mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-10-01 13:29:18 +02:00
Code Issues Projects Releases Packages Wiki Activity
mariadb/sql/opt_hints_parser.cc
Dave Gosselin 16f2b7e3df MDEV-36125 [NO_]INDEX_MERGE Hint 
Introduces NO_INDEX_MERGE and INDEX_MERGE, which control whether or
not index merge strategies are used during query optimization.  Here
is an example query from the tests:

  SET optimizer_switch='index_merge_intersection=off';
  EXPLAIN SELECT /*+ INDEX_MERGE(t1 f4, f2) */ COUNT(*) FROM t1
  WHERE f4 = 'h' AND f2 = 2;

with the hint in place, the query plan will employ the index_merge
intersect strategy (abbreviated EXPLAIN output):

  type		Extra
  index_merge	Using intersect(f2,f4); Using where; Using index

The presence of the [NO_]INDEX_MERGE hint overrides the optimizer's
choice of keys during the index merge optimization.  As we see in
the above example, keys f2 and f4 and given and the optimizer will
consider only those keys for this query.

When the hint is given without any particular keys, as in
INDEX_MERGE(table), then all keys are considered.  In this case, the
cheapest index merge among the keys should be used.  When
NO_INDEX_MERGE(table) is given, then index merge is disabled for
that table.

When the hint is given with one or more keys, then only those keys
are considered.  In the case of NO_INDEX_MERGE, those keys are
excluded.  This can lead to no merged indexes at all, because
there may not be sufficient row-ordered read columns available for
consideration.

The index merge strategies of intersection, union, and sort union
cannot themselves be directly controlled via the hints.  In combination
with the optimizer switches for the same, the strategy may be
indirectly controlled but this is not guaranteed.

When the hint directs the optimizer such that insufficient ROR scans
are available, thus leading to a situation where the INDEX_MERGE hint
cannot be honored, the server will emit a warning to that effect.

In the hints module (opt_hints*{cc,h}), this commit adds some
index merge-specific functionality to make interpreting hint state
at callsites in the optimizer cleaner and more intuitive.
Particularly, we add a bit field to the table hints class which
indicates the keys that are marked by an [NO_]INDEX_MERGE hint, if
present.  A new function, index_merge_hint (and associated new
helper functions) relies on this field when interpreting index merge
hint state for the optimizer.

If there are no index merges available prior to attemping to find
a suitable union/sort union, then the optimizer will not attempt
it.  This change results in optimizer trace output which does not
include the 'analyzing_index_merge_union' block when there are no
merges.

Parts of this implementation based on MySQL commit
ebcb981807e3d91a64782e89d48e1a25622eafea
2025-09-05 15:10:13 +03:00

1318 lines
41 KiB
C++
Raw Permalink Blame History

/*
Copyright (c) 2024, MariaDB
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; version 2 of
the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1335 USA
*/
#include "opt_hints_parser.h"
#include "sql_error.h"
#include "mysqld_error.h"
#include "sql_class.h"
#include "sql_show.h"
#include "opt_hints.h"
using Parser= Optimizer_hint_parser;
extern struct st_opt_hint_info opt_hint_info[];
// Forward declaration of functions
void print_warn(THD *thd, uint err_code, opt_hints_enum hint_type,
bool hint_state,
const Lex_ident_sys *qb_name_arg,
const Lex_ident_sys *table_name_arg,
const Lex_ident_sys *key_name_arg,
const Printable_parser_rule *hint);
Opt_hints_qb *get_qb_hints(Parse_context *pc);
Opt_hints_qb *find_qb_hints(Parse_context *pc,
const Lex_ident_sys &qb_name,
opt_hints_enum hint_type,
bool hint_state);
Opt_hints_global *get_global_hints(Parse_context *pc);
Opt_hints_table *get_table_hints(Parse_context *pc,
const Lex_ident_sys &table_name,
Opt_hints_qb *qb);
void append_table_name(THD *thd, String *str, const LEX_CSTRING &table_name,
const LEX_CSTRING &qb_name);
static const Lex_ident_sys null_ident_sys;
Parse_context::Parse_context(THD *thd, st_select_lex *select)
: thd(thd),
mem_root(thd->mem_root),
select(select)
{}
Optimizer_hint_tokenizer::TokenID
Optimizer_hint_tokenizer::find_keyword(const LEX_CSTRING &str)
{
switch (str.length)
{
case 3:
if ("BKA"_Lex_ident_column.streq(str)) return TokenID::keyword_BKA;
if ("BNL"_Lex_ident_column.streq(str)) return TokenID::keyword_BNL;
if ("MRR"_Lex_ident_column.streq(str)) return TokenID::keyword_MRR;
break;
case 5:
if ("MERGE"_Lex_ident_column.streq(str)) return TokenID::keyword_MERGE;
if ("INDEX"_Lex_ident_column.streq(str)) return TokenID::keyword_INDEX;
break;
case 6:
if ("NO_BKA"_Lex_ident_column.streq(str)) return TokenID::keyword_NO_BKA;
if ("NO_BNL"_Lex_ident_column.streq(str)) return TokenID::keyword_NO_BNL;
if ("NO_ICP"_Lex_ident_column.streq(str)) return TokenID::keyword_NO_ICP;
if ("NO_MRR"_Lex_ident_column.streq(str)) return TokenID::keyword_NO_MRR;
break;
case 7:
if ("QB_NAME"_Lex_ident_column.streq(str))
return TokenID::keyword_QB_NAME;
break;
case 8:
if ("SEMIJOIN"_Lex_ident_column.streq(str))
return TokenID::keyword_SEMIJOIN;
if ("SUBQUERY"_Lex_ident_column.streq(str))
return TokenID::keyword_SUBQUERY;
