Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
Comment: | Move the examples on lang_explain.html to new document eqp.html. |
---|---|
Downloads: | Tarball | ZIP archive |
Timelines: | family | ancestors | descendants | both | trunk |
Files: | files | file ages | folders |
SHA1: |
0c8678b577986758037af8ee13e3720f |
User & Date: | dan 2010-11-13 19:00:19.000 |
Context
2010-11-15
| ||
13:15 | Fix a typo from the fts3.html document. (check-in: c547f4b6c1 user: drh tags: trunk) | |
2010-11-13
| ||
19:00 | Move the examples on lang_explain.html to new document eqp.html. (check-in: 0c8678b577 user: dan tags: trunk) | |
17:10 | Update EXPLAIN QUERY PLAN examples to match changes to sqlite. (check-in: 3f8c0a0c1f user: dan tags: trunk) | |
Changes
Added pages/eqp.in.
> > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > | 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 | <tcl>hd_keywords {explain query plan}</tcl> <title>EXPLAIN QUERY PLAN</title> <style>codeblock { display:block;white-space:pre;font-family:fixed }</style> <table_of_contents> <h1>The EXPLAIN QUERY PLAN Command</h1> <p style="margin-left:10ex;margin-right:10ex"> <b>Warning: The data returned by the EXPLAIN QUERY PLAN command is intended for interactive debugging only. It may change dramatically between SQLite releases. Applications should not depend on the results of an EXPLAIN QUERY PLAN command.</b> <p>The [EXPLAIN|EXPLAIN QUERY PLAN] SQL command is used to obtain a high-level description of the strategy or plan that SQLite uses to implement a specific SQL query. Most significantly, it reports on the way in which the query uses database indices. In interpreting and using this information to optimize database schemas and queries, users might find the documents describing how SQLite [indexing|plans] and [optimizer|optimizes] queries useful. <p>An EXPLAIN QUERY PLAN command returns zero or more rows of four columns each. The column names are "selectid", "order", "from", "detail". Each of the first three column always contains an integer value. The final column, "detail", which contains most of the useful information, always contains a text value. <p>EXPLAIN QUERY PLAN is most useful when used with a SELECT statement, but may also be used on other statements that read data from database tables (e.g. UPDATE, DELETE, INSERT INTO ... SELECT). <h2>Table and Index Scans</h2> <p> When processing a SELECT (or other) statement, SQLite may retrieve data from database tables in a variety of ways. It may scan through all the records in a table (a full-table scan), scan a contiguous subset of the records in a table based on the rowid index, scan a contiguous subset of the entries in a database [CREATE TABLE|index], or use a combination of the above strategies in a single scan. The various ways in which SQLite may retrieve data from a table or index are described in detail [strategies|here]. <p> For each table the query reads data from, the output of EXPLAIN QUERY PLAN includes a record for which the value in the "detail" column begins with either "SCAN" or "SEARCH". "SCAN" is used for a full-table scan, including cases where SQLite iterates through all records in a table in an order defined by an index. "SEARCH" indicates that only a subset of the table rows are visited. Each SCAN or SEARCH record includes the following information: <ul> <li> The name of the table data is read from. <li> Whether or not an index or [automatic indexing|automatic index] is used. <li> Whether or not the [covering index] optimization applies. <li> The selectivity of the subset of records scanned. <li> The estimated number of rows that SQLite expects the scan to visit. </ul> <p> For example, the following EXPLAIN QUERY PLAN command operates on a SELECT statement that is implemented by performing a full-table scan on table t1: ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT a, b FROM t1 WHERE a=1; 0|0|0|SCAN TABLE t1 (~100000 rows) </codeblock>)^ <p> SQLite estimates that the full-table