed against writes and
294 performs the entire index build with a single scan of the table. Other
295 transactions can still read the table, but if they try to insert,
296 update, or delete rows in the table they will block until the index
297 build is finished. This could have a severe effect if the system is a
298 live production database. Very large tables can take many hours to be
299 indexed, and even for smaller tables, an index build can lock out
300 writers for periods that are unacceptably long for a production system.
301 PostgreSQL supports building indexes without locking out writes. This
303 method is invoked by specifying the CONCURRENTLY option of CREATE
304 INDEX. When this option is used, PostgreSQL must perform two scans of
305 the table, and in addition it must wait for all existing transactions
306 that could potentially modify or use the index to terminate. Thus this
307 method requires more total work than a standard index build and takes
308 significantly longer to complete. However, since it allows normal
309 operations to continue while the index is built, this method is useful
310 for adding new indexes in a production environment. Of course, the
311 extra CPU and I/O load imposed by the index creation might slow other
312 operations.
313 In a concurrent index build, the index is actually entered as an
315 “invalid” index into the system catalogs in one transaction, then two
316 table scans occur in two more transactions. Before each table scan, the
317 index build must wait for existing transactions that have modified the
318 table to terminate. After the second scan, the index build must wait
319 for any transactions that have a snapshot (see Chapter 13) predating
320 the second scan to terminate, including transactions used by any phase
321 of concurrent index builds on other tables, if the indexes involved are
322 partial or have columns that are not simple column references. Then
323 finally the index can be marked “valid” and ready for use, and the
324 CREATE INDEX command terminates. Even then, however, the index may not
325 be immediately usable for queries: in the worst case, it cannot be used
326 as long as transactions exist that predate the start of the index
327 build.
328 If a problem arises while scanning the table, such as a deadlock or a
330 uniqueness violation in a unique index, the CREATE INDEX command will
331 fail but leave behind an “invalid” index. This index will be ignored
332 for querying purposes because it might be incomplete; however it will
333 still consume update overhead. The psql \d command will report such an
334 index as INVALID:
335 postgres=# \d tab
337 Table "public.tab"
338 Column | Type | Collation | Nullable | Default
339 --------+---------+-----------+----------+---------
340 col | integer | | |
341 Indexes:
342 "idx" btree (col) INVALID
343 The recommended recovery method in such cases is to drop the index and
345 try again to perform CREATE INDEX CONCURRENTLY. (Another possibility is
346 to rebuild the index with REINDEX INDEX CONCURRENTLY).
347 Another caveat when building a unique index concurrently is that the
349 uniqueness constraint is already being enforced against other
350 transactions when the second table scan begins. This means that
351 constraint violations could be reported in other queries prior to the
352 index becoming available for use, or even in cases where the index
353 build eventually fails. Also, if a failure does occur in the second
354 scan, the “invalid” index continues to enforce its uniqueness
355 constraint afterwards.
356 Concurrent builds of expression indexes and partial indexes are
358 supported. Errors occurring in the evaluation of these expressions
359 could cause behavior similar to that described above for unique
360 constraint violations.
361 Regular index builds permit other regular index builds on the same
363 table to occur simultaneously, but only one concurrent index build can
364 occur on a table at a time. In either case, schema modification of the
365 table is not allowed while the index is being built. Another difference
366 is that a regular CREATE INDEX command can be performed within a
367 transaction block, but CREATE INDEX CONCURRENTLY cannot.
368 Concurrent builds for indexes on partitioned tables are currently not
370 supported. However, you may concurrently build the index on each
371 partition individually and then finally create the partitioned index
372 non-concurrently in order to reduce the time where writes to the
373 partitioned table will be locked out. In this case, building the
374 partitioned index is a metadata only operation.
375
377 See Chapter 11 for information about when indexes can be used, when
378 they are not used, and in which particular situations they can be
379 useful.
380 Currently, only the B-tree, GiST, GIN, and BRIN index methods support
382 multiple-key-column indexes. Whether there can be multiple key columns
383 is independent of whether INCLUDE columns can be added to the index.
384 Indexes can have up to 32 columns, including INCLUDE columns. (This
385 limit can be altered when building PostgreSQL.) Only B-tree currently
386 supports unique indexes.
387 An operator class with optional parameters can be specified for each
389 column of an index. The operator class identifies the operators to be
390 used by the index for that column. For example, a B-tree index on
391 four-byte integers would use the int4_ops class; this operator class
392 includes comparison functions for four-byte integers. In practice the
393 default operator class for the column's data type is usually
394 sufficient. The main point of having operator classes is that for some
395 data types, there could be more than one meaningful ordering. For
396 example, we might want to sort a complex-number data type either by
397 absolute value or by real part. We could do this by defining two
398 operator classes for the data type and then selecting the proper class
399 when creating an index. More information about operator classes is in
400 Section 11.10 and in Section 38.16.
401 When CREATE INDEX is invoked on a partitioned table, the default
403 behavior is to recurse to all partitions to ensure they all have
404 matching indexes. Each partition is first checked to determine whether
405 an equivalent index already exists, and if so, that index will become
406 attached as a partition index to the index being created, which will
407 become its parent index. If no matching index exists, a new index will
408 be created and automatically attached; the name of the new index in
409 each partition will be determined as if no index name had been
410 specified in the command. If the ONLY option is specified, no recursion
411 is done, and the index is marked invalid. (ALTER INDEX ... ATTACH
412 PARTITION marks the index valid, once all partitions acquire matching
413 indexes.) Note, however, that any partition that is created in the
414 future using CREATE TABLE ... PARTITION OF will automatically have a
415 matching index, regardless of whether ONLY is specified.