if ("NO_MERGE"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_MERGE;
if ("NO_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_INDEX;
break;
case 9:
if ("LOOSESCAN"_Lex_ident_column.streq(str))
return TokenID::keyword_LOOSESCAN;
break;
case 10:
if ("FIRSTMATCH"_Lex_ident_column.streq(str))
return TokenID::keyword_FIRSTMATCH;
if ("INTOEXISTS"_Lex_ident_column.streq(str))
return TokenID::keyword_INTOEXISTS;
if ("JOIN_ORDER"_Lex_ident_column.streq(str))
return TokenID::keyword_JOIN_ORDER;
if ("JOIN_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_JOIN_INDEX;
break;
case 11:
if ("NO_SEMIJOIN"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_SEMIJOIN;
if ("DUPSWEEDOUT"_Lex_ident_column.streq(str))
return TokenID::keyword_DUPSWEEDOUT;
if ("JOIN_PREFIX"_Lex_ident_column.streq(str))
return TokenID::keyword_JOIN_PREFIX;
if ("JOIN_SUFFIX"_Lex_ident_column.streq(str))
return TokenID::keyword_JOIN_SUFFIX;
if ("ORDER_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_ORDER_INDEX;
if ("GROUP_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_GROUP_INDEX;
if ("INDEX_MERGE"_Lex_ident_column.streq(str))
return TokenID::keyword_INDEX_MERGE;
break;
case 13:
if ("NO_JOIN_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_JOIN_INDEX;
break;
case 14:
if ("NO_ORDER_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_ORDER_INDEX;
if ("NO_GROUP_INDEX"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_GROUP_INDEX;
if ("NO_INDEX_MERGE"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_INDEX_MERGE;
break;
case 15:
if ("MATERIALIZATION"_Lex_ident_column.streq(str))
return TokenID::keyword_MATERIALIZATION;
break;
case 16:
if ("JOIN_FIXED_ORDER"_Lex_ident_column.streq(str))
return TokenID::keyword_JOIN_FIXED_ORDER;
break;
case 18:
if ("MAX_EXECUTION_TIME"_Lex_ident_column.streq(str))
return TokenID::keyword_MAX_EXECUTION_TIME;
if ("SPLIT_MATERIALIZED"_Lex_ident_column.streq(str))
return TokenID::keyword_SPLIT_MATERIALIZED;
break;
case 21:
if ("NO_RANGE_OPTIMIZATION"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_RANGE_OPTIMIZATION;
if ("NO_SPLIT_MATERIALIZED"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_SPLIT_MATERIALIZED;
break;
case 26:
if ("DERIVED_CONDITION_PUSHDOWN"_Lex_ident_column.streq(str))
return TokenID::keyword_DERIVED_CONDITION_PUSHDOWN;
break;
case 29:
if ("NO_DERIVED_CONDITION_PUSHDOWN"_Lex_ident_column.streq(str))
return TokenID::keyword_NO_DERIVED_CONDITION_PUSHDOWN;
break;
}
if (str.length > 0 && (str.str[0] >= '0' && str.str[0] <= '9'))
{
/*
If all characters are digits, qualify the token as a number,
otherwise as an identifier
*/
for(size_t i = 1; i < str.length; i++)
{
if (str.str[i] < '0' || str.str[i] > '9')
return TokenID::tIDENT;
}
return TokenID::tUNSIGNED_NUMBER;
}
return TokenID::tIDENT;
}
Optimizer_hint_tokenizer::Token
Optimizer_hint_tokenizer::get_token(CHARSET_INFO *cs)
{
get_spaces();
if (eof())
return Token(Lex_cstring(m_ptr, m_ptr), TokenID::tEOF);
const char head= m_ptr[0];
if (head == '`' || head=='"')
{
const Token_with_metadata delimited_ident= get_quoted_string();
/*
Consider only non-empty quoted strings as identifiers.
Table and index names cannot be empty in MariaDB.
Let's also disallow empty query block names.
Note, table aliases can actually be empty:
SELECT ``.a FROM t1 ``;
But let's disallow them in hints for simplicity, to handle
all identifiers in the same way in the hint parser.
*/
if (delimited_ident.length > 2)
return Token(delimited_ident, TokenID::tIDENT);
/*
If the string is empty, "unget" it to have a good
syntax error position in the message text.
The point is to include the empty string in the error message:
EXPLAIN EXTENDED SELECT ... QB_NAME(``) ...; -->
Optimizer hint syntax error near '``) ...' at line 1
*/
m_ptr-= delimited_ident.length;
return Token(Lex_cstring(m_ptr, m_ptr), TokenID::tNULL);
}
const Token_with_metadata ident= get_ident();
if (ident.length)
return Token(ident, ident.m_extended_chars ?
TokenID::tIDENT : find_keyword(ident));
if (!get_char(','))
return Token(Lex_cstring(m_ptr - 1, 1), TokenID::tCOMMA);
if (!get_char('@'))
return Token(Lex_cstring(m_ptr - 1, 1), TokenID::tAT);
if (!get_char('('))
return Token(Lex_cstring(m_ptr - 1, 1), TokenID::tLPAREN);
if (!get_char(')'))
return Token(Lex_cstring(m_ptr - 1, 1), TokenID::tRPAREN);
return Token(Lex_cstring(m_ptr, m_ptr), TokenID::tNULL);
}
// This method is for debug purposes
bool Parser::parse_token_list(THD *thd)
{
for ( ; ; m_look_ahead_token= get_token(m_cs))
{
char tmp[200];
my_snprintf(tmp, sizeof(tmp), "TOKEN: %d %.*s",
(int) m_look_ahead_token.id(),
(int) m_look_ahead_token.length,
m_look_ahead_token.str);
push_warning(thd, Sql_condition::WARN_LEVEL_WARN,
ER_UNKNOWN_ERROR, tmp);
if (m_look_ahead_token.id() == TokenID::tNULL ||
m_look_ahead_token.id() == TokenID::tEOF)
break;
}
return true; // Success
}
void Parser::push_warning_syntax_error(THD *thd, uint start_lineno)
{
DBUG_ASSERT(m_start <= m_ptr);
DBUG_ASSERT(m_ptr <= m_end);
const char *msg= ER_THD(thd, ER_WARN_OPTIMIZER_HINT_SYNTAX_ERROR);
ErrConvString txt(m_look_ahead_token.str, strlen(m_look_ahead_token.str),
thd->variables.character_set_client);
/*
start_lineno is the line number on which the whole hint started.
