2014-02-26 19:11:54 +01:00
|
|
|
/*****************************************************************************
|
|
|
|
|
2016-06-21 14:21:03 +02:00
|
|
|
Copyright (c) 1996, 2016, Oracle and/or its affiliates. All Rights Reserved.
|
2018-02-15 10:22:03 +02:00
|
|
|
Copyright (c) 2017, 2018, MariaDB Corporation.
|
2014-02-26 19:11:54 +01:00
|
|
|
|
|
|
|
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 Street, Suite 500, Boston, MA 02110-1335 USA
|
|
|
|
|
|
|
|
*****************************************************************************/
|
|
|
|
|
|
|
|
/**************************************************//**
|
|
|
|
@file include/row0upd.h
|
|
|
|
Update of a row
|
|
|
|
|
|
|
|
Created 12/27/1996 Heikki Tuuri
|
|
|
|
*******************************************************/
|
|
|
|
|
|
|
|
#ifndef row0upd_h
|
|
|
|
#define row0upd_h
|
|
|
|
|
|
|
|
#include "data0data.h"
|
|
|
|
#include "row0types.h"
|
|
|
|
#include "btr0types.h"
|
|
|
|
#include "dict0types.h"
|
|
|
|
#include "trx0types.h"
|
2016-12-30 15:04:10 +02:00
|
|
|
#include "btr0pcur.h"
|
|
|
|
#include "que0types.h"
|
|
|
|
#include "pars0types.h"
|
2014-02-26 19:11:54 +01:00
|
|
|
|
|
|
|
/*********************************************************************//**
|
|
|
|
Creates an update vector object.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return own: update vector object */
|
2014-02-26 19:11:54 +01:00
|
|
|
UNIV_INLINE
|
|
|
|
upd_t*
|
|
|
|
upd_create(
|
|
|
|
/*=======*/
|
|
|
|
ulint n, /*!< in: number of fields */
|
|
|
|
mem_heap_t* heap); /*!< in: heap from which memory allocated */
|
|
|
|
/*********************************************************************//**
|
|
|
|
Returns the number of fields in the update vector == number of columns
|
|
|
|
to be updated by an update vector.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return number of fields */
|
2014-02-26 19:11:54 +01:00
|
|
|
UNIV_INLINE
|
|
|
|
ulint
|
|
|
|
upd_get_n_fields(
|
|
|
|
/*=============*/
|
|
|
|
const upd_t* update); /*!< in: update vector */
|
|
|
|
#ifdef UNIV_DEBUG
|
|
|
|
/*********************************************************************//**
|
|
|
|
Returns the nth field of an update vector.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return update vector field */
|
2014-02-26 19:11:54 +01:00
|
|
|
UNIV_INLINE
|
|
|
|
upd_field_t*
|
|
|
|
upd_get_nth_field(
|
|
|
|
/*==============*/
|
|
|
|
const upd_t* update, /*!< in: update vector */
|
|
|
|
ulint n); /*!< in: field position in update vector */
|
|
|
|
#else
|
|
|
|
# define upd_get_nth_field(update, n) ((update)->fields + (n))
|
|
|
|
#endif
|
2016-12-30 15:04:10 +02:00
|
|
|
|
2014-02-26 19:11:54 +01:00
|
|
|
/*********************************************************************//**
|
|
|
|
Sets an index field number to be updated by an update vector field. */
|
|
|
|
UNIV_INLINE
|
|
|
|
void
|
|
|
|
upd_field_set_field_no(
|
|
|
|
/*===================*/
|
|
|
|
upd_field_t* upd_field, /*!< in: update vector field */
|
|
|
|
ulint field_no, /*!< in: field number in a clustered
|
|
|
|
index */
|
2017-09-29 16:19:28 +03:00
|
|
|
dict_index_t* index);
|
2016-08-12 11:17:45 +03:00
|
|
|
|
|
|
|
/** set field number to a update vector field, marks this field is updated
|
|
|
|
@param[in,out] upd_field update vector field
|
|
|
|
@param[in] field_no virtual column sequence num
|
|
|
|
@param[in] index index */
|
|
|
|
UNIV_INLINE
|
|
|
|
void
|
|
|
|
upd_field_set_v_field_no(
|
|
|
|
upd_field_t* upd_field,
|
|
|
|
ulint field_no,
|
|
|
|
dict_index_t* index);
|
2014-02-26 19:11:54 +01:00
|
|
|
/*********************************************************************//**
|
|
|
|
Returns a field of an update vector by field_no.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return update vector field, or NULL */
|
2014-02-26 19:11:54 +01:00
|
|
|
UNIV_INLINE
|
|
|
|
const upd_field_t*
|
|
|
|
upd_get_field_by_field_no(
|
|
|
|
/*======================*/
|
|
|
|
const upd_t* update, /*!< in: update vector */
|
2016-08-12 11:17:45 +03:00
|
|
|
ulint no, /*!< in: field_no */
|
2016-09-06 09:43:16 +03:00
|
|
|
bool is_virtual) /*!< in: if it is a virtual column */
|
|
|
|
MY_ATTRIBUTE((warn_unused_result));
|
2014-02-26 19:11:54 +01:00
|
|
|
/*********************************************************************//**
|
|
|
|
Writes into the redo log the values of trx id and roll ptr and enough info
|
|
|
|
to determine their positions within a clustered index record.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return new pointer to mlog */
|
2014-02-26 19:11:54 +01:00
|
|
|
byte*
|
|
|
|
row_upd_write_sys_vals_to_log(
|
|
|
|
/*==========================*/
|
|
|
|
dict_index_t* index, /*!< in: clustered index */
|
|
|
|
trx_id_t trx_id, /*!< in: transaction id */
