1
0
mirror of https://github.com/sqlite/sqlite.git synced 2025-11-18 10:21:03 +03:00
Commit Graph

1372 Commits

Author SHA1 Message Date
drh
559656196b Simplify the sqlite3VdbeGet() routine. Smaller, faster, and easier to maintain.
FossilOrigin-Name: 5dbb255a9377a6c2619fbac519f18caa4d8cb23257dfba0ffb9e36dd9dc16627
2017-09-16 20:58:41 +00:00
drh
2def2f7e0a Remove an unnecessary parameter from selectInnerLoop().
FossilOrigin-Name: dd95887f82739d05585edfb70913fdaaab154de46c4d3113dd32d283767fb6fc
2017-09-15 17:40:34 +00:00
drh
4dd89d5a24 Sometimes a TK_COLUMN Expr node can have Expr.pTab==0 if it is a reference
to an expression column in an index on and expression.  Fix for ticket
[aa98619ad08ddcab].

FossilOrigin-Name: d0da791ba0edfb65186459345e43500d8364a086a5a1651d828fecc1a1dd1edb
2017-08-14 14:53:24 +00:00
drh
1112cc797b Properly dequote column names in tables constructed by an aggregate SELECT.
FossilOrigin-Name: 7e0d3e9cb071873564b7916c022aba5cd9f3b8ebab9dba787ecd7113c5b7816a
2017-08-14 01:33:07 +00:00
drh
a485ad191f In the KeyInfo object, refactor the nField and nXField elements into
nKeyField and nAllField, which are more useful and run a little faster.

FossilOrigin-Name: aea5990eab5e85f92df966aa641db2271c81052010ad2d80982475c4275a1284
2017-08-02 22:43:14 +00:00
drh
f35f2f92ed Move the generation of output column names earlier, to right after name
resolution and before query transformations such as flattening.  This prevents
the names from getting mangled by query transformations, and obviates hacks
in the query flattener that attempt to work around the name mangling.
The resulting code is smaller and faster and gives more consistent output.
This is an alternative fix to ticket [de3403bf5ae5f72ed].

FossilOrigin-Name: 09834279aeca3bda63de684a369ed64f2cbf587b5f5df1454c0a3c009a1337ad
2017-07-29 16:01:55 +00:00
dan
46133d98a6 Do not use the flattening optimization for a sub-query on the RHS of a LEFT
JOIN if that subquery reads data from a virtual table. Because it prevents the
planner from creating an automatic index on the results of the sub-query, the
flattening optimization sometimes slows things down in this case.

FossilOrigin-Name: 9dbae1df75219e2accd8993015ce0ffc21966d7e69d04178b1a2cc46207fe260
2017-07-18 17:34:41 +00:00
drh
ec360a8d01 Clarification and simplification and result column naming logic.
FossilOrigin-Name: 68824a439b76a4cca05609a02de7abdc42bd1d26afbfcd047b90001c610d3c56
2017-07-12 14:10:19 +00:00
drh
155507b344 Always make "column%d" column-names 1-based, never 0-based.
FossilOrigin-Name: 70096c505d702a9646da24613da387cee19afcf395d0294b3797c5ab50bb3ee2
2017-07-09 18:55:29 +00:00
drh
d7ca600ede Make sure the columns of a VALUES() clause are named "column%d" even if
the VALUES() clause is a subquery in the FROM clause of an outer query.

FossilOrigin-Name: acf3b9cc9c3932431979995a1dceacc06c659ab400fad95ce3728ff8895a022b
2017-07-09 00:30:58 +00:00
drh
269ba80469 Add the count-of-view optimization when compiled using
SQLITE_COUNTOFVIEW_OPTIMIZATION.

FossilOrigin-Name: d1ba200234f40b84327c7fc28c2584ed069da80e97578df71114d1a9ba9c559c
2017-07-04 19:34:36 +00:00
dan
5aa550cf3b Consider the values bound to SQL variables when determining whether or not a
partial index may be used.