Add the line number of the current tokenizer position inside the hint
(in case hints are written in multiple lines).
*/
push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN,
ER_PARSE_ERROR, ER_THD(thd, ER_PARSE_ERROR),
msg, txt.ptr(), start_lineno + lineno());
}
bool
Parser::Table_name_list_container::add(Optimizer_hint_parser *p,
Table_name &&elem)
{
Table_name *pe= (Table_name*) p->m_thd->alloc(sizeof(*pe));
if (!pe)
return true;
*pe= std::move(elem);
return push_back(pe, p->m_thd->mem_root);
}
bool Parser::Hint_param_table_list_container::add(Optimizer_hint_parser *p,
Hint_param_table &&elem)
{
Hint_param_table *pe= (Hint_param_table*) p->m_thd->alloc(sizeof(*pe));
if (!pe)
return true;
*pe= std::move(elem);
return push_back(pe, p->m_thd->mem_root);
}
bool Parser::Hint_param_index_list_container::add(Optimizer_hint_parser *p,
Hint_param_index &&elem)
{
Hint_param_index *pe= (Hint_param_index*) p->m_thd->alloc(sizeof(*pe));
if (!pe)
return true;
*pe= std::move(elem);
return push_back(pe, p->m_thd->mem_root);
}
bool Parser::Hint_list_container::add(Optimizer_hint_parser *p, Hint &&elem)
{
Hint *pe= new (p->m_thd->mem_root) Hint;
if (!pe)
return true;
*pe= std::move(elem);
return push_back(pe, p->m_thd->mem_root);
}
bool Parser::Semijoin_strategy_list_container::add(Optimizer_hint_parser *p,
Semijoin_strategy &&elem)
{
Semijoin_strategy *pe= (Semijoin_strategy*) p->m_thd->alloc(sizeof(*pe));
if (!pe)
return true;
*pe= std::move(elem);
return push_back(pe, p->m_thd->mem_root);
}
/*
Resolve a parsed table level hint, i.e. set up proper Opt_hint_* structures
which will be used later during query preparation and optimization.
Return value:
- false: no critical errors, warnings on duplicated hints,
unresolved query block names, etc. are allowed
- true: critical errors detected, break further hints processing
*/
bool Parser::Table_level_hint::resolve(Parse_context *pc) const
{
const Table_level_hint_type &table_level_hint_type= *this;
opt_hints_enum hint_type;
bool hint_state; // ON or OFF
switch (table_level_hint_type.id())
{
case TokenID::keyword_BNL:
hint_type= BNL_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_BNL:
hint_type= BNL_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_BKA:
hint_type= BKA_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_BKA:
hint_type= BKA_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_DERIVED_CONDITION_PUSHDOWN:
hint_type= DERIVED_CONDITION_PUSHDOWN_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_DERIVED_CONDITION_PUSHDOWN:
hint_type= DERIVED_CONDITION_PUSHDOWN_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_MERGE:
hint_type= MERGE_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_MERGE:
hint_type= MERGE_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_SPLIT_MATERIALIZED:
hint_type= SPLIT_MATERIALIZED_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_SPLIT_MATERIALIZED:
hint_type= SPLIT_MATERIALIZED_HINT_ENUM;
hint_state= false;
break;
default:
DBUG_ASSERT(0);
return true;
}
if (const At_query_block_name_opt_table_name_list &
at_query_block_name_opt_table_name_list= *this)
{
// this is @ query_block_name opt_table_name_list
const Lex_ident_sys qb_name_sys= Query_block_name::to_ident_sys(pc->thd);
Opt_hints_qb *qb= find_qb_hints(pc, qb_name_sys, hint_type, hint_state);
if (qb == NULL)
return false;
if (at_query_block_name_opt_table_name_list.is_empty())
{
// e.g. BKA(@qb1)
if (qb->set_switch(hint_state, hint_type, false))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, hint_state,
&qb_name_sys, nullptr, nullptr, nullptr);
}
return false;
}
else
{
// e.g. BKA(@qb1 t1, t2, t3)
const Opt_table_name_list &opt_table_name_list= *this;
for (const Table_name &table : opt_table_name_list)
{
const Lex_ident_sys table_name_sys= table.to_ident_sys(pc->thd);
Opt_hints_table *tab= get_table_hints(pc, table_name_sys, qb);
if (!tab)
return false;
if (tab->set_switch(hint_state, hint_type, true))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, hint_state,
&qb_name_sys, &table_name_sys, nullptr,
nullptr);
}
}
}
}
else
{
// this is opt_hint_param_table_list
const Opt_hint_param_table_list &opt_hint_param_table_list= *this;
Opt_hints_qb *qb= find_qb_hints(pc, Lex_ident_sys(), hint_type, hint_state);
if (qb == NULL)
return false;
if (opt_hint_param_table_list.is_empty())
{
// e.g. BKA()
if (qb->set_switch(hint_state, hint_type, false))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, hint_state,
&null_ident_sys, nullptr, nullptr, nullptr);
}
return false;
}
for (const Hint_param_table &table : opt_hint_param_table_list)
{
// e.g. BKA(t1@qb1, t2@qb2, t3)
const Lex_ident_sys qb_name_sys= table.Query_block_name::
to_ident_sys(pc->thd);
Opt_hints_qb *qb= find_qb_hints(pc, qb_name_sys, hint_type, hint_state);
if (qb == NULL)
return false;
const Lex_ident_sys table_name_sys= table.Table_name::
to_ident_sys(pc->thd);
Opt_hints_table *tab= get_table_hints(pc, table_name_sys, qb);
if (!tab)
return false;
if (tab->set_switch(hint_state, hint_type, true))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, hint_state,
&qb_name_sys, &table_name_sys, nullptr, nullptr);
}
}
}
return false;
}
/*
Resolve a parsed index level hint, i.e. set up proper Opt_hint_* structures
which will be used later during query preparation and optimization.
