database/sqlite
1
0
Fork 0
mirror of https://github.com/sqlite/sqlite.git synced 2025-11-18 10:21:03 +03:00
Code Activity
23,964 Commits 1,610 Branches 185 Tags
94e615a6acf3a1519d98e85a6665c712edb5ea8a
Commit Graph

1577 Commits

Author SHA1 Message Date
drh
4dd96a8315 Honor a high likelihood() on range constraints. 
FossilOrigin-Name: 401235edf40fcd665eaf426cf5155ac6855e8537
 2014-10-24 15:26:29 +00:00
drh
9769efcc4c Get the likelihood() functions working on operators like BETWEEN that 
create virtual terms in the WHERE-clause analysis.

FossilOrigin-Name: 03d0498d0f24bec2383d5d79edf25069effecd59
 2014-10-24 14:32:21 +00:00
drh
d7d7147022 Change the 0x800 bit of SQLITE_TESTCTRL_OPTIMIZATIONS so that it disables 
the loading of STAT3 and STAT4 content, not just the using of that content.
Change the internal name of that bit to SQLITE_Stat34.

FossilOrigin-Name: ca3b00c44ec52d209642a5ba9ef82e085fac39db
 2014-10-22 19:57:16 +00:00
drh
9f07cf7b2e Take steps to avoid misestimating range query costs based on STAT4 data 
due to the roundoff error of converting from integers to LogEst and back
to integers.

FossilOrigin-Name: 3c933bf95f291f7957580d823dce92c981375a5c
 2014-10-22 15:27:05 +00:00
drh
f7f2e84a9c Further tuning of the cost estimates for skip-scan loops, especially for cases 
when skip-scan loops are in competition with regular loops.

FossilOrigin-Name: a27861c28c4791e51d797aa37e9cca806cb58775
 2014-10-21 18:16:21 +00:00
drh
1b131b7a7a Improvements to the WHERETRACE debugging logic. 
FossilOrigin-Name: ec1e942f08548695ff02645b3f3cd6bb2516bc9a
 2014-10-21 16:01:40 +00:00
drh
c8bbce1e6a If a skip-scan is a proper subset of some other scan, then adjust the 
cost of the skip-scan upward so that it is more costly than the other scan.
Such a cost imbalance can arise under STAT4 because of difficulties in getting
an accurate estimate for skip-scans.

FossilOrigin-Name: f4b22a2620a5dc48949048c2ecbd226755d4b2c3
 2014-10-21 01:05:09 +00:00
drh
4929047d1a Make sure that a GROUP BY that also implements an ORDER BY scans the table 
in the correct order.  Fix for ticket [ba7cbfaedc7e6].

FossilOrigin-Name: 7a32fdfd4be2138c0ab00f3dc6f54a70e4e07be4
 2014-10-11 02:12:58 +00:00
drh
c631faa922 Remove an unnecessary conditional from the EXPLAIN QUERY PLAN logic. 
FossilOrigin-Name: c5dc83ebded914f07286b7f98d0a50c28c16f609
 2014-10-11 01:22:16 +00:00
dan
b6453201d2 Account for the ASC/DESC properties of ORDER BY expressions when using the same index for GROUP BY and ORDER BY. Candidate fix for [ba7cbfaedc]. 
FossilOrigin-Name: 2a9573962b837973c4959465d8a5f2641d109a5a
 2014-10-10 20:52:53 +00:00
drh
1f8817cb32 Restrict the scope of the explainIndexRange() function in where.c. 
FossilOrigin-Name: c30124520027f0f860223bf842e2f09db3dafb5f
 2014-10-10 19:15:35 +00:00
drh
98545bb2ed Add the estimated number of output rows to the EXPLAIN QUERY PLAN output 
if compiled with SQLITE_EXPLAIN_ESTIMATED_ROWS.  This feature is off by
default for the time being.

FossilOrigin-Name: daa8314fba9dc3c4f5e7fbda42c97604fbfc4392
 2014-10-10 17:20:39 +00:00
drh
6c97789d2e Simplify the code used to generate the text for EXPLAIN QUERY PLAN. 
FossilOrigin-Name: beea1efc3a49cad08087fcbb18dbce71c873fe57
 2014-10-10 15:47:46 +00:00
drh
681fca0018 Fix a potential problem in the whereRangeScanEst() routine when STAT4 is 
active.  The problem was introduced by recent enhancements.

FossilOrigin-Name: 68e1b4de700b5291f79249a03e1a750c6b2c9ae4
 2014-10-10 15:01:46 +00:00
drh
8e0a8f681a Remove an always-true branch in whereRangeScanEst(). Replace it with an 
assert().

