MDEV-16426 Optimizer erroneously treats equal constants of different formats as same
A cleanup for MDEV-14630: fixing a crash in Item_decimal::eq().
Problems:
- old implementations of Item_decimal::eq() and
Item_temporal_literal::eq() were not symmetric
with Item_param::eq(), this caused MDEV-11361.
- old implementations for DECIMAL and temporal data types
did not take into account that in case when eq() is called
with binary_cmp==true, {{eq()}} should check not only equality
of the two values, but also equality if their decimal precision.
This cuases MDEV-16426.
- Item_decimal::eq() crashes with "item" pointing
to a non-DECIMAL value. Before MDEV-14630
non-DECIMAL values were filtered out by the test:
type() == item->type()
as literals of different types had different type().
After MDEV-14630 type() for literals of all data types return CONST_ITEM.
This caused failures in tests:
./mtr engines/iuds.insert_number
./mtr --ps --embedded main.explain_slowquerylog
(revealed by buildbot)
The essence of the fix:
Making literals and Item_param reuse the same code to avoid
asymmetries between Item_param::eq(Item_literal) and
Item_literal::eq(Item_param), now and in the future, and to
avoid code duplication between Item_literal and Item_param.
Adding tests for "decimals" for DECIMAL and temporal data types,
to treat constants of different scale as not equal when "binary_cmp"
is "true".
Details:
1. Adding a helper class Item_const to extract constant values from Items easier
2. Deriving Item_basic_value from Item_const
3. Joining Type_handler::Item_basic_value_eq() and Item_basic_value_bin_eq()
into a single method with an extra "binary_cmp" argument
(it looks simple this way) and renaming the new method to Item_const_eq().
Modifying its implementations to operate with
Item_const instead of Item_basic_value.
4. Adding a new class Type_handler_hex_hybrid,
to handle hex constants like 0x616263.
5. Removing Item::VARBIN_ITEM and fixing Item_hex_constant to
use type_handler_hex_hybrid instead of type_handler_varchar.
Item_hex_hybrid::type() now returns CONST_ITEM, like all
other literals do.
6. Move virtual methods Item::type_handler_for_system_time() and
Item::cast_to_int_type_handler() from Item to Type_handler.
7. Removing Item_decimal::eq() and Item_temporal_literal::eq().
These classes are now handled by the generic Item_basic_value::eq().
8. Implementing Type_handler_temporal_result::Item_const_eq()
and Type_handler_decimal_result::Item_const_eq(),
this fixes MDEV-11361.
9. Adding tests for "decimals" into
Type_handler_decimal_result::Item_const_eq() and
Type_handler_temporal_result::Item_const_eq()
in case if "binary_cmp" is true.
This fixes MDEV-16426.
10. Moving Item_cache out of Item_basic_value.
They share nothing. It simplifies implementation
of Item_basic_value::eq(). Deriving Item_cache
directly from Item.
11. Adding class DbugStringItemTypeValue, which
used Item::print() internally, and using
in instead of the old debug printing code.
This gives nicer output in func_debug.result.
Changes N5 and N6 do not directly relate to the bugs fixed,
but make the code fully symmetric across all literal types.
Without a new handler Type_handler_hex_hybrid we'd have
to keep two code branches (for regular literals and for
hex hybrid literals).
The problem described in the bug report happened because the code
did not test check_cols(1) after fix_fields() in a few places.
Additionally, fix_fields() could be called multiple times for SP variables,
because they are all fixed at a early stage in append_for_log().
Solution:
1. Adding a few helper methods
- fix_fields_if_needed()
- fix_fields_if_needed_for_scalar()
- fix_fields_if_needed_for_bool()
- fix_fields_if_needed_for_order_by()
and using it in many cases instead of fix_fields() where
the "fixed" status is not definitely known to be "false".
2. Adding DBUG_ASSERT(!fixed) into Item_splocal*::fix_fields()
to catch double execution.
3. Adding tests.
As a good side effect, the patch removes a lot of duplicate code (~60 lines):
if (!item->fixed &&
item->fix_fields(..) &&
item->check_cols(1))
return true;
Now the boolean data type is preserved in hybrid functions and MIN/MAX,
so COALESCE(bool_expr,bool_expr) and MAX(bool_expr) are correctly
detected by JSON_OBJECT() as being boolean rather than numeric expressions.