Return value:
- false: no critical errors, warnings on duplicated hints,
unresolved query block names, etc. are allowed
- true: critical errors detected, break further hints processing
Taxonomy of index hints
- 2 levels of hints:
- table level hints: only table name specified but no index names
- index level hints: both table name and index names specified
- 2 kinds of hints:
- global: [NO_]INDEX
- non-global: [NO_]JOIN_INDEX, [NO_]GROUP_INDEX, [NO_]ORDER_INDEX
- 4 types of hints:
- [NO_]JOIN_INDEX
- [NO_]GROUP_INDEX
- [NO_]ORDER_INDEX
- [NO_]INDEX
Conflict checking
- A conflict happens if and only if
- for a table level hint
- a hint of the same type or opposite kind has already been specified
for the same table
- for a index level hint
- the same type of hint has already been specified for the same
table or for the same index, OR
- the opposite kind of hint has already been specified for the
same index
- For a multi index hint like JOIN_INDEX(t1 i1, i2, i3), it conflicts
with a previous hint if any of the JOIN_INDEX(t1 i1), JOIN_INDEX(t1
i2), JOIN_INDEX(t1 i3) conflicts with a previous hint
When a hint type is specified for an index, it is also marked as
specified with the same switch state for the table
*/
bool Parser::Index_level_hint::resolve(Parse_context *pc) const
{
const Index_level_hint_type &index_level_hint_type= *this;
opt_hints_enum hint_type;
bool hint_state; // ON or OFF
/*
Intended for the [NO_]INDEX_MERGE hints originally, this is set
true when the hint should be set also on the table as well as
the index. However, for such a hint set on the table, the check_parent flag
is inverted from this value; that is, if set_also_on_table is
true, then check_parent will be false (see below). This is
because the presence of the hint on the table is treated differently
than the hint applying to all keys on the table. For example,
INDEX_MERGE(table) means that the hint applies to all keys while
INDEX_MERGE(table, k1) means that the hint applies to key k1. In
both cases, however, we set the hint both as an index-level hint and
as a table-level hint, but when inspecting it during index_merge_hint,
it is given special treatment (see that function for details).
*/
bool set_also_on_table= false;
switch (index_level_hint_type.id())
{
case TokenID::keyword_NO_ICP:
hint_type= ICP_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_MRR:
hint_type= MRR_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_MRR:
hint_type= MRR_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_NO_RANGE_OPTIMIZATION:
hint_type= NO_RANGE_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_INDEX:
hint_type= INDEX_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_INDEX:
hint_type= INDEX_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_JOIN_INDEX:
hint_type= JOIN_INDEX_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_JOIN_INDEX:
hint_type= JOIN_INDEX_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_ORDER_INDEX:
hint_type= ORDER_INDEX_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_ORDER_INDEX:
hint_type= ORDER_INDEX_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_GROUP_INDEX:
hint_type= GROUP_INDEX_HINT_ENUM;
hint_state= true;
break;
case TokenID::keyword_NO_GROUP_INDEX:
hint_type= GROUP_INDEX_HINT_ENUM;
hint_state= false;
break;
case TokenID::keyword_INDEX_MERGE:
hint_type= INDEX_MERGE_HINT_ENUM;
hint_state= true;
set_also_on_table= true;
break;
case TokenID::keyword_NO_INDEX_MERGE:
hint_type= INDEX_MERGE_HINT_ENUM;
hint_state= false;
set_also_on_table= true;
break;
default:
DBUG_ASSERT(0);
return true;
}
const Hint_param_table_ext &table_ext= *this;
const Lex_ident_sys qb_name_sys= table_ext.Query_block_name::
to_ident_sys(pc->thd);
const Lex_ident_sys table_name_sys= table_ext.Table_name::
to_ident_sys(pc->thd);
Opt_hints_qb *qb= find_qb_hints(pc, qb_name_sys, hint_type, hint_state);
if (qb == NULL)
return false;
Opt_hints_table *tab= get_table_hints(pc, table_name_sys, qb);
if (!tab)
return false;
const Lex_ident_sys key_conflict(
STRING_WITH_LEN("another hint was already specified for this index"));
/*
If no index names are given, this is a table level hint, for example:
GROUP_INDEX(t1), NO_MRR(t2).