FossilOrigin-Name: 42e48fd3a6a6219d9bd6135d821b38c5157922ba
 2014-10-08 19:33:54 +00:00
drh
69afd9980e Fix the STAT4 range scan estimates for DESC indexes. 
FossilOrigin-Name: e3fe84005259ef9a6027d25793514cebb2d4e7e0
 2014-10-08 02:53:25 +00:00
drh
9a7b41d74a More intuitive labels on ".wheretrace" output. 
FossilOrigin-Name: adcb3fed489b580221c7bf2692a60e24248b23a0
 2014-10-08 00:08:08 +00:00
drh
36be4c49e4 Enable the query planner to deal with WHERE clauses that have OR terms 
nested within AND terms that are nested within OR terms.  Also remove an
unused function declaration.

FossilOrigin-Name: b6b289182f6590288ebc7b9efbcb29b6b4480538
 2014-09-30 17:31:23 +00:00
drh
0a99ba3bc8 Further enhancements to the "wheretrace" debugging output. 
FossilOrigin-Name: 670993eb8113f386cb2cb8b1507917f6da3b4d98
 2014-09-30 17:03:35 +00:00
drh
5265149c0d Enhanced debug output for OR-logic in the query loop optimizer. 
FossilOrigin-Name: 2e375eae473e4a9f2e7870d59e22ba39051ecbce
 2014-09-30 14:14:19 +00:00
drh
c90713d3d2 Show tree diagrams of data structures in the debugging output when the 0x100 
bit is set on sqlite3WhereTrace or sqlite3SelectTrace.

FossilOrigin-Name: 92e0b4bd4d75e8b000586e51a07b3e181d9af20b
 2014-09-30 13:46:49 +00:00
drh
4fa4a54f7e Remove the SQLITE_ENABLE_TREE_EXPLAIN compile-time option. Add alternative 
debugging display routines: sqlite3TreeViewExpr(), sqlite3TreeViewExprList(),
and sqlite3TreeViewSelect().

FossilOrigin-Name: 4ff51325d6b41d0c59e303b573700ec80c51d216
 2014-09-30 12:33:33 +00:00
drh
14f0e2128a Adjust skip-scan cost estimates slightly so that a full table scan is 
preferred over a skip-scan to a column with only two distinct values.

FossilOrigin-Name: ae9a42b268ad3f7d21a5813bb931e795c6917014
 2014-09-23 01:40:59 +00:00
drh
7e07433fe7 Tune the query planner to be more aggressive about using automatic indexes 
on views and subqueries for which there is not opportunity to declare a
persistent schema index.

FossilOrigin-Name: 41de1643bfc9ae25e20790d707e2789b665baa2b
 2014-09-22 14:30:51 +00:00
drh
137fd4fda2 Make sure that the sorting-index pre-filter recognizes that a rowid reference 
might be sortable.  This fixes a performance regression.

FossilOrigin-Name: 72727b68cd07969165f1f0943cc7e1a265436653
 2014-09-19 02:01:37 +00:00
drh
75179ded97 Reduce the number of arguments to RecordCompare functions from 4 to 3, 
resulting in a small performance increase.

FossilOrigin-Name: 8239c35aedd583af79505378bb7dbb78346a3f45
 2014-09-16 14:37:35 +00:00
peter.d.reid
60ec914c74 Fix typos in comments. No code changes. 
FossilOrigin-Name: e62aab5e9290503869e1f4d5e0fefd2b4dee0a69
 2014-09-06 16:39:46 +00:00
mistachkin
6b9da12856 Fix a couple of typos in comments. No changes to code. 
FossilOrigin-Name: a758465e3cfa7e0cb8749d097cd6fb5f86b60955
 2014-09-06 02:00:41 +00:00
drh
d8b77e20fc Query planner heuristic update: 
When doing a full table scan on a table that has an equality constraint on
an unindexed column, do not allow the estimated number of output rows to
be greater than half the total number of rows in the table.

FossilOrigin-Name: 73954f93c4c6f880c6e01d0d130e3fed40fd4106
 2014-09-06 01:35:57 +00:00
drh
ae914d783a Improved WHERETRACE messages for the estimated output row reductions from 
range scans.

FossilOrigin-Name: fdd478bb11eb9e244a7a1fb628ac27c53830fb24
 2014-08-28 19:38:22 +00:00
drh
b34fc5be6d Add a testcase() macro to help verfity the row estimation logic. 
FossilOrigin-Name: 6ba609522ea7325341fad93bdb5f869c6506dea4
 2014-08-28 17:20:37 +00:00
dan
7fa0631ece Fix a problem causing an inaccurate stat4-based estimate for the number of rows visited by a range scan. 
FossilOrigin-Name: 052d89b554a947b7e420b7feb6650f5069b7cefa
 2014-08-28 16:01:31 +00:00
drh
5418b1298f Fix a formatting error ("%d" needed in place of "%g") on a WHERETRACE 
macro inside of the query planner.  This fix applies to debugging logic
only.

