2 11.8. Partial Indexes #
4 A partial index is an index built over a subset of a table; the subset
5 is defined by a conditional expression (called the predicate of the
6 partial index). The index contains entries only for those table rows
7 that satisfy the predicate. Partial indexes are a specialized feature,
8 but there are several situations in which they are useful.
10 One major reason for using a partial index is to avoid indexing common
11 values. Since a query searching for a common value (one that accounts
12 for more than a few percent of all the table rows) will not use the
13 index anyway, there is no point in keeping those rows in the index at
14 all. This reduces the size of the index, which will speed up those
15 queries that do use the index. It will also speed up many table update
16 operations because the index does not need to be updated in all cases.
17 Example 11.1 shows a possible application of this idea.
19 Example 11.1. Setting up a Partial Index to Exclude Common Values
21 Suppose you are storing web server access logs in a database. Most
22 accesses originate from the IP address range of your organization but
23 some are from elsewhere (say, employees on dial-up connections). If
24 your searches by IP are primarily for outside accesses, you probably do
25 not need to index the IP range that corresponds to your organization's
28 Assume a table like this:
29 CREATE TABLE access_log (
35 To create a partial index that suits our example, use a command such as
37 CREATE INDEX access_log_client_ip_ix ON access_log (client_ip)
38 WHERE NOT (client_ip > inet '192.168.100.0' AND
39 client_ip < inet '192.168.100.255');
41 A typical query that can use this index would be:
44 WHERE url = '/index.html' AND client_ip = inet '212.78.10.32';
46 Here the query's IP address is covered by the partial index. The
47 following query cannot use the partial index, as it uses an IP address
48 that is excluded from the index:
51 WHERE url = '/index.html' AND client_ip = inet '192.168.100.23';
53 Observe that this kind of partial index requires that the common values
54 be predetermined, so such partial indexes are best used for data
55 distributions that do not change. Such indexes can be recreated
56 occasionally to adjust for new data distributions, but this adds
59 Another possible use for a partial index is to exclude values from the
60 index that the typical query workload is not interested in; this is
61 shown in Example 11.2. This results in the same advantages as listed
62 above, but it prevents the “uninteresting” values from being accessed
63 via that index, even if an index scan might be profitable in that case.
64 Obviously, setting up partial indexes for this kind of scenario will
65 require a lot of care and experimentation.
67 Example 11.2. Setting up a Partial Index to Exclude Uninteresting
70 If you have a table that contains both billed and unbilled orders,
71 where the unbilled orders take up a small fraction of the total table
72 and yet those are the most-accessed rows, you can improve performance
73 by creating an index on just the unbilled rows. The command to create
74 the index would look like this:
75 CREATE INDEX orders_unbilled_index ON orders (order_nr)
76 WHERE billed is not true;
78 A possible query to use this index would be:
79 SELECT * FROM orders WHERE billed is not true AND order_nr < 10000;
81 However, the index can also be used in queries that do not involve
82 order_nr at all, e.g.:
83 SELECT * FROM orders WHERE billed is not true AND amount > 5000.00;
85 This is not as efficient as a partial index on the amount column would
86 be, since the system has to scan the entire index. Yet, if there are
87 relatively few unbilled orders, using this partial index just to find
88 the unbilled orders could be a win.
90 Note that this query cannot use this index:
91 SELECT * FROM orders WHERE order_nr = 3501;
93 The order 3501 might be among the billed or unbilled orders.
95 Example 11.2 also illustrates that the indexed column and the column
96 used in the predicate do not need to match. PostgreSQL supports partial
97 indexes with arbitrary predicates, so long as only columns of the table
98 being indexed are involved. However, keep in mind that the predicate
99 must match the conditions used in the queries that are supposed to
100 benefit from the index. To be precise, a partial index can be used in a
101 query only if the system can recognize that the WHERE condition of the
102 query mathematically implies the predicate of the index. PostgreSQL
103 does not have a sophisticated theorem prover that can recognize
104 mathematically equivalent expressions that are written in different
105 forms. (Not only is such a general theorem prover extremely difficult
106 to create, it would probably be too slow to be of any real use.) The
107 system can recognize simple inequality implications, for example “x <
108 1” implies “x < 2”; otherwise the predicate condition must exactly
109 match part of the query's WHERE condition or the index will not be
110 recognized as usable. Matching takes place at query planning time, not
111 at run time. As a result, parameterized query clauses do not work with
112 a partial index. For example a prepared query with a parameter might
113 specify “x < ?” which will never imply “x < 2” for all possible values
116 A third possible use for partial indexes does not require the index to
117 be used in queries at all. The idea here is to create a unique index
118 over a subset of a table, as in Example 11.3. This enforces uniqueness
119 among the rows that satisfy the index predicate, without constraining
122 Example 11.3. Setting up a Partial Unique Index
124 Suppose that we have a table describing test outcomes. We wish to
125 ensure that there is only one “successful” entry for a given subject
126 and target combination, but there might be any number of “unsuccessful”
127 entries. Here is one way to do it:
135 CREATE UNIQUE INDEX tests_success_constraint ON tests (subject, target)
138 This is a particularly efficient approach when there are few successful
139 tests and many unsuccessful ones. It is also possible to allow only one
140 null in a column by creating a unique partial index with an IS NULL
143 Finally, a partial index can also be used to override the system's
144 query plan choices. Also, data sets with peculiar distributions might
145 cause the system to use an index when it really should not. In that
146 case the index can be set up so that it is not available for the
147 offending query. Normally, PostgreSQL makes reasonable choices about
148 index usage (e.g., it avoids them when retrieving common values, so the
149 earlier example really only saves index size, it is not required to
150 avoid index usage), and grossly incorrect plan choices are cause for a
153 Keep in mind that setting up a partial index indicates that you know at
154 least as much as the query planner knows, in particular you know when
155 an index might be profitable. Forming this knowledge requires
156 experience and understanding of how indexes in PostgreSQL work. In most
157 cases, the advantage of a partial index over a regular index will be
158 minimal. There are cases where they are quite counterproductive, as in
161 Example 11.4. Do Not Use Partial Indexes as a Substitute for
164 You might be tempted to create a large set of non-overlapping partial
166 CREATE INDEX mytable_cat_1 ON mytable (data) WHERE category = 1;
167 CREATE INDEX mytable_cat_2 ON mytable (data) WHERE category = 2;
168 CREATE INDEX mytable_cat_3 ON mytable (data) WHERE category = 3;
170 CREATE INDEX mytable_cat_N ON mytable (data) WHERE category = N;
172 This is a bad idea! Almost certainly, you'll be better off with a
173 single non-partial index, declared like
174 CREATE INDEX mytable_cat_data ON mytable (category, data);
176 (Put the category column first, for the reasons described in
177 Section 11.3.) While a search in this larger index might have to
178 descend through a couple more tree levels than a search in a smaller
179 index, that's almost certainly going to be cheaper than the planner
180 effort needed to select the appropriate one of the partial indexes. The
181 core of the problem is that the system does not understand the
182 relationship among the partial indexes, and will laboriously test each
183 one to see if it's applicable to the current query.
185 If your table is large enough that a single index really is a bad idea,
186 you should look into using partitioning instead (see Section 5.12).
187 With that mechanism, the system does understand that the tables and
188 indexes are non-overlapping, so far better performance is possible.
190 More information about partial indexes can be found in [ston89b],
191 [olson93], and [seshadri95].