Otherwise this is a group of index-level hints:
NO_INDEX(t1 idx1, idx2) NO_ICP(t2 idx_a, idx_b, idx_c)
*/
if (is_empty() || set_also_on_table) // Table level hint
{
uint warn_code= 0;
if (is_compound_hint(hint_type) &&
is_index_hint_conflicting(tab, nullptr, hint_type))
{
warn_code= ER_WARN_CONFLICTING_COMPOUND_INDEX_HINT_FOR_TABLE;
}
else if (tab->set_switch(hint_state, hint_type, false))
{
warn_code= ER_WARN_CONFLICTING_INDEX_HINT_FOR_TABLE;
}
if (warn_code != 0)
{
print_warn(pc->thd, warn_code, hint_type, hint_state, &qb_name_sys,
&table_name_sys, nullptr, this);
}
else if (is_compound_hint(hint_type))
{
tab->get_key_hint_bitmap(hint_type)->parsed_hint= this;
}
if (is_empty())
return false;
}
// Key names for a compound hint are first collected into the array:
Mem_root_array<std::pair<Opt_hints_key *,
bool /* whether a new one was created */>>
key_hints(pc->thd->mem_root);
bool is_conflicting= false;
for (const Hint_param_index &index_name : *this)
{
const Lex_ident_sys index_name_sys= index_name.to_ident_sys(pc->thd);
bool new_opt_key_hint_created= false;
Opt_hints_key *key= (Opt_hints_key *)tab->find_by_name(index_name_sys);
if (!key)
{
key= new (pc->thd->mem_root)
Opt_hints_key(index_name_sys, tab, pc->thd->mem_root);
new_opt_key_hint_created= true;
}
if (!is_compound_hint(hint_type))
{
if (key->set_switch(hint_state, hint_type, !set_also_on_table))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_INDEX_HINT_FOR_KEY,
hint_type, hint_state, &qb_name_sys, &table_name_sys,
&index_name_sys, nullptr);
continue;
}
if (new_opt_key_hint_created)
tab->register_child(key);
}
else
{
bool is_specified= tab->is_specified(hint_type) ||
key->is_specified(hint_type);
if (is_specified || is_index_hint_conflicting(tab, key, hint_type))
{
is_conflicting= true;
uint warn_code;
if (is_specified)
{
warn_code= tab->is_specified(hint_type) ?
ER_WARN_CONFLICTING_INDEX_HINT_FOR_TABLE :
ER_WARN_CONFLICTING_INDEX_HINT_FOR_KEY;
}
else
{
warn_code= ER_WARN_CONFLICTING_COMPOUND_INDEX_HINT_FOR_KEY;
}
print_warn(pc->thd, warn_code, hint_type, hint_state,
&qb_name_sys, &table_name_sys, nullptr, this);
break;
}
key_hints.push_back({ key, new_opt_key_hint_created });
}
}
if (is_compound_hint(hint_type) && !is_conflicting)
{
/*
Process key names collected for a compound hint. They have already been
checked for conflicts/duplication above, so there is no need to examine
the `set_switch()` return value
*/
for (size_t i= 0; i < key_hints.size(); i++)
{
std::pair<Opt_hints_key *, bool> key= key_hints.at(i);
key.first->set_switch(hint_state, hint_type, true);
if (key.second)
tab->register_child(key.first);
}
tab->get_key_hint_bitmap(hint_type)->parsed_hint= this;
tab->set_switch(hint_state, hint_type, false);
}
return false;
}
void Parser::Index_level_hint::append_args(THD *thd, String *str) const
{
if (is_empty()) // Empty list of index names, no additional info
return;
bool first_index_name= true;
for (const Hint_param_index &index_name : *this)
{
if (!first_index_name)
str->append(STRING_WITH_LEN(","));
append_identifier(thd, str, &index_name);
first_index_name= false;
}
}
/*
Resolve a parsed query block name hint, i.e. set up proper Opt_hint_*
structures which will be used later during query preparation and optimization.
Return value:
- false: no critical errors, warnings on duplicated hints,
unresolved query block names, etc. are allowed
- true: critical errors detected, break further hints processing
*/
bool Parser::Qb_name_hint::resolve(Parse_context *pc) const
{
Opt_hints_qb *qb= pc->select->opt_hints_qb;
DBUG_ASSERT(qb);
const Lex_ident_sys qb_name_sys= Query_block_name::to_ident_sys(pc->thd);
if (qb->get_name().str || // QB name is already set
qb->get_parent()->find_by_name(qb_name_sys)) // Name is already used
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, QB_NAME_HINT_ENUM, true,
&qb_name_sys, nullptr, nullptr, nullptr);
return false;
}
qb->set_name(qb_name_sys);
return false;
}
void Parser::Semijoin_hint::fill_strategies_map(Opt_hints_qb *qb) const
{
// Loop for hints like SEMIJOIN(firstmatch, dupsweedout)
const Hint_param_opt_sj_strategy_list &hint_param_strategy_list= *this;
for (const Semijoin_strategy &strat : hint_param_strategy_list)
add_strategy_to_map(strat.id(), qb);
// Loop for hints like SEMIJOIN(@qb1 firstmatch, dupsweedout)
const Opt_sj_strategy_list &opt_sj_strategy_list= *this;
for (const Semijoin_strategy &strat : opt_sj_strategy_list)
add_strategy_to_map(strat.id(), qb);
}
void Parser::Semijoin_hint::add_strategy_to_map(TokenID token_id,
Opt_hints_qb *qb) const
{
switch(token_id)
{
case TokenID::keyword_DUPSWEEDOUT:
qb->semijoin_strategies_map |= OPTIMIZER_SWITCH_DUPSWEEDOUT;
break;
case TokenID::keyword_FIRSTMATCH:
qb->semijoin_strategies_map |= OPTIMIZER_SWITCH_FIRSTMATCH;
break;
case TokenID::keyword_LOOSESCAN:
qb->semijoin_strategies_map |= OPTIMIZER_SWITCH_LOOSE_SCAN;
break;
case TokenID::keyword_MATERIALIZATION:
qb->semijoin_strategies_map |= OPTIMIZER_SWITCH_MATERIALIZATION;
break;
default:
DBUG_ASSERT(0);
}
}
/*
Resolve a parsed semijoin hint, i.e. set up proper Opt_hint_* structures
which will be used later during query preparation and optimization.