FossilOrigin-Name: 7b59c353b805c64689b4ae9df347705acbb5f116346ad77af8ce087da7893747
2017-06-24 18:10:29 +00:00
dan
bd11a2acbb Ensure that the query planner knows that any column of a flattened LEFT JOIN
can be NULL even if that column is labeled with "NOT NULL".
Fix for ticket [892fc34f173e99d8].

FossilOrigin-Name: 483462682d3a57fb9dd85b4772596e9738f1694a454b8ebbc480b9452733e88d
2017-06-20 17:43:26 +00:00
drh
eafc6dfe2b When reusing a materialized view, make sure the estimated number of
output rows is correctly initialized to avoid confusing the query planner.

FossilOrigin-Name: 87aceb417a813a2925b636cffa660fe25f3b7e3b3c84bb2108951bf0af644376
2017-06-11 19:51:36 +00:00
drh
d4ab003d4e Avoid allocating excess memory to the KeyInfo objects.
FossilOrigin-Name: df7859435161f3ecf36635ce574bf73573516a1f2c6e75be320619364e7067d1
2017-05-31 13:45:59 +00:00
drh
979dd1be57 Optimizations to the Walker object and its methods to make the code a little
smaller and to help it run a little faster.

FossilOrigin-Name: 6854a34ed708259f2280f7ee56cec09f7fc99810dc739dc2814ddeae286aa2c4
2017-05-29 14:26:07 +00:00
drh
d8a295669c Correctly initialize the iSelectId of FROM clause terms that are a self
join of a reused materialized subquery.  Without this, the EXPLAIN QUERY PLAN
output for the query will identify the subquery using the uninitialized
(and arbitrary) iSelectId.

FossilOrigin-Name: 43c9ae371f6250fee98a7c4011726eff8ad37f5a97add4f490ac3a2dd501a0d2
2017-05-29 13:09:24 +00:00
drh
dc6de47925 The SQLITE_EXTRA_IFNULLROW compile-time option causes OP_IfNullRow opcodes
to be issued for references to the right-hand side table of *any* flattened
join, not just LEFT JOINs.  This puts extra stress on the OP_IfNUllRow opcodes
for testing purposes.

FossilOrigin-Name: 1a074c8a2bc0b28918ef905339d11a21d30101b4ea8c06c8b3faca7d17237538
2017-05-25 11:39:50 +00:00
drh
3c790f2a83 Disable the LEFT JOIN flattening optimization for aggregate queries, as it
does not currently work.  Further fix for ticket [cad1ab4cb7b0fc344].

FossilOrigin-Name: 05ada741554b9fd00befcf7b4083637f53de68699ae5210ac18e0773ad1a9910
2017-05-23 15:33:41 +00:00
drh
7c1544e051 Ensure that the expression rewriter inside the query flattener decends into
the substructure of the TK_IF_NULL_ROW operator.  This is a continuation
of the fix for ticket [cad1ab4cb7b0fc344].

FossilOrigin-Name: 28d2902d8f4dfbbfc610f271e00d353b0baae386b56b482f96b88836410be129
2017-05-23 12:44:57 +00:00
drh
10d33c1781 When flattening a query, make sure iTable attribute of TK_IF_NULL_ROW
operators (that result from a prior flattening of a LEFT JOIN) are updated
correctly.  Fix for ticket [cad1ab4cb7b0fc344].

FossilOrigin-Name: 919d36e6ea112c8c94f01e02611c48e78c441657e6898a466cfc39f1e0001403
2017-05-23 01:29:35 +00:00
drh
2336c935af Change the SQLITE_READ authorization call for unreferenced tables to use
an empty string for the column name, as this is less likely to impact legacy
authorization callbacks that assume column names are always non-NULL.

FossilOrigin-Name: 4139953ab528f20fa346409810edcb22adb6c1edc9d22f40b1b077ef842a2441
2017-05-11 12:05:23 +00:00
drh
701caf1eb1 Invoke the SQLITE_READ authorizer callback with a NULL column name for any
table referenced by a query but from when no columns are extracted.

FossilOrigin-Name: 92ab1f7257d2866c69eaaf4cf85990677b911ef425e9c5a36a96978cccfb551c
2017-05-10 16:12:00 +00:00
drh
3f4df4c6f3 Reuse the same materialization of a view when that view appears in a query
more than once, such as in a self-join.