416 For index methods that support ordered scans (currently, only B-tree),
418 the optional clauses ASC, DESC, NULLS FIRST, and/or NULLS LAST can be
419 specified to modify the sort ordering of the index. Since an ordered
420 index can be scanned either forward or backward, it is not normally
421 useful to create a single-column DESC index — that sort ordering is
422 already available with a regular index. The value of these options is
423 that multicolumn indexes can be created that match the sort ordering
424 requested by a mixed-ordering query, such as SELECT ... ORDER BY x ASC,
425 y DESC. The NULLS options are useful if you need to support “nulls sort
426 low” behavior, rather than the default “nulls sort high”, in queries
427 that depend on indexes to avoid sorting steps.
428 The system regularly collects statistics on all of a table's columns.
430 Newly-created non-expression indexes can immediately use these
431 statistics to determine an index's usefulness. For new expression
432 indexes, it is necessary to run ANALYZE or wait for the autovacuum
433 daemon to analyze the table to generate statistics for these indexes.
434 For most index methods, the speed of creating an index is dependent on
436 the setting of maintenance_work_mem. Larger values will reduce the time
437 needed for index creation, so long as you don't make it larger than the
438 amount of memory really available, which would drive the machine into
439 swapping.
440 PostgreSQL can build indexes while leveraging multiple CPUs in order to
442 process the table rows faster. This feature is known as parallel index
443 build. For index methods that support building indexes in parallel
444 (currently, only B-tree), maintenance_work_mem specifies the maximum
445 amount of memory that can be used by each index build operation as a
446 whole, regardless of how many worker processes were started. Generally,
447 a cost model automatically determines how many worker processes should
448 be requested, if any.
449 Parallel index builds may benefit from increasing maintenance_work_mem
451 where an equivalent serial index build will see little or no benefit.
452 Note that maintenance_work_mem may influence the number of worker
453 processes requested, since parallel workers must have at least a 32MB
454 share of the total maintenance_work_mem budget. There must also be a
455 remaining 32MB share for the leader process. Increasing
456 max_parallel_maintenance_workers may allow more workers to be used,
457 which will reduce the time needed for index creation, so long as the
458 index build is not already I/O bound. Of course, there should also be
459 sufficient CPU capacity that would otherwise lie idle.
460 Setting a value for parallel_workers via ALTER TABLE directly controls
462 how many parallel worker processes will be requested by a CREATE INDEX
463 against the table. This bypasses the cost model completely, and
464 prevents maintenance_work_mem from affecting how many parallel workers
465 are requested. Setting parallel_workers to 0 via ALTER TABLE will
466 disable parallel index builds on the table in all cases.
467 Tip
469 You might want to reset parallel_workers after setting it as part
470 of tuning an index build. This avoids inadvertent changes to query
471 plans, since parallel_workers affects all parallel table scans.
472 While CREATE INDEX with the CONCURRENTLY option supports parallel
474 builds without special restrictions, only the first table scan is
475 actually performed in parallel.
476 Use DROP INDEX to remove an index.
478 Like any long-running transaction, CREATE INDEX on a table can affect
480 which tuples can be removed by concurrent VACUUM on any other table.
481 Prior releases of PostgreSQL also had an R-tree index method. This
483 method has been removed because it had no significant advantages over
484 the GiST method. If USING rtree is specified, CREATE INDEX will
485 interpret it as USING gist, to simplify conversion of old databases to
486 GiST.
487 Each backend running CREATE INDEX will report its progress in the
489 pg_stat_progress_create_index view. See Section 28.4.2 for details.
490
492 To create a unique B-tree index on the column title in the table films:
493 CREATE UNIQUE INDEX title_idx ON films (title);
495 To create a unique B-tree index on the column title with included
497 columns director and rating in the table films:
498 CREATE UNIQUE INDEX title_idx ON films (title) INCLUDE (director, rating);
500 To create a B-Tree index with deduplication disabled:
502 CREATE INDEX title_idx ON films (title) WITH (deduplicate_items = off);
504 To create an index on the expression lower(title), allowing efficient
506 case-insensitive searches:
507 CREATE INDEX ON films ((lower(title)));
509 (In this example we have chosen to omit the index name, so the system
511 will choose a name, typically films_lower_idx.)
512 To create an index with non-default collation:
514 CREATE INDEX title_idx_german ON films (title COLLATE "de_DE");
516 To create an index with non-default sort ordering of nulls:
518 CREATE INDEX title_idx_nulls_low ON films (title NULLS FIRST);
520 To create an index with non-default fill factor:
522 CREATE UNIQUE INDEX title_idx ON films (title) WITH (fillfactor = 70);
524 To create a GIN index with fast updates disabled:
526 CREATE INDEX gin_idx ON documents_table USING GIN (locations) WITH (fastupdate = off);
528 To create an index on the column code in the table films and have the
530 index reside in the tablespace indexspace:
531 CREATE INDEX code_idx ON films (code) TABLESPACE indexspace;
533 To create a GiST index on a point attribute so that we can efficiently
535 use box operators on the result of the conversion function:
536 CREATE INDEX pointloc
538 ON points USING gist (box(location,location));
539 SELECT * FROM points
540 WHERE box(location,location) && '(0,0),(1,1)'::box;
541 To create an index without locking out writes to the table:
543 CREATE INDEX CONCURRENTLY sales_quantity_index ON sales_table (quantity);
545
547 CREATE INDEX is a PostgreSQL language extension. There are no
548 provisions for indexes in the SQL standard.
549
551 ALTER INDEX (ALTER_INDEX(7)), DROP INDEX (DROP_INDEX(7)), REINDEX(7),
552 Section 28.4.2
553PostgreSQL 15.4 2023 CREATE INDEX(7)