FossilOrigin-Name: c931ca2b7763cf894ece26ca4775d638876107e9
 2014-08-28 13:42:13 +00:00
dan
0829169f1e When determining whether or not a partial index is usable, do not assume that the cursor number assigned to each table in the query is the same as its index in the FROM clause. Fix for ticket [98d973b8f5]. 
FossilOrigin-Name: fcebca166f15431764b82a8b267f11d28386e975
 2014-08-27 17:37:20 +00:00
dan
f741e0491e In cases where stat4 data is available but cannot be used because the rhs of a range constraint is too complex a expression, fall back to using the default estimates for number of rows scanned. 
FossilOrigin-Name: e06dc6f0c35f87c44292c71677111b74f073a5c4
 2014-08-25 18:29:38 +00:00
drh
f3f69ac942 Enhancements to skip-scan such that it is operable when a middle column of 
an index is skipped while the left-most column is constrained in the WHERE
clause.

FossilOrigin-Name: bc985caa7816f1f873ad8e4467c5278399f315ce
 2014-08-20 23:38:07 +00:00
dan
e2c278513d Fix a buffer overrun in the previous commit. 
FossilOrigin-Name: 43c59c85436dc8001c81f4aac7f5231b13d741cb
 2014-08-08 17:25:33 +00:00
dan
50ae31e6f6 Because SQLite internally calculates query plan costs using a logarithmic scale, very large estimated sorting costs can cause all other estimated costs to be rounded down to zero. In these cases break ties between plans with the same total cost by comparing the costs with sorting excluded. This is an alternative fix for the same problem as addressed by [2af630c572]. 
FossilOrigin-Name: 299b9570279ded7158d22349ef93384286a5c755
 2014-08-08 16:52:28 +00:00
drh
f2a90306ca Clarify the computation of compatible isOrdered by in the plan solver of 
the query planner.

FossilOrigin-Name: b5e8fd575a80334160de0aac8084ed5cd28816a5
 2014-08-07 20:37:01 +00:00
drh
b28ce65f45 Remove the extraneous debugging printf() from the previous check-in. 
FossilOrigin-Name: 8f04d2c0084afa6381e78847c9aa296498d448cb
 2014-08-07 20:25:37 +00:00
drh
ddef5dc044 When the estimated cost to do a sort overwhelms the estimated cost to do 
individual table lookups, make sure that the table lookup costs are still
taken into consideration when selecting the lookup algorithm.

FossilOrigin-Name: ec5d84ba69c100d9565425ed74040a49e410ea03
 2014-08-07 16:50:00 +00:00
drh
3a85625d87 Begin making changes to the IN operator in an attempt to make it run faster 
and to make the code easier to understand.

FossilOrigin-Name: ee0fd6aaf94cda1dce3fe752bfe3b0f83e0043f1
 2014-08-01 14:46:57 +00:00
drh
5f1d1d9c87 Refactoring: Change "pIndex->onError!=OE_None" to use a macro: 
"IsUniqueIndex(pIndex)".  Easier to understand that way.

FossilOrigin-Name: e75b26ee357bb3d3c1a539b05d633ebf314726d7
 2014-07-31 22:59:04 +00:00
drh
d15f87ecfd Fix a bug in the whereRangeSkipScanEst() procedure (added by 
check-in [d09ca6d5efad3e4cfa]) where it fails to consider the possibility
of a ROWID column when computing the affinity of a table column.

FossilOrigin-Name: 6aea2258dd7d3967900cefe4cd0af48a7ee7e169
 2014-07-24 22:41:20 +00:00
drh
dbd9486d5b Add experimental "costmult" logic. Only enabled when compiled with 
-DSQLITE_ENABLE_COSTMULT.

FossilOrigin-Name: 729ece40885ed7f52c5981364833fc62281a388b
 2014-07-23 23:57:42 +00:00
drh
3526319b48 Add the OP_ReopenIdx opcode that works like OP_OpenRead except that it becomes 
a no-op if the cursor is already open on the same index.  Update the
OR-optimization logic to make use of OP_ReopenIdx in order to avoid
unnecessary cursor open requests sent to the B-Tree layer.

FossilOrigin-Name: 77f412caf0192d3e7fecb377d6d72123d8b64424
 2014-07-22 20:02:19 +00:00
drh
a45fdc7b1e The optimization of check-in [b67a6e33f2] does not work (it generates 
incorrect VDBE code) if an OR term is AND-ed with a constant expression.
So back that optimization out and add a test case to make sure it does not
get added back in.

FossilOrigin-Name: eed754fe93644f4df082eac0c0a7ffc5c78ccd10
 2014-07-22 19:14:42 +00:00
drh
358406fc57 Expire prepared statements after running ANALYZE. 
FossilOrigin-Name: b083a961f8ac3a6158e822574d73cd63e3103b09
 2014-07-22 14:42:16 +00:00
drh
83a305f20e Enhance the comment on whereLoopAddBtree(). No changes to code. 
FossilOrigin-Name: b22dd165da227a52d88b17a91e80a1701dce61ad
 2014-07-22 12:05:32 +00:00
drh
53ad7e48ab For the OR-optimization, avoid generating OP_OpenRead opcodes that reopen 
exactly the same index.

FossilOrigin-Name: b67a6e33f23ce5f5d9a545fa9d6700a7ed636901
 2014-07-22 00:40:45 +00:00