- Removing tests of item->type() against INT_ITEM and replacing
them to calls of new method item->is_bool_literal().
- Changing constant conditions to use Item_bool() instead of Item_int().
1. Adding new methods:
- Item::is_order_clause_position()
- Item_splocal::is_valid_limit_clause_variable_with_error()
- Type_handler::is_order_clause_position_type()
- is_limit_clause_valid_type()
and changing all tests related to the ORDER and LIMIT clauses
like "item->type()==INT_ITEM" to these new methods.
2. Adding a helper function prepare_param() in sql_analyse.cc
and replacing three pieces of duplicate code to prepare_param() calls.
Replacing the test "item->type()!=Item::INT_ITEM" to an equivalent
condition using item->basic_const_item() and type_handler()->result_type().
Detailed: changes:
1. Moving Field specific code into new methods on Field:
- Field *Field::create_tmp_field(...)
- virtual void init_for_tmp_table(...)
2. Removing virtual Item::create_tmp_field().
Adding instead a new virtual method Item::create_tmp_field_ex().
Note, a virtual create_tmp_field() still exists, but only for Item_sum.
This resembles 10.0 code structure. Perhaps create_tmp_field() should
be removed from Item_sum and Item_sum descendants should override
create_tmp_field_ex() directly. This can be done in a separate commit.
3. Adding helper classes Tmp_field_src and Tmp_field_param,
to make the API for Item::create_tmp_field_ex() smaller
and easier to extend in the future.
4. Decomposing the public function create_tmp_field() into
virtual implementations for Item and a number of its descendants:
- Item_basic_value
- Item_sp_variable
- Item_name_const
- Item_result_field
- Item_field
- Item_ref
- Item_type_holder
- Item_row
- Item_func_sp
- Item_func_user_var
- Item_sum
- Item_sum_field
- Item_proc
5. Adding DBUG_ASSERT-only virtual implementations for
Item types that should not appear in create_tmp_table_ex(),
for easier debugging:
- Item_nodeset_func
- Item_nodeset_to_const_comparator
- Item_null_result
- Item_copy
- Item_ident_for_show
- Item_user_var_as_out_param
6. Moving public function create_tmp_field_from_field()
as a method to Item_field.
7. Removing Item::set_result_field(). It's not needed any more.
8. Cleanup: Removing the enum value "EXPR_CACHE_ITEM",
as it's not used for a very long time.
- Using array_elements() instead of a constant to iterate through an array
- Adding some comments
- Adding new-line function comments
- Using STRING_WITH_LEN instead of C_STRING_WITH_LEN
Queries involving rollup need all aggregate function to have copy_or_same function where we create a copy
of item_sum items for each sum level.
Implemented copy_or_same function for the custom aggregate function class (Item_sum_sp)
The logic and the implementation scheme are similar with the
MDEV-9197 Pushdown conditions into non-mergeable views/derived tables
How the push down is made on the example:
select * from t1
where a>3 and b>10 and
(a,b) in (select x,max(y) from t2 group by x);
-->
select * from t1
where a>3 and b>10 and
(a,b) in (select x,max(y)
from t2
where x>3
group by x
having max(y)>10);
The implementation scheme:
1. Search for the condition cond that depends only on the fields
from the left part of the IN subquery (left_part)
2. Find fields F_group in the select of the right part of the
IN subquery (right_part) that are used in the GROUP BY
3. Extract from the cond condition cond_where that depends only on the
fields from the left_part that stay at the same places in the left_part
(have the same indexes) as the F_group fields in the projection of the
right_part
4. Transform cond_where so it can be pushed into the WHERE clause of the
right_part and delete cond_where from the cond
5. Transform cond so it can be pushed into the HAVING clause of the right_part
The optimization is made in the
Item_in_subselect::pushdown_cond_for_in_subquery() and is controlled by the
variable condition_pushdown_for_subquery.
New test file in_subq_cond_pushdown.test is created.
There are also some changes made for setup_jtbm_semi_joins().
Now it is decomposed into the 2 procedures: setup_degenerate_jtbm_semi_joins()
that is called before optimize_cond() for cond and setup_jtbm_semi_joins()
that is called after optimize_cond().
New setup_jtbm_semi_joins() is made in the way so that the result of its work is
the same as if it was called before optimize_cond().