FossilOrigin-Name: 9e35c89dbe744312f612e507b51ff9a5bb656def75392d25bc19fc638548cd1e
2017-05-02 17:54:19 +00:00
drh
a1f5c0c90b Move terms of the HAVING clause that reference only columns in the GROUP BY
clause over to the WHERE clause, resulting in a faster query plan.

FossilOrigin-Name: 47cbb471d056c8e1834a5ca72491404a3bfb273b5ff7bdd84b98d263938ea874
2017-05-02 16:55:07 +00:00
drh
c478c8daec Remove an unnecessary branch.
FossilOrigin-Name: a33179596f534067a58d68b77160f11ab13272b29f912d7cbe15ea00bbf03ade
2017-05-01 19:53:12 +00:00
drh
ed71298085 Futher bug fixes to the function that determines when a materialized view
can be reused.

FossilOrigin-Name: c64fe3a1695925693385d313e9ad2a1d8cb37ddaa8cc19920ae0978c91bc4c2c
2017-05-01 17:04:35 +00:00
drh
33543c237b Minor bug fixes and performance enhancement.
FossilOrigin-Name: b2aae559581d05e4e8caaddfbd137cb275d582e82d58872c6f42e462dc859b1c
2017-05-01 16:37:20 +00:00
drh
e08e8d6b61 Initial implementation of an optimization that attempts to reuse the same
materialization of a view on a self-join of the view.

FossilOrigin-Name: 478c34b9a8b5127d13024e10307aa832f160b89720c46424dd17555bd36f590d
2017-05-01 15:15:41 +00:00
dan
ab31a8450b Automatically transfer terms from the HAVING clause to the WHERE clause of an
aggregate query in cases where the result of evaluating the term depends only
one one or more of the GROUP BY expressions (and on no other inputs).

FossilOrigin-Name: 5375a3ce56f1d993b13b469fe33ec7679948f53940f62a15ddbaeb8aaa26a22c
2017-04-29 20:53:09 +00:00
drh
7555581fc1 Add a single testcase() macro to the subquery processing logic.
FossilOrigin-Name: 4e1df76e3d85922648e0e1cce73a266c3b1ed4511ace259ec0a01d7693af9e6f
2017-04-29 19:29:49 +00:00
drh
31d6fd5507 Add the TK_IF_NULL_ROW opcode to deal with non-NULL result columns in the
result set of a view or subquery on the RHS of a LEFT JOIN that gets flattened.

FossilOrigin-Name: 3a5860d86fadcf924316707918bf283d26c53b1473e5e67f5cff59d18c2a7742
2017-04-14 19:03:10 +00:00
drh
399c7e21e4 An initial attempt to optimize VIEWs that occur as the right operand of a
LEFT JOIN.  This particular check-in does not work correctly because it does
not deal with the case of columns in the VIEW that return non-NULL even when
all columns in the table of the VIEW are NULL because of the LEFT JOIN.

FossilOrigin-Name: 1838a59c8a1c151bd6fc822b0ffef661803cf0e4704c917e74a04567b81740b9
2017-04-14 17:18:45 +00:00
drh
46967de2ec Simplify the interface to the subst() routines that are part of the
query flattener by collecting common parameters into the SubstContext object
and passing around a pointer to that object.

FossilOrigin-Name: e651074be28b2077b8a298f513e2e699ceca99164ace9869ea50873c0637fd2e
2017-04-14 12:39:37 +00:00
drh
7c88bcd473 Remove an incorrect ALWAYS().
FossilOrigin-Name: f956f6ae6b0863c70c4a2b227338d5c332e87f2ec6f3fcdf2233c14c5b78ca2b
2017-04-14 12:27:45 +00:00
drh
ef90a6b8dd Very slight smaller and faster sqlite3SelectNew()
FossilOrigin-Name: 4143650c4ce32289d2301cdfc69bb10877246420f656bed122886f6803fc956a
2017-04-11 16:44:39 +00:00
drh
eac5fc0476 Improved comments. Fix a problem when an indexed expression is used in an
ORDER BY clause.