|
|
|
|
roll_ptr_t roll_ptr,/*!< in: roll ptr of the undo log record */
|
|
|
|
byte* log_ptr,/*!< pointer to a buffer of size > 20 opened
|
|
|
|
in mlog */
|
|
|
|
mtr_t* mtr); /*!< in: mtr */
|
|
|
|
/*********************************************************************//**
|
|
|
|
Updates the trx id and roll ptr field in a clustered index record when
|
|
|
|
a row is updated or marked deleted. */
|
|
|
|
UNIV_INLINE
|
|
|
|
void
|
|
|
|
row_upd_rec_sys_fields(
|
|
|
|
/*===================*/
|
|
|
|
rec_t* rec, /*!< in/out: record */
|
|
|
|
page_zip_des_t* page_zip,/*!< in/out: compressed page whose
|
|
|
|
uncompressed part will be updated, or NULL */
|
|
|
|
dict_index_t* index, /*!< in: clustered index */
|
|
|
|
const ulint* offsets,/*!< in: rec_get_offsets(rec, index) */
|
|
|
|
const trx_t* trx, /*!< in: transaction */
|
2018-02-08 12:51:19 +02:00
|
|
|
roll_ptr_t roll_ptr);/*!< in: DB_ROLL_PTR to the undo log */
|
2014-02-26 19:11:54 +01:00
|
|
|
/*********************************************************************//**
|
|
|
|
Sets the trx id or roll ptr field of a clustered index entry. */
|
|
|
|
void
|
|
|
|
row_upd_index_entry_sys_field(
|
|
|
|
/*==========================*/
|
|
|
|
dtuple_t* entry, /*!< in/out: index entry, where the memory
|
|
|
|
buffers for sys fields are already allocated:
|
|
|
|
the function just copies the new values to
|
|
|
|
them */
|
|
|
|
dict_index_t* index, /*!< in: clustered index */
|
|
|
|
ulint type, /*!< in: DATA_TRX_ID or DATA_ROLL_PTR */
|
|
|
|
ib_uint64_t val); /*!< in: value to write */
|
|
|
|
/*********************************************************************//**
|
|
|
|
Creates an update node for a query graph.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return own: update node */
|
2014-02-26 19:11:54 +01:00
|
|
|
upd_node_t*
|
|
|
|
upd_node_create(
|
|
|
|
/*============*/
|
|
|
|
mem_heap_t* heap); /*!< in: mem heap where created */
|
|
|
|
/***********************************************************//**
|
|
|
|
Writes to the redo log the new values of the fields occurring in the index. */
|
|
|
|
void
|
|
|
|
row_upd_index_write_log(
|
|
|
|
/*====================*/
|
|
|
|
const upd_t* update, /*!< in: update vector */
|
|
|
|
byte* log_ptr,/*!< in: pointer to mlog buffer: must
|
|
|
|
contain at least MLOG_BUF_MARGIN bytes
|
|
|
|
of free space; the buffer is closed
|
|
|
|
within this function */
|
|
|
|
mtr_t* mtr); /*!< in: mtr into whose log to write */
|
|
|
|
/***********************************************************//**
|
|
|
|
Returns TRUE if row update changes size of some field in index or if some
|
|
|
|
field to be updated is stored externally in rec or update.
|
|
|
|
@return TRUE if the update changes the size of some field in index or
|
|
|
|
the field is external in rec or update */
|
|
|
|
ibool
|
|
|
|
row_upd_changes_field_size_or_external(
|
|
|
|
/*===================================*/
|
|
|
|
dict_index_t* index, /*!< in: index */
|
|
|
|
const ulint* offsets,/*!< in: rec_get_offsets(rec, index) */
|
|
|
|
const upd_t* update);/*!< in: update vector */
|
|
|
|
/***********************************************************//**
|
|
|
|
Returns true if row update contains disowned external fields.
|
|
|
|
@return true if the update contains disowned external fields. */
|
|
|
|
bool
|
|
|
|
row_upd_changes_disowned_external(
|
|
|
|
/*==============================*/
|
|
|
|
const upd_t* update) /*!< in: update vector */
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((nonnull, warn_unused_result));
|
2014-02-26 19:11:54 +01:00
|
|
|
/***********************************************************//**
|
|
|
|
Replaces the new column values stored in the update vector to the
|
|
|
|
record given. No field size changes are allowed. This function is
|
|
|
|
usually invoked on a clustered index. The only use case for a
|
|
|
|
secondary index is row_ins_sec_index_entry_by_modify() or its
|
|
|
|
counterpart in ibuf_insert_to_index_page(). */
|
|
|
|
void
|
|
|
|
row_upd_rec_in_place(
|
|
|
|
/*=================*/
|
|
|
|
rec_t* rec, /*!< in/out: record where replaced */
|
|
|
|
dict_index_t* index, /*!< in: the index the record belongs to */
|
|
|
|
const ulint* offsets,/*!< in: array returned by rec_get_offsets() */
|
|
|
|
const upd_t* update, /*!< in: update vector */
|
|
|
|
page_zip_des_t* page_zip);/*!< in: compressed page with enough space
|
|
|
|
available, or NULL */
|
2016-12-30 15:04:10 +02:00
|
|
|
|
2014-02-26 19:11:54 +01:00
|
|
|
/***************************************************************//**
|
|
|
|
Builds an update vector from those fields which in a secondary index entry
|
|
|
|
differ from a record that has the equal ordering fields. NOTE: we compare
|
|
|
|
the fields as binary strings!