The code that is common for pushdown into materialized derived and into materialized
IN subqueries is factored out into pushdown_cond_for_derived(),
Item_in_subselect::pushdown_cond_for_in_subquery() and
st_select_lex::pushdown_cond_into_where_clause().
rename to post_fix_fields_part_expr_processor()
because it's only used after fix_fields in
fix_fields_part_func() and can be used for
various post-fix_fields fixups
MDEV-16100 FOR SYSTEM_TIME erroneously resolves string user variables as transaction IDs
Problem:
Vers_history_point::resolve_unit() tested item->result_type() before
item->fix_fields() was called.
- Item_func_get_user_var::result_type() returned REAL_RESULT by default.
This caused MDEV-16100.
- Item_func_sp::result_type() crashed on assert.
This caused MDEV-16094
Changes:
1. Adding item->fix_fields() into Vers_history_point::resolve_unit()
before using data type specific properties of the history point
expression.
2. Adding a new virtual method Type_handler::Vers_history_point_resolve_unit()
3. Implementing type-specific
Type_handler_xxx::Type_handler::Vers_history_point_resolve_unit()
in the way to:
a. resolve temporal and general purpose string types to TIMESTAMP
b. resolve BIT and general purpose INT types to TRANSACTION
c. disallow use of non-relevant data type expressions in FOR SYSTEM_TIME
Note, DOUBLE and DECIMAL data types are disallowed intentionally.
- DOUBLE does not have enough precision to hold huge BIGINT UNSIGNED values
- DECIMAL rounds on conversion to INT
Both lack of precision and rounding might potentionally lead to
very unpredictable results when a wrong transaction ID would be chosen.
If one really wants dangerous use of DOUBLE and DECIMAL, explicit CAST
can be used:
FOR SYSTEM_TIME AS OF CAST(double_or_decimal AS UNSIGNED)
QQ: perhaps DECIMAL(N,0) could still be allowed.
4. Adding a new virtual method Item::type_handler_for_system_time(),
to make HEX hybrids and bit literals work as TRANSACTION rather
than TIMESTAMP.
5. sql_yacc.yy: replacing the rule temporal_literal to "TIMESTAMP TEXT_STRING".
Other temporal literals now resolve to TIMESTAMP through the new
Type_handler methods. No special grammar needed. This removed
a few shift/resolve conflicts.
(TIMESTAMP related conflicts in "history_point:" will be removed separately)
6. Removing the "timestamp_only" parameter from
vers_select_conds_t::resolve_units() and Vers_history_point::resolve_unit().
It was a hint telling that a table did not have any TRANSACTION-aware
system time columns, so it's OK to resolve to TIMESTAMP in case of uncertainty.
In the new reduction it works as follows:
- the decision between TIMESTAMP and TRANSACTION is first made
based only on the expression data type only
- then, in case if the expression resolved to TRANSACTION, the table
is checked if TRANSACTION-aware columns really exist.
This way is safer against possible ALTER TABLE statements changing
ROW START and ROW END columns from "BIGINT UNSIGNED" to "TIMESTAMP(x)"
or the other way around.
multiple times with different arguments.
If the ON expression of an outer join is an OR formula with one
of the disjunct being a constant formula then the expression
cannot be null-rejected if the constant formula is true. Otherwise
it can be null-rejected and if so the outer join can be converted
into inner join. This optimization was added in the patch for
mdev-4817. Yet the code had a defect: if the query was used in
a stored procedure with parameters and the constant item contained
some of them then the value of this constant item depended on the
values of the parameters. With some parameters it may be true,
for others not. The validity of conversion to inner join is checked
only once and it happens only for the first call of procedure.
So if the parameters in the first call allowed the conversion it
was done and next calls used the transformed query though there
could be calls whose parameters made the conversion invalid.
Fixed by cheking whether the constant disjunct in the ON expression
originally contained an SP parameter. If so the expression is not
considered as null-rejected. For this check a new item's attribute
was intruduced: Item::with_param. It is calculated for each item
by fix fields() functions.
Also moved the call of optimize_constant_subqueries() in
JOIN::optimize after the call of simplify_joins(). The reason
for this is that after the optimization introduced by the patch
for mdev-4817 simplify_joins() can use the results of execution
of non-expensive constant subqueries and this is not valid.
This patch does the following:
1. Makes Field_vers_trx_id::type_handler() return
&type_handler_vers_trx_id rather than &type_handler_longlong.
Fixes Item_func::convert_const_compared_to_int_field() to
test field_item->type_handler() against &type_handler_vers_trx_id,
instead of testing field_item->vers_trx_id().