Return value:
- false: no critical errors, warnings on duplicated hints,
unresolved query block names, etc. are allowed
- true: critical errors detected, break further hints processing
*/
bool Parser::Semijoin_hint::resolve(Parse_context *pc) const
{
const Semijoin_hint_type &semijoin_hint_type= *this;
bool hint_state; // true - SEMIJOIN(), false - NO_SEMIJOIN()
if (semijoin_hint_type.id() == TokenID::keyword_SEMIJOIN)
hint_state= true;
else
hint_state= false;
Opt_hints_qb *qb;
if (const At_query_block_name_opt_strategy_list &
at_query_block_name_opt_strategy_list __attribute__((unused)) = *this)
{
/*
This is @ query_block_name opt_strategy_list,
e.g. SEMIJOIN(@qb1) or SEMIJOIN(@qb1 firstmatch, loosescan)
*/
const Lex_ident_sys qb_name= Query_block_name::to_ident_sys(pc->thd);
qb= resolve_for_qb_name(pc, hint_state, &qb_name);
}
else
{
// This is opt_strategy_list, e.g. SEMIJOIN(loosescan, dupsweedout)
Lex_ident_sys empty_qb_name= Lex_ident_sys();
qb= resolve_for_qb_name(pc, hint_state, &empty_qb_name);
}
if (qb)
qb->semijoin_hint= this;
return false;
}
/*
Helper function to be called by Semijoin_hint::resolve().
Return value:
- pointer to Opt_hints_qb if the hint was resolved successfully
- NULL if the hint was ignored
*/
Opt_hints_qb* Parser::Semijoin_hint::
resolve_for_qb_name(Parse_context *pc, bool hint_state,
const Lex_ident_sys *qb_name) const
{
Opt_hints_qb *qb= find_qb_hints(pc, *qb_name, SEMIJOIN_HINT_ENUM, hint_state);
if (!qb)
return nullptr;
if (qb->subquery_hint)
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, SEMIJOIN_HINT_ENUM,
hint_state, qb_name, nullptr, nullptr, this);
return nullptr;
}
if (qb->set_switch(hint_state, SEMIJOIN_HINT_ENUM, false))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, SEMIJOIN_HINT_ENUM,
hint_state, qb_name, nullptr, nullptr, this);
return nullptr;
}
fill_strategies_map(qb);
return qb;
}
void Parser::Semijoin_hint::append_args(THD *thd, String *str) const
{
// Loop for hints without query block name, e.g. SEMIJOIN(firstmatch, dupsweedout)
const Hint_param_opt_sj_strategy_list &hint_param_strategy_list= *this;
uint32 len_before= str->length();
for (const Semijoin_strategy &strat : hint_param_strategy_list)
{
if (str->length() > len_before)
str->append(STRING_WITH_LEN(", "));
append_strategy_name(strat.id(), str);
}
// Loop for hints with query block name, e.g. SEMIJOIN(@qb1 firstmatch, dupsweedout)
const Opt_sj_strategy_list &opt_sj_strategy_list= *this;
for (const Semijoin_strategy &strat : opt_sj_strategy_list)
{
if (str->length() > len_before)
str->append(STRING_WITH_LEN(", "));
append_strategy_name(strat.id(), str);
}
}
void Parser::Semijoin_hint::
append_strategy_name(TokenID token_id, String *str) const
{
switch(token_id)
{
case TokenID::keyword_DUPSWEEDOUT:
str->append(STRING_WITH_LEN("DUPSWEEDOUT"));
break;
case TokenID::keyword_FIRSTMATCH:
str->append(STRING_WITH_LEN("FIRSTMATCH"));
break;
case TokenID::keyword_LOOSESCAN:
str->append(STRING_WITH_LEN("LOOSESCAN"));
break;
case TokenID::keyword_MATERIALIZATION:
str->append(STRING_WITH_LEN("MATERIALIZATION"));
break;
default:
DBUG_ASSERT(0);
}
}
/*
Resolve a parsed subquery hint, i.e. set up proper Opt_hint_* structures
which will be used later during query preparation and optimization.
Return value:
- false: no critical errors, warnings on duplicated hints,
unresolved query block names, etc. are allowed
- true: critical errors detected, break further hints processing
*/
bool Parser::Subquery_hint::resolve(Parse_context *pc) const
{
Opt_hints_qb *qb;
if (const At_query_block_name_subquery_strategy &
at_query_block_name_subquery_strategy= *this)
{
/*
This is @ query_block_name subquery_strategy,
e.g. SUBQUERY(@qb1 INTOEXISTS)
*/
const Lex_ident_sys qb_name= Query_block_name::to_ident_sys(pc->thd);
const Subquery_strategy &strat= at_query_block_name_subquery_strategy;
qb= resolve_for_qb_name(pc, strat.id(), &qb_name);
}
else
{
// This is subquery_strategy, e.g. SUBQUERY(MATERIALIZATION)
Lex_ident_sys empty_qb_name= Lex_ident_sys();
const Hint_param_subquery_strategy &strat= *this;
qb= resolve_for_qb_name(pc, strat.id(), &empty_qb_name);
}
if (qb)
qb->subquery_hint= this;
return false;
}
/*
Helper function to be called by Subquery_hint::resolve().