|
2016-08-12 11:17:45 +03:00
|
|
|
@return own: update vector of differing fields */
|
2014-02-26 19:11:54 +01:00
|
|
|
upd_t*
|
|
|
|
row_upd_build_sec_rec_difference_binary(
|
|
|
|
/*====================================*/
|
|
|
|
const rec_t* rec, /*!< in: secondary index record */
|
|
|
|
dict_index_t* index, /*!< in: index */
|
|
|
|
const ulint* offsets,/*!< in: rec_get_offsets(rec, index) */
|
|
|
|
const dtuple_t* entry, /*!< in: entry to insert */
|
|
|
|
mem_heap_t* heap) /*!< in: memory heap from which allocated */
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((warn_unused_result, nonnull));
|
2016-09-06 09:43:16 +03:00
|
|
|
/** Builds an update vector from those fields, excluding the roll ptr and
|
2014-02-26 19:11:54 +01:00
|
|
|
trx id fields, which in an index entry differ from a record that has
|
|
|
|
the equal ordering fields. NOTE: we compare the fields as binary strings!
|
2016-09-06 09:43:16 +03:00
|
|
|
@param[in] index clustered index
|
|
|
|
@param[in] entry clustered index entry to insert
|
|
|
|
@param[in] rec clustered index record
|
|
|
|
@param[in] offsets rec_get_offsets(rec,index), or NULL
|
|
|
|
@param[in] no_sys skip the system columns
|
|
|
|
DB_TRX_ID and DB_ROLL_PTR
|
|
|
|
@param[in] trx transaction (for diagnostics),
|
|
|
|
or NULL
|
|
|
|
@param[in] heap memory heap from which allocated
|
|
|
|
@param[in,out] mysql_table NULL, or mysql table object when
|
|
|
|
user thread invokes dml
|
2014-02-26 19:11:54 +01:00
|
|
|
@return own: update vector of differing fields, excluding roll ptr and
|
|
|
|
trx id */
|
2016-08-12 11:17:45 +03:00
|
|
|
upd_t*
|
2014-02-26 19:11:54 +01:00
|
|
|
row_upd_build_difference_binary(
|
2016-09-06 09:43:16 +03:00
|
|
|
dict_index_t* index,
|
|
|
|
const dtuple_t* entry,
|
|
|
|
const rec_t* rec,
|
|
|
|
const ulint* offsets,
|
|
|
|
bool no_sys,
|
|
|
|
trx_t* trx,
|
|
|
|
mem_heap_t* heap,
|
|
|
|
TABLE* mysql_table)
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((nonnull(1,2,3,7), warn_unused_result));
|
MDEV-11369 Instant ADD COLUMN for InnoDB
For InnoDB tables, adding, dropping and reordering columns has
required a rebuild of the table and all its indexes. Since MySQL 5.6
(and MariaDB 10.0) this has been supported online (LOCK=NONE), allowing
concurrent modification of the tables.
This work revises the InnoDB ROW_FORMAT=REDUNDANT, ROW_FORMAT=COMPACT
and ROW_FORMAT=DYNAMIC so that columns can be appended instantaneously,
with only minor changes performed to the table structure. The counter
innodb_instant_alter_column in INFORMATION_SCHEMA.GLOBAL_STATUS
is incremented whenever a table rebuild operation is converted into
an instant ADD COLUMN operation.
ROW_FORMAT=COMPRESSED tables will not support instant ADD COLUMN.
Some usability limitations will be addressed in subsequent work:
MDEV-13134 Introduce ALTER TABLE attributes ALGORITHM=NOCOPY
and ALGORITHM=INSTANT
MDEV-14016 Allow instant ADD COLUMN, ADD INDEX, LOCK=NONE
The format of the clustered index (PRIMARY KEY) is changed as follows:
(1) The FIL_PAGE_TYPE of the root page will be FIL_PAGE_TYPE_INSTANT,
and a new field PAGE_INSTANT will contain the original number of fields
in the clustered index ('core' fields).
If instant ADD COLUMN has not been used or the table becomes empty,
or the very first instant ADD COLUMN operation is rolled back,
the fields PAGE_INSTANT and FIL_PAGE_TYPE will be reset
to 0 and FIL_PAGE_INDEX.
(2) A special 'default row' record is inserted into the leftmost leaf,
between the page infimum and the first user record. This record is
distinguished by the REC_INFO_MIN_REC_FLAG, and it is otherwise in the
same format as records that contain values for the instantly added
columns. This 'default row' always has the same number of fields as
the clustered index according to the table definition. The values of
'core' fields are to be ignored. For other fields, the 'default row'
will contain the default values as they were during the ALTER TABLE
statement. (If the column default values are changed later, those
values will only be stored in the .frm file. The 'default row' will
contain the original evaluated values, which must be the same for
every row.) The 'default row' must be completely hidden from
higher-level access routines. Assertions have been added to ensure
that no 'default row' is ever present in the adaptive hash index
or in locked records. The 'default row' is never delete-marked.
(3) In clustered index leaf page records, the number of fields must
reside between the number of 'core' fields (dict_index_t::n_core_fields
introduced in this work) and dict_index_t::n_fields. If the number
of fields is less than dict_index_t::n_fields, the missing fields
are replaced with the column value of the 'default row'.