2. Removes VERS_TRX_ID related code from
Type_handler_hybrid_field_type::aggregate_for_comparison(),
because "BIGINT UNSIGNED GENERATED ALWAYS AS ROW {START|END}"
columns behave just like a BIGINT in a regular comparison,
i.e. when not inside AS OF.
3. Removes
- Type_handler_hybrid_field_type::m_vers_trx_id;
- Type_handler_hybrid_field_type::m_flags;
because a "BIGINT UNSIGNED GENERATED ALWAYS AS ROW {START|END}"
behaves like a regular BIGINT column when in UNION.
4. Removes Field::vers_trx_id(), Item::vers_trx_id(), Item::field_flags()
They are not needed anymore. See N1.
Problems:
1. Unlike Item_field::fix_fields(),
Item_sum_sp::fix_length_and_dec() and Item_func_sp::fix_length_and_dec()
did not run the code which resided in adjust_max_effective_column_length(),
therefore they did not extend max_length for the integer return data types
from the user-specified length to the maximum length according to
the data type capacity.
2. The code in adjust_max_effective_column_length() was not correct
for TEXT data, because Field_blob::max_display_length()
multiplies to mbmaxlen. So TEXT variants were unintentionally
promoted to the next longer data type for multi-byte character
sets: TINYTEXT->TEXT, TEXT->MEDIUMTEXT, MEDIUMTEXT->LONGTEXT.
3. Item_sum_sp::create_table_field_from_handler()
Item_func_sp::create_table_field_from_handler()
erroneously called tmp_table_field_from_field_type(),
which converted VARCHAR(>512) to TEXT variants.
So "CREATE..SELECT spfunc()" erroneously converted
VARCHAR to TEXT. This was wrong, because stored
functions have explicitly declared data types,
which should be preserved.
Solution:
- Removing Type_std_attributes(const Field *)
and using instead Type_std_attributes::set() in combination
with field->type_str_attributes() all around the code, e.g.:
Type_std_attributes::set(field->type_std_attributes())
These two ways of copying attributes from a Field
to an Item duplicated each other, and were slightly
different in how to mix max_length and mbmaxlen.
- Removing adjust_max_effective_column_length() and
fixing Field::type_std_attributes() to do all necessary
type-specific calculations , so no further adjustments
is needed.
Field::type_std_attributes() is now called from all affected methods:
Item_field::fix_fields()
Item_sum_sp::fix_length_and_dec()
Item_func_sp::fix_length_and_dec()
This fixes the problem N1.
- Making Field::type_std_attributes() virtual, to make
sure that type-specific adjustments a properly done
by individual Field_xxx classes. Implementing
Field_blob::type_std_attributes() in the way that
no TEXT promotion is done.
This fixes the problem N2.
- Fixing Item_sum_sp::create_table_field_from_handler()
Item_func_sp::create_table_field_from_handler() to
call create_table_field_from_handler() instead of
tmp_table_field_from_field_type() to avoid
VARCHAR->TEXT conversion on "CREATE..SELECT spfunc()".
- Recording mysql-test/suite/compat/oracle/r/sp-param.result
as "CREATE..SELECT spfunc()" now correctly
preserve the data type as specified in the RETURNS clause.
- Adding new tests
Renaming methods:
- Field::make_field(Send_field*) to make_send_field(..)
- Item::make_field(THD *,Send_field *) to make_send_field(..)
- Item::init_make_field(Send_field *, enum_field_type) to init_make_send_field(..)
These names looked similar to other functions that are used
for a very different purpose (creating Field instances):
- Public function "Field * make_field(..)"
- Method "Field *Column_defitinion::make_field(..)"
The rename makes it's easier to search the code using "grep".
Do not assume that it's always item->field_type() - this is not the case
in temporal comparisons (e.g. when comparing DATETIME column with a TIME
literal).
- CREATE PACKAGE [BODY] statements are now
entirely written to mysql.proc with type='PACKAGE' and type='PACKAGE BODY'.