Return value:
- pointer to Opt_hints_qb if the hint was resolved successfully
- NULL if the hint was ignored
*/
Opt_hints_qb* Parser::Subquery_hint::
resolve_for_qb_name(Parse_context *pc, TokenID token_id,
const Lex_ident_sys *qb_name) const
{
Opt_hints_qb *qb= find_qb_hints(pc, *qb_name, SUBQUERY_HINT_ENUM, true);
if (!qb)
return nullptr;
if (qb->semijoin_hint)
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, SUBQUERY_HINT_ENUM,
true, qb_name, nullptr, nullptr, this);
return nullptr;
}
if (qb->set_switch(true, SUBQUERY_HINT_ENUM, false))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, SUBQUERY_HINT_ENUM,
true, qb_name, nullptr, nullptr, this);
return nullptr;
}
set_subquery_strategy(token_id, qb);
return qb;
}
void Parser::Subquery_hint::set_subquery_strategy(TokenID token_id,
Opt_hints_qb *qb) const
{
switch(token_id)
{
case TokenID::keyword_INTOEXISTS:
qb->subquery_strategy= SUBS_IN_TO_EXISTS;
break;
case TokenID::keyword_MATERIALIZATION:
qb->subquery_strategy= SUBS_MATERIALIZATION;
break;
default:
DBUG_ASSERT(0);
}
}
void Parser::Subquery_hint::append_args(THD *thd, String *str) const
{
TokenID token_id;
if (const At_query_block_name_subquery_strategy &
at_query_block_name_subquery_strategy= *this)
{
const Subquery_strategy &strat= at_query_block_name_subquery_strategy;
token_id= strat.id();
}
else
{
const Hint_param_subquery_strategy& hint_param_strat= *this;
token_id= hint_param_strat.id();
}
switch(token_id)
{
case TokenID::keyword_INTOEXISTS:
str->append(STRING_WITH_LEN("INTOEXISTS"));
break;
case TokenID::keyword_MATERIALIZATION:
str->append(STRING_WITH_LEN("MATERIALIZATION"));
break;
default:
DBUG_ASSERT(0);
}
}
/*
This is the first step of MAX_EXECUTION_TIME() hint resolution. It is invoked
during the parsing phase, but at this stage some essential information is
not yet available, preventing a full validation of the hint.
Particularly, the type of SQL command, mark of a stored procedure execution
or whether SELECT_LEX is not top-level (i.e., a subquery) are not yet set.
However, some basic checks like the numeric argument validation or hint
duplication check can still be performed.
The second step of hint validation is performed during the JOIN preparation
phase, within Opt_hints_global::resolve(). By this point, all necessary
information is up-to-date, allowing the hint to be fully resolved
*/
bool Parser::Max_execution_time_hint::resolve(Parse_context *pc) const
{
const Unsigned_Number& hint_arg= *this;
const ULonglong_null time_ms= hint_arg.get_ulonglong();
if (time_ms.is_null() || time_ms.value() == 0 || time_ms.value() > INT_MAX32)
{
print_warn(pc->thd, ER_BAD_OPTION_VALUE, MAX_EXEC_TIME_HINT_ENUM,
true, NULL, NULL, NULL, this);
return false;
}
Opt_hints_global *global_hint= get_global_hints(pc);
if (global_hint->is_specified(MAX_EXEC_TIME_HINT_ENUM))
{
// Hint duplication: /*+ MAX_EXECUTION_TIME ... MAX_EXECUTION_TIME */
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, MAX_EXEC_TIME_HINT_ENUM, true,
NULL, NULL, NULL, this);
return false;
}
global_hint->set_switch(true, MAX_EXEC_TIME_HINT_ENUM, false);
global_hint->max_exec_time_hint= this;
global_hint->max_exec_time_select_lex= pc->select;
return false;
}
void Parser::Join_order_hint::append_args(THD *thd, String *str) const
{
bool first_table_name= true;
for (const Parser::Table_name_and_Qb& tbl : table_names)
{
if (!first_table_name)
str->append(STRING_WITH_LEN(","));
append_table_name(thd, str, tbl.table_name, tbl.qb_name);
first_table_name= false;
}
}
/*
Resolve a parsed join order hint, i.e. set up proper Opt_hint_* structures
which will be used later during query preparation and optimization.