Note: The number of fields in the record may shrink if some of the
last instantly added columns are updated to the value that is
in the 'default row'. The function btr_cur_trim() implements this
'compression' on update and rollback; dtuple::trim() implements it
on insert.
(4) In ROW_FORMAT=COMPACT and ROW_FORMAT=DYNAMIC records, the new
status value REC_STATUS_COLUMNS_ADDED will indicate the presence of
a new record header that will encode n_fields-n_core_fields-1 in
1 or 2 bytes. (In ROW_FORMAT=REDUNDANT records, the record header
always explicitly encodes the number of fields.)
We introduce the undo log record type TRX_UNDO_INSERT_DEFAULT for
covering the insert of the 'default row' record when instant ADD COLUMN
is used for the first time. Subsequent instant ADD COLUMN can use
TRX_UNDO_UPD_EXIST_REC.
This is joint work with Vin Chen (陈福荣) from Tencent. The design
that was discussed in April 2017 would not have allowed import or
export of data files, because instead of the 'default row' it would
have introduced a data dictionary table. The test
rpl.rpl_alter_instant is exactly as contributed in pull request #408.
The test innodb.instant_alter is based on a contributed test.
The redo log record format changes for ROW_FORMAT=DYNAMIC and
ROW_FORMAT=COMPACT are as contributed. (With this change present,
crash recovery from MariaDB 10.3.1 will fail in spectacular ways!)
Also the semantics of higher-level redo log records that modify the
PAGE_INSTANT field is changed. The redo log format version identifier
was already changed to LOG_HEADER_FORMAT_CURRENT=103 in MariaDB 10.3.1.
Everything else has been rewritten by me. Thanks to Elena Stepanova,
the code has been tested extensively.
When rolling back an instant ADD COLUMN operation, we must empty the
PAGE_FREE list after deleting or shortening the 'default row' record,
by calling either btr_page_empty() or btr_page_reorganize(). We must
know the size of each entry in the PAGE_FREE list. If rollback left a
freed copy of the 'default row' in the PAGE_FREE list, we would be
unable to determine its size (if it is in ROW_FORMAT=COMPACT or
ROW_FORMAT=DYNAMIC) because it would contain more fields than the
rolled-back definition of the clustered index.
UNIV_SQL_DEFAULT: A new special constant that designates an instantly
added column that is not present in the clustered index record.
len_is_stored(): Check if a length is an actual length. There are
two magic length values: UNIV_SQL_DEFAULT, UNIV_SQL_NULL.
dict_col_t::def_val: The 'default row' value of the column. If the
column is not added instantly, def_val.len will be UNIV_SQL_DEFAULT.
dict_col_t: Add the accessors is_virtual(), is_nullable(), is_instant(),
instant_value().
dict_col_t::remove_instant(): Remove the 'instant ADD' status of
a column.
dict_col_t::name(const dict_table_t& table): Replaces
dict_table_get_col_name().
dict_index_t::n_core_fields: The original number of fields.
For secondary indexes and if instant ADD COLUMN has not been used,
this will be equal to dict_index_t::n_fields.
dict_index_t::n_core_null_bytes: Number of bytes needed to
represent the null flags; usually equal to UT_BITS_IN_BYTES(n_nullable).
dict_index_t::NO_CORE_NULL_BYTES: Magic value signalling that
n_core_null_bytes was not initialized yet from the clustered index
root page.
dict_index_t: Add the accessors is_instant(), is_clust(),
get_n_nullable(), instant_field_value().
dict_index_t::instant_add_field(): Adjust clustered index metadata
for instant ADD COLUMN.
dict_index_t::remove_instant(): Remove the 'instant ADD' status
of a clustered index when the table becomes empty, or the very first
instant ADD COLUMN operation is rolled back.
dict_table_t: Add the accessors is_instant(), is_temporary(),
supports_instant().
dict_table_t::instant_add_column(): Adjust metadata for
instant ADD COLUMN.
dict_table_t::rollback_instant(): Adjust metadata on the rollback
of instant ADD COLUMN.
prepare_inplace_alter_table_dict(): First create the ctx->new_table,
and only then decide if the table really needs to be rebuilt.
We must split the creation of table or index metadata from the
creation of the dictionary table records and the creation of
the data. In this way, we can transform a table-rebuilding operation
into an instant ADD COLUMN operation. Dictionary objects will only
be added to cache when table rebuilding or index creation is needed.
The ctx->instant_table will never be added to cache.
dict_table_t::add_to_cache(): Modified and renamed from
dict_table_add_to_cache(). Do not modify the table metadata.
Let the callers invoke dict_table_add_system_columns() and if needed,
set can_be_evicted.
dict_create_sys_tables_tuple(), dict_create_table_step(): Omit the
system columns (which will now exist in the dict_table_t object
already at this point).
dict_create_table_step(): Expect the callers to invoke
dict_table_add_system_columns().
pars_create_table(): Before creating the table creation execution
graph, invoke dict_table_add_system_columns().
row_create_table_for_mysql(): Expect all callers to invoke
dict_table_add_system_columns().
create_index_dict(): Replaces row_merge_create_index_graph().
innodb_update_n_cols(): Renamed from innobase_update_n_virtual().