- CREATE PACKAGE BODY now supports IF NOT EXISTS
- DROP PACKAGE BODY now supports IF EXISTS
- CREATE OR REPLACE PACKAGE [BODY] is now supported
- CREATE PACKAGE [BODY] now support the DEFINER clause:
CREATE DEFINER user@host PACKAGE pkg ... END;
CREATE DEFINER user@host PACKAGE BODY pkg ... END;
- CREATE PACKAGE [BODY] now supports SQL SECURITY and COMMENT clauses, e.g.:
CREATE PACKAGE p1 SQL SECURITY INVOKER COMMENT "comment" AS ... END;
- Package routines are now created from the package CREATE PACKAGE BODY
statement and don't produce individual records in mysql.proc.
- CREATE PACKAGE BODY now supports package-wide variables.
Package variables can be read and set inside package routines.
Package variables are stored in a separate sp_rcontext,
which is cached in THD on the first packate routine call.
- CREATE PACKAGE BODY now supports the initialization section.
- All public routines (i.e. declared in CREATE PACKAGE)
must have implementations in CREATE PACKAGE BODY
- Only public package routines are available outside of the package
- {CREATE|DROP} PACKAGE [BODY] now respects CREATE ROUTINE and ALTER ROUTINE
privileges
- "GRANT EXECUTE ON PACKAGE BODY pkg" is now supported
- SHOW CREATE PACKAGE [BODY] is now supported
- SHOW PACKAGE [BODY] STATUS is now supported
- CREATE and DROP for PACKAGE [BODY] now works for non-current databases
- mysqldump now supports packages
- "SHOW {PROCEDURE|FUNCTION) CODE pkg.routine" now works for package routines
- "SHOW PACKAGE BODY CODE pkg" now works (the package initialization section)
- A new package body level MDL was added
- Recursive calls for package procedures are now possible
- Routine forward declarations in CREATE PACKATE BODY are now supported.
- Package body variables now work as SP OUT parameters
- Package body variables now work as SELECT INTO targets
- Package body variables now support ROW, %ROWTYPE, %TYPE
Lots of changes:
* calculate the current history partition in ::external_lock(),
not in ::write_row() or ::update_row()
* remove dynamically collected per-partition row_end stats
* no full table scan in open_table_from_share to calculate these
stats, no manual MDL/thr_locks in open_table_from_share
* no shared stats in TABLE_SHARE = no mutexes or condition waits when
calculating current history partition
* always compare timestamps, don't convert them to MYSQL_TIME
(avoid DST ambiguity, and it's faster too)
* correct interval handling, 1 month = 1 month, not 30 * 24 * 3600 seconds
* save/restore first partition start time, and count intervals from there
* only allow to drop first partitions if INTERVAL
* when adding new history partitions, split the data in the last history
parition, if it was overflowed
* show partition boundaries in INFORMATION_SCHEMA.PARTITIONS
The problem was that Item_func_hybrid_field_type::get_date() did not
convert the result to the correct data type, so MYSQL_TIME::time_type
of the get_date() result could be not in sync with field_type().
Changes:
1. Adding two new classes Datetime and Date to store MYSQL_TIMESTAMP_DATETIME
and MYSQL_TIMESTAMP_DATE values respectively
(in addition to earlier added class Time, for MYSQL_TIMESTAMP_TIME values).
2. Adding Item_func_hybrid_field_type::time_op().
It performs the operation using TIME representation,
and always returns a MYSQL_TIME value with time_type=MYSQL_TIMESTAMP_TIME.
Implementing time_op() for all affected children classes.
3. Fixing all implementations of date_op() to perform the operation
using strictly DATETIME representation. Now they always return a MYSQL_TIME
value with time_type=MYSQL_TIMESTAMP_{DATE|DATETIME},
according to the result data type.
4. Removing assignment of ltime.time_type to mysql_timestamp_type()
from all val_xxx_from_date_op(), because now date_op() makes sure
to return a proper MYSQL_TIME value with a good time_type (and other member)
5. Adding Item_func_hybrid_field_type::val_xxx_from_time_op().
6. Overriding Type_handler_time_common::Item_func_hybrid_field_type_val_xxx()
to call val_xxx_from_time_op() instead of val_xxx_from_date_op().
7. Modified Item_func::get_arg0_date() to return strictly a TIME value
if TIME_TIME_ONLY is passed, or return strictly a DATETIME value otherwise.
If args[0] returned a value of a different temporal type,
(for example a TIME value when TIME_TIME_ONLY was not passed,
or a DATETIME value when TIME_TIME_ONLY was passed), the conversion
is automatically applied.
Earlier, get_arg0_date() did not guarantee a result in
accordance to TIME_TIME_ONLY flag.