FossilOrigin-Name: c59eaf2b7cd2596733f349fc0fe979f71fd29bd73152a0c57066e0b69d5b7a4a
2017-04-11 01:01:27 +00:00
drh
aca19e19a8 Proof of concept for the ability to use the expression columns in an index
on expressions in place of equivalent expressions in the result set or in
the WHERE clause.  This check-in compiles but is mostly untested.

FossilOrigin-Name: a52ef2ad7c0e14b78b801f16a1f6ea8d8fa9ae5d7d810e18dd24c600b662a312
2017-04-07 19:41:31 +00:00
drh
dbd6a7dc06 Split off sqlite3DbFreeNN() from sqlite3DbFree() and use it in cases where
we know that the argument to be freed is never NULL.

FossilOrigin-Name: ad90e8bb5e47945607c8fb47b6ade8cfc52a9b684805cc40132629be0ecc14cc
2017-04-05 12:39:49 +00:00
drh
43606175e2 Combine the ExprList_item objects and the ExprList wrapper into a single
memory allocation, for improved performance and reduced footprint.

FossilOrigin-Name: 2b6560ad88b92820c383bcdc1e30c06f8b081ef7c6d9b1af71d2bb76c83e35cd
2017-04-05 11:32:13 +00:00
dan
c3becddb75 Fix a segfault that could occur following an OOM error in the
flattenSubquery() routine.

FossilOrigin-Name: c6dda3f752c184f441624c9993e77d5031063d79a0e177b6e25a9886514a742e
2017-03-13 14:30:40 +00:00
dan
067cd837fb Avoid a performance problem when very large "VALUES(..), (..), (..)" terms are
used in queries.

FossilOrigin-Name: f5306ad6816cc377036685cdae227e762885229c
2017-02-03 19:16:39 +00:00
dan
f91c1318f4 Changes to allow some multi-row UPDATE statements to avoid the two-pass
approach.

FossilOrigin-Name: 46db23ccd116ce5b9d949f9293be8a2818411b46
2017-01-10 20:04:38 +00:00
drh
a67b5cb6b3 Avoid duplicate b-tree searches in the duplicate row detector used to
implement DISTINCT.

FossilOrigin-Name: d577dda0a7fbfacda57e8cad2bc4651d2a05d813
2017-01-07 00:56:01 +00:00
dan
ac56ab7e48 Fix handling the case where a sub-query in a FROM clause is itself a UNION
ALL, and one side of that UNION ALL is a query on a view that includes an
ORDER BY. Fix for ticket [190c2507].

FossilOrigin-Name: 590ca83b8e8cdd5d24ed7f10f43e540aa0627f22
2017-01-05 17:23:11 +00:00
drh
f99dd359d0 Change the OP_IfNotZero opcode so that it decrements register P1 by 1 rather
than the value in P3, and so that it only decrements if originally positive.
This avoids decrementing the smallest 64-bit signed integer.

FossilOrigin-Name: 165c044686212fbf7182dd560ad1e57eb4cc9838
2016-12-18 17:42:00 +00:00
dan
69b9383edf Fix a problem causing the planner to generate sub-optimal plans for some queries
that use recursive WITH sub-queries with LIMIT clauses.

FossilOrigin-Name: 053a149cc8244a7f85137129cfcb8622efe90306
2016-12-16 15:05:40 +00:00
drh
79df77821a Refactor the Table.nRef field as Table.nTabRef for easier grepping.
FossilOrigin-Name: 9cae4c2e300e20304ced0dc8c1415c4922185928
2016-12-14 14:07:35 +00:00
drh
2700acaacd More changes to take advantage of the sqlite3VdbeAppendP4() method.
FossilOrigin-Name: 83bc5e40af9b20afeed008bf3e2669b7ac9e2dc8
2016-12-08 01:38:24 +00:00
dan
44c5604ced Prevent the flattening or where-term push-down optimizations from obscuring
misuses of SQL row values that can lead to crashes or assert() failures.

FossilOrigin-Name: 433d16ff3adfede3be53d5b0e0512f37e225591b
2016-12-07 15:38:37 +00:00