Call my_error() if an error occurs.
btr_cur_instant_init(), btr_cur_instant_init_low(),
btr_cur_instant_root_init():
Load additional metadata from the clustered index and set
dict_index_t::n_core_null_bytes. This is invoked
when table metadata is first loaded into the data dictionary.
dict_boot(): Initialize n_core_null_bytes for the four hard-coded
dictionary tables.
dict_create_index_step(): Initialize n_core_null_bytes. This is
executed as part of CREATE TABLE.
dict_index_build_internal_clust(): Initialize n_core_null_bytes to
NO_CORE_NULL_BYTES if table->supports_instant().
row_create_index_for_mysql(): Initialize n_core_null_bytes for
CREATE TEMPORARY TABLE.
commit_cache_norebuild(): Call the code to rename or enlarge columns
in the cache only if instant ADD COLUMN is not being used.
(Instant ADD COLUMN would copy all column metadata from
instant_table to old_table, including the names and lengths.)
PAGE_INSTANT: A new 13-bit field for storing dict_index_t::n_core_fields.
This is repurposing the 16-bit field PAGE_DIRECTION, of which only the
least significant 3 bits were used. The original byte containing
PAGE_DIRECTION will be accessible via the new constant PAGE_DIRECTION_B.
page_get_instant(), page_set_instant(): Accessors for the PAGE_INSTANT.
page_ptr_get_direction(), page_get_direction(),
page_ptr_set_direction(): Accessors for PAGE_DIRECTION.
page_direction_reset(): Reset PAGE_DIRECTION, PAGE_N_DIRECTION.
page_direction_increment(): Increment PAGE_N_DIRECTION
and set PAGE_DIRECTION.
rec_get_offsets(): Use the 'leaf' parameter for non-debug purposes,
and assume that heap_no is always set.
Initialize all dict_index_t::n_fields for ROW_FORMAT=REDUNDANT records,
even if the record contains fewer fields.
rec_offs_make_valid(): Add the parameter 'leaf'.
rec_copy_prefix_to_dtuple(): Assert that the tuple is only built
on the core fields. Instant ADD COLUMN only applies to the
clustered index, and we should never build a search key that has
more than the PRIMARY KEY and possibly DB_TRX_ID,DB_ROLL_PTR.
All these columns are always present.
dict_index_build_data_tuple(): Remove assertions that would be
duplicated in rec_copy_prefix_to_dtuple().
rec_init_offsets(): Support ROW_FORMAT=REDUNDANT records whose
number of fields is between n_core_fields and n_fields.
cmp_rec_rec_with_match(): Implement the comparison between two
MIN_REC_FLAG records.
trx_t::in_rollback: Make the field available in non-debug builds.
trx_start_for_ddl_low(): Remove dangerous error-tolerance.
A dictionary transaction must be flagged as such before it has generated
any undo log records. This is because trx_undo_assign_undo() will mark
the transaction as a dictionary transaction in the undo log header
right before the very first undo log record is being written.
btr_index_rec_validate(): Account for instant ADD COLUMN
row_undo_ins_remove_clust_rec(): On the rollback of an insert into
SYS_COLUMNS, revert instant ADD COLUMN in the cache by removing the
last column from the table and the clustered index.
row_search_on_row_ref(), row_undo_mod_parse_undo_rec(), row_undo_mod(),
trx_undo_update_rec_get_update(): Handle the 'default row'
as a special case.
dtuple_t::trim(index): Omit a redundant suffix of an index tuple right
before insert or update. After instant ADD COLUMN, if the last fields
of a clustered index tuple match the 'default row', there is no
need to store them. While trimming the entry, we must hold a page latch,
so that the table cannot be emptied and the 'default row' be deleted.
btr_cur_optimistic_update(), btr_cur_pessimistic_update(),
row_upd_clust_rec_by_insert(), row_ins_clust_index_entry_low():
Invoke dtuple_t::trim() if needed.
row_ins_clust_index_entry(): Restore dtuple_t::n_fields after calling
row_ins_clust_index_entry_low().
rec_get_converted_size(), rec_get_converted_size_comp(): Allow the number
of fields to be between n_core_fields and n_fields. Do not support
infimum,supremum. They are never supposed to be stored in dtuple_t,
because page creation nowadays uses a lower-level method for initializing
them.
rec_convert_dtuple_to_rec_comp(): Assign the status bits based on the
number of fields.
btr_cur_trim(): In an update, trim the index entry as needed. For the
'default row', handle rollback specially. For user records, omit
fields that match the 'default row'.
btr_cur_optimistic_delete_func(), btr_cur_pessimistic_delete():
Skip locking and adaptive hash index for the 'default row'.
row_log_table_apply_convert_mrec(): Replace 'default row' values if needed.
In the temporary file that is applied by row_log_table_apply(),
we must identify whether the records contain the extra header for
instantly added columns. For now, we will allocate an additional byte
for this for ROW_T_INSERT and ROW_T_UPDATE records when the source table
has been subject to instant ADD COLUMN. The ROW_T_DELETE records are
fine, as they will be converted and will only contain 'core' columns
(PRIMARY KEY and some system columns) that are converted from dtuple_t.
rec_get_converted_size_temp(), rec_init_offsets_temp(),
rec_convert_dtuple_to_temp(): Add the parameter 'status'.
REC_INFO_DEFAULT_ROW = REC_INFO_MIN_REC_FLAG | REC_STATUS_COLUMNS_ADDED:
An info_bits constant for distinguishing the 'default row' record.
rec_comp_status_t: An enum of the status bit values.
rec_leaf_format: An enum that replaces the bool parameter of
rec_init_offsets_comp_ordinary().