scan will visit approximately 1,000,000 records. If the query were able to use an index, then the SCAN/SEARCH record would include the name of the index and, for a SEARCH record, an indication of how the subset of rows visited is identified. For example: ^(<codeblock> sqlite> CREATE INDEX i1 ON t1(a); sqlite> EXPLAIN QUERY PLAN SELECT a, b FROM t1 WHERE a=1; 0|0|0|SEARCH TABLE t1 USING INDEX i1 (a=?) (~10 rows) </codeblock>)^ <p> The output above shows that in this case, SQLite uses index "i1" to optimize a WHERE clause filter of the form (a=?) - in this case "a=1". SQLite estimates that scanning the subset of index entries that match the "a=1" filter means scanning through approximately 10 records. In this case it is not possible to use index i1 as a [covering index], but if it were, the SCAN or SEARCH record would report that as well. For example: ^(<codeblock> sqlite> CREATE INDEX i2 ON t1(a, b); sqlite> EXPLAIN QUERY PLAN SELECT a, b FROM t1 WHERE a=1; 0|0|0|SEARCH TABLE t1 USING COVERING INDEX i2 (a=?) (~10 rows) </codeblock>)^ <p> All joins in SQLite are [join order|implemented using nested scans]. When a SELECT query that features a join is analyzed using EXPLAIN QUERY PLAN, one SCAN or SEARCH record is output for each nested loop. For example: ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT t1.*, t2.* FROM t1, t2 WHERE t1.a=1 AND t1.b>2; 0|0|0|SEARCH TABLE t1 USING COVERING INDEX i2 (a=? AND b>?) (~3 rows) 0|1|1|SCAN TABLE t2 (~1000000 rows) </codeblock>)^ <p> The second column of output (column "order") indicates the nesting order. In this case, the scan of table t1 using index i2 is the outer loop (order=0) and the full-table scan of table t2 (order=1) is the inner loop. The third column (column "from"), indicates the position in the FROM clause of the SELECT statement that the table associated with each scan occurs in. In the case above, table t1 occupies the first position in the FROM clause, so the value of column "from" is 0 in the first record. Table t2 is in the second position, so the "from" column for the corresponding SCAN record is set to 1. In the following example, the positions of t1 and t2 in the FROM clause of the SELECT are reversed. The query strategy remains the same, but the values in the "from" column of the output are adjusted accordingly. ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT t1.*, t2.* FROM t2, t1 WHERE t1.a=1 AND t1.b>2; 0|0|1|SEARCH TABLE t1 USING COVERING INDEX i2 (a=? AND b>?) (~3 rows) 0|1|0|SCAN TABLE t2 (~1000000 rows) </codeblock>)^ <p> In the example above, SQLite estimates that the outer loop scan will visit approximately 3 rows, and the inner loop scan approximately 1,000,000. If you observe that SQLite's estimates are wildly inaccurate (and appear to be causing it to generate sub-optimal query plans), your queries may benefit from running the [ANALYZE] command on the database. <p> If the WHERE clause of a query contains an OR expression, then SQLite might use the [or-connected-terms|"OR by union"] strategy (also described [or optimization|here]). In this case there will be two SEARCH records, one for each index, with the same values in both the "order" and "from" columns. For example: ^(<codeblock> sqlite> CREATE INDEX i3 ON t1(b); sqlite> EXPLAIN QUERY PLAN SELECT * FROM t1 WHERE a=1 OR b=2; 0|0|0|SEARCH TABLE t1 USING COVERING INDEX i2 (a=?) (~10 rows) 0|0|0|SEARCH TABLE t1 USING INDEX i3 (b=?) (~10 rows) </codeblock>)^ <h2>Temporary Sorting B-Trees</h2> <p> If a SELECT query contains an ORDER BY, GROUP BY or DISTINCT clause, SQLite may need to use a temporary b-tree structure to perform an <a href="http://en.wikipedia.org/wiki/Insertion_sort">insertion sort</a> of the output rows. Or, it may [sorting|use an index]. Using an index is almost always much more efficient than performing an online insertion sort. If a temporary b-tree is required, a record is added to the EXPLAIN QUERY PLAN output with the "detail" field set to a string value of the form "USE TEMP B-TREE FOR xxx", where xxx is one of "ORDER BY", "GROUP BY" or "DISTINCT". For example: ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT c, d FROM t2 ORDER BY c; 0|0|0|SCAN TABLE t2 (~1000000 rows) 0|0|0|USE TEMP B-TREE FOR ORDER BY </codeblock>)^ <p> In this case using the temporary b-tree can be avoided by creating an index on t2(c), as follows: ^(<codeblock> sqlite> CREATE INDEX i4 ON t2(c); sqlite> EXPLAIN QUERY PLAN SELECT c, d FROM t2 ORDER BY c; 0|0|0|SCAN TABLE t2 USING INDEX i4 (~1000000 rows) </codeblock>)^ <h2>Subqueries</h2> <p> In all the examples above, the first column (column "selectid") is always set to 0. If a query contains sub-selects, either as part of the FROM clause or as part of SQL expressions, then the output of EXPLAIN QUERY PLAN also includes a report for each sub-select. Each sub-select is assigned a distinct, non-zero "selectid" value. The top-level SELECT statement is always assigned the selectid value 0. For example: ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT (SELECT b FROM t1 WHERE a=0), (SELECT a FROM t1 WHERE b=t2.c) FROM t2; 0|0|0|SCAN TABLE t2 (~1000000 rows) 0|0|0|EXECUTE SCALAR SUBQUERY 1 1|0|0|SEARCH TABLE t1 USING COVERING INDEX i2 (a=?) (~10 rows) 0|0|0|EXECUTE CORRELATED SCALAR SUBQUERY 2 2|0|0|SEARCH TABLE t1 USING INDEX i3 (b=?) (~10 rows) </codeblock>)^ <p> The example above contains a pair of scalar subqueries assigned selectid values 1 and 2. As well as a SCAN record, there are also 2 "EXECUTE" records associated with the top level subquery (selectid 0), indicating that subqueries 1 and 2 are executed by the top level query in a scalar context. The CORRELATED qualifier present in the EXECUTE record associated with scalar subquery 2 indicates that the query must be run separately for each row visited by the top level query. Its absence in the record associated with subquery 1 means that the subquery is only run once and the result cached. In other words, subquery 2 may be more performance critical, as it may be run many times whereas subquery 1 is only ever run once. <p> Unless the [flattening optimization] is applied, if a subquery appears in the FROM clause of a SELECT statement, SQLite executes the subquery and stores the results in a temporary table. It then uses the contents of the temporary table in place of the subquery to execute the parent query. This is shown in the output of EXPLAIN QUERY PLAN by substituting a "SCAN SUBQUERY" record for the "SCAN TABLE" record that normally appears for each element in the FROM clause. For example: ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT count(*) FROM (SELECT max(b) AS x FROM t1 GROUP BY a) GROUP BY x; 1|0|0|SCAN TABLE t1 USING COVERING INDEX i2 (~1000000 rows) 0|0|0|SCAN SUBQUERY 1 (~1000000 rows) 0|0|0|USE TEMP B-TREE FOR GROUP BY </codeblock>)^ <p> If the [flattening optimization] is used on a subquery in the FROM clause of a SELECT statement, then the output of EXPLAIN QUERY PLAN reflects this. For example, in the following there is no "SCAN SUBQUERY" record even though there is a subquery in the FROM clause of the top level SELECT. Instead, since the flattening optimization does apply in this case, the EXPLAIN QUERY PLAN report shows that the top level query is implemented using a nested loop join of tables t1 and t2. ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT * FROM (SELECT * FROM t2 WHERE c=1), t1; 0|0|0|SEARCH TABLE t2 USING INDEX i4 (c=?) (~10 rows) 0|1|1|SCAN TABLE t1 (~1000000 rows) </codeblock>)^ <h2>Compound Queries</h2> <p> Each component query of a [compound query] (UNION, UNION ALL, EXCEPT or INTERSECT) is assigned its own selectid and reported on separately. A single record is output for the parent (compound query) identifying the operation, and whether or not a temporary b-tree is used to implement it. For example: ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT a FROM t1 UNION SELECT c FROM t2; 1|0|0|SCAN TABLE t1 (~1000000 rows) 2|0|0|SCAN TABLE t2 (~1000000 