Return value:
- false: no critical errors, warnings on duplicated hints,
unresolved query block names, etc. are allowed
- true: critical errors detected, break further hints processing
*/
bool Parser::Join_order_hint::resolve(Parse_context *pc)
{
const Join_order_hint_type &join_order_hint_type= *this;
switch (join_order_hint_type.id())
{
case TokenID::keyword_JOIN_FIXED_ORDER:
hint_type= JOIN_FIXED_ORDER_HINT_ENUM;
break;
case TokenID::keyword_JOIN_ORDER:
hint_type= JOIN_ORDER_HINT_ENUM;
break;
case TokenID::keyword_JOIN_PREFIX:
hint_type= JOIN_PREFIX_HINT_ENUM;
break;
case TokenID::keyword_JOIN_SUFFIX:
hint_type= JOIN_SUFFIX_HINT_ENUM;
break;
default:
DBUG_ASSERT(0);
return true;
}
Opt_hints_qb *qb= nullptr;
Lex_ident_sys qb_name;
if (const At_query_block_name_opt_table_name_list &at_qb_tab_list= *this)
{
// this is @ query_block_name opt_table_name_list
qb_name= Query_block_name::to_ident_sys(pc->thd);
qb= find_qb_hints(pc, qb_name, hint_type, true);
// Compose `tables_names` list for warnings and final hints resolving
const Opt_table_name_list &opt_table_name_list= at_qb_tab_list;
for (const Table_name &table : opt_table_name_list)
{
Parser::Table_name_and_Qb *tbl_qb= new (pc->mem_root)
Parser::Table_name_and_Qb(table.to_ident_sys(pc->thd), Lex_ident_sys());
if (!tbl_qb)
return true;
table_names.push_back(tbl_qb, pc->mem_root);
}
}
else
{
// this is opt_hint_param_table_list, query block name is not specified
qb= find_qb_hints(pc, Lex_ident_sys(), hint_type, true);
const Opt_hint_param_table_list &opt_hint_param_table_list= *this;
for (const Hint_param_table &table : opt_hint_param_table_list)
{
// e.g. JOIN_ORDER(t1@qb1, t2@qb2, t3)
Parser::Table_name_and_Qb *tbl_qb=
new (pc->mem_root) Parser::Table_name_and_Qb(
table.Table_name::to_ident_sys(pc->thd),
table.Query_block_name::to_ident_sys(pc->thd));
table_names.push_back(tbl_qb, pc->mem_root);
}
}
if (qb == nullptr)
return false;
if ((hint_type != JOIN_FIXED_ORDER_HINT_ENUM && table_names.is_empty()) ||
(hint_type == JOIN_FIXED_ORDER_HINT_ENUM && !table_names.is_empty()))
{
/*
Skipping table name(s) only allowed and required for the
JOIN_FIXED_ORDER hint and is not allowed for other hint types
*/
print_warn(pc->thd, ER_WARN_MALFORMED_HINT, hint_type, true,
&qb_name, nullptr, nullptr, this);
return false;
}
if (hint_type == JOIN_FIXED_ORDER_HINT_ENUM)
{
/*
This is JOIN_ORDER_FIXED() or JOIN_ORDER_FIXED(@qb1)
There can be only one JOIN_ORDER_FIXED hint in a query block,
other hints are not allowed in this case
*/
if (qb->has_join_order_hints()|| qb->join_fixed_order)
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, true,
&qb_name, nullptr, nullptr, this);
return false;
}
qb->join_fixed_order= this;
qb->set_switch(true, hint_type, false);
pc->select->options |= SELECT_STRAIGHT_JOIN;
return false;
}
// Finished with processing of JOIN_FIXED_ORDER()
DBUG_ASSERT(hint_type != JOIN_FIXED_ORDER_HINT_ENUM);
/*
Hints except JOIN_ORDER() must not duplicate. If there is JOIN_ORDER_FIXED()
already, then other hints are not allowed for this query block
*/
if ((qb->get_switch(hint_type) && hint_type != JOIN_ORDER_HINT_ENUM) ||
qb->join_fixed_order)
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, true,
&qb_name, nullptr, nullptr, this);
return false;
}
switch(hint_type)
{
case JOIN_PREFIX_HINT_ENUM:
if (qb->join_prefix || qb->add_join_order_hint(this))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, true,
&qb_name, nullptr, nullptr, this);
return false;
}
qb->join_prefix= this;
qb->set_switch(true, JOIN_PREFIX_HINT_ENUM, false);
break;
case JOIN_SUFFIX_HINT_ENUM:
if (qb->join_suffix || qb->add_join_order_hint(this))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, true,
&qb_name, nullptr, nullptr, this);
return false;
}
qb->join_suffix= this;
qb->set_switch(true, JOIN_SUFFIX_HINT_ENUM, false);
break;
case JOIN_ORDER_HINT_ENUM:
// Multiple JOIN_ORDER() hints are allowed
if (qb->add_join_order_hint(this))
{
print_warn(pc->thd, ER_WARN_CONFLICTING_HINT, hint_type, true,
&qb_name, nullptr, nullptr, this);
return false;
}
qb->set_switch(true, JOIN_ORDER_HINT_ENUM, false);
break;
default:
DBUG_ASSERT(0);
}
return false;
}
void Parser::Max_execution_time_hint::append_args(THD *thd, String *str) const
{
const Unsigned_Number& hint_arg= *this;
str->append(ErrConvString(hint_arg.str, hint_arg.length,
&my_charset_latin1).lex_cstring());
}
ulonglong Parser::Max_execution_time_hint::get_milliseconds() const
{
const Unsigned_Number& hint_arg= *this;
return hint_arg.get_ulonglong().value();
}
bool Parser::Hint_list::resolve(Parse_context *pc) const
{
if (pc->thd->lex->create_view)
{
// we're creating or modifying a view, hints are not allowed here
push_warning(pc->thd, Sql_condition::WARN_LEVEL_WARN,
ER_HINTS_INSIDE_VIEWS_NOT_SUPPORTED,
ER_THD(pc->thd, ER_HINTS_INSIDE_VIEWS_NOT_SUPPORTED));
return false;
}
if (!get_qb_hints(pc))
return true;
for (Hint_list::iterator li= this->begin(); li != this->end(); ++li)
{
Parser::Hint &hint= *li;
if (const Table_level_hint &table_hint= hint)
{
if (table_hint.resolve(pc))
return true;
}
else if (const Index_level_hint &index_hint= hint)
{
if (index_hint.resolve(pc))
return true;
}
else if (const Qb_name_hint &qb_hint= hint)
{
if (qb_hint.resolve(pc))
return true;
}
else if (const Max_execution_time_hint &max_hint= hint)
{
if (max_hint.resolve(pc))
return true;
}
else if (const Semijoin_hint &sj_hint= hint)
{
if (sj_hint.resolve(pc))
return true;
}
else if (const Subquery_hint &subq_hint= hint)
{
if (subq_hint.resolve(pc))
return true;
}
else if (Join_order_hint &join_order_hint= hint)
{
if (join_order_hint.resolve(pc))
return true;
}
else {
DBUG_ASSERT(0);
}
}
return false;
}
Reference in a new issue View git blame Copy permalink