2017-10-06 07:00:05 +03:00
|
|
|
/** Apply an update vector to an index entry.
|
|
|
|
@param[in,out] entry index entry to be updated; the clustered index record
|
|
|
|
must be covered by a lock or a page latch to prevent
|
|
|
|
deletion (rollback or purge)
|
|
|
|
@param[in] index index of the entry
|
|
|
|
@param[in] update update vector built for the entry
|
|
|
|
@param[in,out] heap memory heap for copying off-page columns */
|
2014-02-26 19:11:54 +01:00
|
|
|
void
|
|
|
|
row_upd_index_replace_new_col_vals_index_pos(
|
MDEV-11369 Instant ADD COLUMN for InnoDB
For InnoDB tables, adding, dropping and reordering columns has
required a rebuild of the table and all its indexes. Since MySQL 5.6
(and MariaDB 10.0) this has been supported online (LOCK=NONE), allowing
concurrent modification of the tables.
This work revises the InnoDB ROW_FORMAT=REDUNDANT, ROW_FORMAT=COMPACT
and ROW_FORMAT=DYNAMIC so that columns can be appended instantaneously,
with only minor changes performed to the table structure. The counter
innodb_instant_alter_column in INFORMATION_SCHEMA.GLOBAL_STATUS
is incremented whenever a table rebuild operation is converted into
an instant ADD COLUMN operation.
ROW_FORMAT=COMPRESSED tables will not support instant ADD COLUMN.
Some usability limitations will be addressed in subsequent work:
MDEV-13134 Introduce ALTER TABLE attributes ALGORITHM=NOCOPY
and ALGORITHM=INSTANT
MDEV-14016 Allow instant ADD COLUMN, ADD INDEX, LOCK=NONE
The format of the clustered index (PRIMARY KEY) is changed as follows:
(1) The FIL_PAGE_TYPE of the root page will be FIL_PAGE_TYPE_INSTANT,
and a new field PAGE_INSTANT will contain the original number of fields
in the clustered index ('core' fields).
If instant ADD COLUMN has not been used or the table becomes empty,
or the very first instant ADD COLUMN operation is rolled back,
the fields PAGE_INSTANT and FIL_PAGE_TYPE will be reset
to 0 and FIL_PAGE_INDEX.
(2) A special 'default row' record is inserted into the leftmost leaf,
between the page infimum and the first user record. This record is
distinguished by the REC_INFO_MIN_REC_FLAG, and it is otherwise in the
same format as records that contain values for the instantly added
columns. This 'default row' always has the same number of fields as
the clustered index according to the table definition. The values of
'core' fields are to be ignored. For other fields, the 'default row'
will contain the default values as they were during the ALTER TABLE
statement. (If the column default values are changed later, those
values will only be stored in the .frm file. The 'default row' will
contain the original evaluated values, which must be the same for
every row.) The 'default row' must be completely hidden from
higher-level access routines. Assertions have been added to ensure
that no 'default row' is ever present in the adaptive hash index
or in locked records. The 'default row' is never delete-marked.
(3) In clustered index leaf page records, the number of fields must
reside between the number of 'core' fields (dict_index_t::n_core_fields
introduced in this work) and dict_index_t::n_fields. If the number
of fields is less than dict_index_t::n_fields, the missing fields
are replaced with the column value of the 'default row'.
Note: The number of fields in the record may shrink if some of the
last instantly added columns are updated to the value that is
in the 'default row'. The function btr_cur_trim() implements this
'compression' on update and rollback; dtuple::trim() implements it
on insert.
(4) In ROW_FORMAT=COMPACT and ROW_FORMAT=DYNAMIC records, the new
status value REC_STATUS_COLUMNS_ADDED will indicate the presence of
a new record header that will encode n_fields-n_core_fields-1 in
1 or 2 bytes. (In ROW_FORMAT=REDUNDANT records, the record header
always explicitly encodes the number of fields.)
We introduce the undo log record type TRX_UNDO_INSERT_DEFAULT for
covering the insert of the 'default row' record when instant ADD COLUMN
is used for the first time. Subsequent instant ADD COLUMN can use
TRX_UNDO_UPD_EXIST_REC.
This is joint work with Vin Chen (陈福荣) from Tencent. The design
that was discussed in April 2017 would not have allowed import or
export of data files, because instead of the 'default row' it would
have introduced a data dictionary table. The test
rpl.rpl_alter_instant is exactly as contributed in pull request #408.
The test innodb.instant_alter is based on a contributed test.
The redo log record format changes for ROW_FORMAT=DYNAMIC and
ROW_FORMAT=COMPACT are as contributed. (With this change present,
crash recovery from MariaDB 10.3.1 will fail in spectacular ways!)
Also the semantics of higher-level redo log records that modify the
PAGE_INSTANT field is changed. The redo log format version identifier
was already changed to LOG_HEADER_FORMAT_CURRENT=103 in MariaDB 10.3.1.
Everything else has been rewritten by me. Thanks to Elena Stepanova,
the code has been tested extensively.
When rolling back an instant ADD COLUMN operation, we must empty the
PAGE_FREE list after deleting or shortening the 'default row' record,
by calling either btr_page_empty() or btr_page_reorganize(). We must
know the size of each entry in the PAGE_FREE list. If rollback left a
freed copy of the 'default row' in the PAGE_FREE list, we would be
unable to determine its size (if it is in ROW_FORMAT=COMPACT or
ROW_FORMAT=DYNAMIC) because it would contain more fields than the
rolled-back definition of the clustered index.