rows) 0|0|0|COMPOUND SUBQUERIES 1 AND 2 USING TEMP B-TREE (UNION) </codeblock>)^ <p> The "USING TEMP B-TREE" clause in the above output indicates that a temporary b-tree structure is used to implement the UNION of the results of the two sub-selects. If the temporary b-tree were not required, as in the following example, the clause is not present. ^(<codeblock> sqlite> EXPLAIN QUERY PLAN SELECT a FROM t1 EXCEPT SELECT d FROM t2 ORDER BY 1; 1|0|0|SCAN TABLE t1 USING COVERING INDEX i2 (~1000000 rows) 2|0|0|SCAN TABLE t2 (~1000000 rows) 2|0|0|USE TEMP B-TREE FOR ORDER BY 0|0|0|COMPOUND SUBQUERIES 1 AND 2 (EXCEPT) </codeblock>)^ <h1>Sample Code</h1> <codeblock> /* ** Argument pStmt is a prepared SQL statement. This function compiles ** an EXPLAIN QUERY PLAN command to report on the prepared statement, ** and prints the report to stdout using printf(). */ int printExplainQueryPlan(sqlite3_stmt *pStmt){ char *zSql; /* Input SQL */ char *zExplain; /* SQL with EXPLAIN QUERY PLAN prepended */ sqlite3_stmt *pExplain; /* Compiled EXPLAIN QUERY PLAN command */ int rc; /* Return code from sqlite3_prepare_v2() */ zSql = sqlite3_sql(pStmt); if( zSql==0 ) return SQLITE_ERROR; zExplain = sqlite3_mprintf("EXPLAIN QUERY PLAN %s", zSql); if( zExplain==0 ) return SQLITE_NOMEM; rc = sqlite3_prepare_v2(sqlite3_db_handle(pStmt), zExplain, -1, &pExplain, 0); sqlite3_free(zExplain); if( rc!=SQLITE_OK ) return rc; while( SQLITE_ROW==sqlite3_step(pExplain) ){ int iSelectid = sqlite3_column_int(pExplain, 0); int iOrder = sqlite3_column_int(pExplain, 1); int iFrom = sqlite3_column_int(pExplain, 2); const char *zDetail = sqlite3_column_text(pExplain, 3); printf("%d %d %d %s\n", iSelectid, iOrder, iFrom, zDetail); } return sqlite3_finalize(pExplain); } </codeblock> |
Changes to pages/fancyformat.tcl.
︙ | ︙ | |||
412 413 414 415 416 417 418 | # this text. The "<div class=startsearch>" tag tells the script that # builds the site-search database not to index any text that occurs # before it. This stops the table of contents from being used for # snippets on search results pages. # set toc [subst { <div class=fancy> | | | | 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 | # this text. The "<div class=startsearch>" tag tells the script that # builds the site-search database not to index any text that occurs # before it. This stops the table of contents from being used for # snippets on search results pages. # set toc [subst { <div class=fancy> <div style="font-size:2em;text-align:center;color:#044a64"> $::Addtoc(title) </div> <div style="font-size:1.5em;margin:1em;color:#044a64"> Table Of Contents</div> <div id=toc> $::Addtoc(toc) </div> <div class=startsearch></div> }] string map [list <table_of_contents> $toc] $::Addtoc(doc) } |
︙ | ︙ |
Changes to pages/lang.in.
︙ | ︙ | |||
1367 1368 1369 1370 1371 1372 1373 | <p>^When the EXPLAIN keyword appears by itself it causes the statement to behave as a query that returns the sequence of [virtual machine instructions] it would have used to execute the command had the EXPLAIN keyword not been present. ^When the EXPLAIN QUERY PLAN phrase appears, the statement returns high-level information regarding the query plan that would have been used. | | | < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < | 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 | <p>^When the EXPLAIN keyword appears by itself it causes the statement to behave as a query that returns the sequence of [virtual machine instructions] it would have used to execute the command had the EXPLAIN keyword not been present. ^When the EXPLAIN QUERY PLAN phrase appears, the statement returns high-level information regarding the query plan that would have been used. The EXPLAIN QUERY PLAN command is described in [explain query plan|more detail here]. <tcl> ############################################################################## Section expression expr {*expression {expression syntax}} BubbleDiagram expr 1 BubbleDiagram literal-value |
︙ | ︙ |