UNIV_SQL_DEFAULT: A new special constant that designates an instantly
added column that is not present in the clustered index record.
len_is_stored(): Check if a length is an actual length. There are
two magic length values: UNIV_SQL_DEFAULT, UNIV_SQL_NULL.
dict_col_t::def_val: The 'default row' value of the column. If the
column is not added instantly, def_val.len will be UNIV_SQL_DEFAULT.
dict_col_t: Add the accessors is_virtual(), is_nullable(), is_instant(),
instant_value().
dict_col_t::remove_instant(): Remove the 'instant ADD' status of
a column.
dict_col_t::name(const dict_table_t& table): Replaces
dict_table_get_col_name().
dict_index_t::n_core_fields: The original number of fields.
For secondary indexes and if instant ADD COLUMN has not been used,
this will be equal to dict_index_t::n_fields.
dict_index_t::n_core_null_bytes: Number of bytes needed to
represent the null flags; usually equal to UT_BITS_IN_BYTES(n_nullable).
dict_index_t::NO_CORE_NULL_BYTES: Magic value signalling that
n_core_null_bytes was not initialized yet from the clustered index
root page.
dict_index_t: Add the accessors is_instant(), is_clust(),
get_n_nullable(), instant_field_value().
dict_index_t::instant_add_field(): Adjust clustered index metadata
for instant ADD COLUMN.
dict_index_t::remove_instant(): Remove the 'instant ADD' status
of a clustered index when the table becomes empty, or the very first
instant ADD COLUMN operation is rolled back.
dict_table_t: Add the accessors is_instant(), is_temporary(),
supports_instant().
dict_table_t::instant_add_column(): Adjust metadata for
instant ADD COLUMN.
dict_table_t::rollback_instant(): Adjust metadata on the rollback
of instant ADD COLUMN.
prepare_inplace_alter_table_dict(): First create the ctx->new_table,
and only then decide if the table really needs to be rebuilt.
We must split the creation of table or index metadata from the
creation of the dictionary table records and the creation of
the data. In this way, we can transform a table-rebuilding operation
into an instant ADD COLUMN operation. Dictionary objects will only
be added to cache when table rebuilding or index creation is needed.
The ctx->instant_table will never be added to cache.
dict_table_t::add_to_cache(): Modified and renamed from
dict_table_add_to_cache(). Do not modify the table metadata.
Let the callers invoke dict_table_add_system_columns() and if needed,
set can_be_evicted.
dict_create_sys_tables_tuple(), dict_create_table_step(): Omit the
system columns (which will now exist in the dict_table_t object
already at this point).
dict_create_table_step(): Expect the callers to invoke
dict_table_add_system_columns().
pars_create_table(): Before creating the table creation execution
graph, invoke dict_table_add_system_columns().
row_create_table_for_mysql(): Expect all callers to invoke
dict_table_add_system_columns().
create_index_dict(): Replaces row_merge_create_index_graph().
innodb_update_n_cols(): Renamed from innobase_update_n_virtual().
Call my_error() if an error occurs.
btr_cur_instant_init(), btr_cur_instant_init_low(),
btr_cur_instant_root_init():
Load additional metadata from the clustered index and set
dict_index_t::n_core_null_bytes. This is invoked
when table metadata is first loaded into the data dictionary.
dict_boot(): Initialize n_core_null_bytes for the four hard-coded
dictionary tables.
dict_create_index_step(): Initialize n_core_null_bytes. This is
executed as part of CREATE TABLE.
dict_index_build_internal_clust(): Initialize n_core_null_bytes to
NO_CORE_NULL_BYTES if table->supports_instant().
row_create_index_for_mysql(): Initialize n_core_null_bytes for
CREATE TEMPORARY TABLE.
commit_cache_norebuild(): Call the code to rename or enlarge columns
in the cache only if instant ADD COLUMN is not being used.
(Instant ADD COLUMN would copy all column metadata from
instant_table to old_table, including the names and lengths.)
PAGE_INSTANT: A new 13-bit field for storing dict_index_t::n_core_fields.
This is repurposing the 16-bit field PAGE_DIRECTION, of which only the
least significant 3 bits were used. The original byte containing
PAGE_DIRECTION will be accessible via the new constant PAGE_DIRECTION_B.
page_get_instant(), page_set_instant(): Accessors for the PAGE_INSTANT.
page_ptr_get_direction(), page_get_direction(),
page_ptr_set_direction(): Accessors for PAGE_DIRECTION.
page_direction_reset(): Reset PAGE_DIRECTION, PAGE_N_DIRECTION.
page_direction_increment(): Increment PAGE_N_DIRECTION
and set PAGE_DIRECTION.
rec_get_offsets(): Use the 'leaf' parameter for non-debug purposes,
and assume that heap_no is always set.
Initialize all dict_index_t::n_fields for ROW_FORMAT=REDUNDANT records,
even if the record contains fewer fields.
rec_offs_make_valid(): Add the parameter 'leaf'.
rec_copy_prefix_to_dtuple(): Assert that the tuple is only built
on the core fields. Instant ADD COLUMN only applies to the
clustered index, and we should never build a search key that has
more than the PRIMARY KEY and possibly DB_TRX_ID,DB_ROLL_PTR.
All these columns are always present.
dict_index_build_data_tuple(): Remove assertions that would be
duplicated in rec_copy_prefix_to_dtuple().
rec_init_offsets(): Support ROW_FORMAT=REDUNDANT records whose
number of fields is between n_core_fields and n_fields.
cmp_rec_rec_with_match(): Implement the comparison between two
MIN_REC_FLAG records.
trx_t::in_rollback: Make the field available in non-debug builds.
trx_start_for_ddl_low(): Remove dangerous error-tolerance.
A dictionary transaction must be flagged as such before it has generated
any undo log records. This is because trx_undo_assign_undo() will mark
the transaction as a dictionary transaction in the undo log header
right before the very first undo log record is being written.
btr_index_rec_validate(): Account for instant ADD COLUMN
row_undo_ins_remove_clust_rec(): On the rollback of an insert into
SYS_COLUMNS, revert instant ADD COLUMN in the cache by removing the
last column from the table and the clustered index.
row_search_on_row_ref(), row_undo_mod_parse_undo_rec(), row_undo_mod(),
trx_undo_update_rec_get_update(): Handle the 'default row'
as a special case.
dtuple_t::trim(index): Omit a redundant suffix of an index tuple right
before insert or update. After instant ADD COLUMN, if the last fields
of a clustered index tuple match the 'default row', there is no
need to store them. While trimming the entry, we must hold a page latch,
so that the table cannot be emptied and the 'default row' be deleted.
btr_cur_optimistic_update(), btr_cur_pessimistic_update(),
row_upd_clust_rec_by_insert(), row_ins_clust_index_entry_low():
Invoke dtuple_t::trim() if needed.
row_ins_clust_index_entry(): Restore dtuple_t::n_fields after calling
row_ins_clust_index_entry_low().
rec_get_converted_size(), rec_get_converted_size_comp(): Allow the number
of fields to be between n_core_fields and n_fields. Do not support
infimum,supremum. They are never supposed to be stored in dtuple_t,
because page creation nowadays uses a lower-level method for initializing
them.
rec_convert_dtuple_to_rec_comp(): Assign the status bits based on the
number of fields.
btr_cur_trim(): In an update, trim the index entry as needed. For the
'default row', handle rollback specially. For user records, omit
fields that match the 'default row'.
btr_cur_optimistic_delete_func(), btr_cur_pessimistic_delete():
Skip locking and adaptive hash index for the 'default row'.
row_log_table_apply_convert_mrec(): Replace 'default row' values if needed.
In the temporary file that is applied by row_log_table_apply(),
we must identify whether the records contain the extra header for
instantly added columns. For now, we will allocate an additional byte
for this for ROW_T_INSERT and ROW_T_UPDATE records when the source table
has been subject to instant ADD COLUMN. The ROW_T_DELETE records are
fine, as they will be converted and will only contain 'core' columns
(PRIMARY KEY and some system columns) that are converted from dtuple_t.
rec_get_converted_size_temp(), rec_init_offsets_temp(),
rec_convert_dtuple_to_temp(): Add the parameter 'status'.
REC_INFO_DEFAULT_ROW = REC_INFO_MIN_REC_FLAG | REC_STATUS_COLUMNS_ADDED:
An info_bits constant for distinguishing the 'default row' record.
rec_comp_status_t: An enum of the status bit values.
rec_leaf_format: An enum that replaces the bool parameter of
rec_init_offsets_comp_ordinary().
2017-10-06 07:00:05 +03:00
|
|
|
dtuple_t* entry,
|
|
|
|
const dict_index_t* index,
|
|
|
|
const upd_t* update,
|
|
|
|
mem_heap_t* heap)
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((nonnull));
|
2014-02-26 19:11:54 +01:00
|
|
|
/***********************************************************//**
|
|
|
|
Replaces the new column values stored in the update vector to the index entry
|
|
|
|
given. */
|
|
|
|
void
|
|
|
|
row_upd_index_replace_new_col_vals(
|
|
|
|
/*===============================*/
|
|
|
|
dtuple_t* entry, /*!< in/out: index entry where replaced;
|
|
|
|
the clustered index record must be
|
|
|
|
covered by a lock or a page latch to
|
|
|
|
prevent deletion (rollback or purge) */
|
|
|
|
dict_index_t* index, /*!< in: index; NOTE that this may also be a
|
|
|
|
non-clustered index */
|
|
|
|
const upd_t* update, /*!< in: an update vector built for the
|
|
|
|
CLUSTERED index so that the field number in
|
|
|
|
an upd_field is the clustered index position */
|
|
|
|
mem_heap_t* heap) /*!< in: memory heap for allocating and
|
|
|
|
copying the new values */
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((nonnull));
|
2014-02-26 19:11:54 +01:00
|
|
|
/***********************************************************//**
|
|
|
|
Replaces the new column values stored in the update vector. */
|
|
|
|
void
|
|
|
|
row_upd_replace(
|
|
|
|
/*============*/
|
|
|
|
dtuple_t* row, /*!< in/out: row where replaced,
|
|
|
|
indexed by col_no;
|
|
|
|
the clustered index record must be
|
|
|
|
covered by a lock or a page latch to
|
|
|
|
prevent deletion (rollback or purge) */
|
|
|
|
row_ext_t** ext, /*!< out, own: NULL, or externally
|
|
|
|
stored column prefixes */
|
|
|
|
const dict_index_t* index, /*!< in: clustered index */
|
|
|
|
const upd_t* update, /*!< in: an update vector built for the
|
|
|
|
clustered index */
|
|
|
|
mem_heap_t* heap); /*!< in: memory heap */
|
2016-08-12 11:17:45 +03:00
|
|
|
/** Replaces the virtual column values stored in a dtuple with that of
|
|
|
|
a update vector.
|
|
|
|
@param[in,out] row dtuple whose column to be updated
|
|
|
|
@param[in] table table
|
|
|
|
@param[in] update an update vector built for the clustered index
|
|
|
|
@param[in] upd_new update to new or old value
|
|
|
|
@param[in,out] undo_row undo row (if needs to be updated)
|
|
|
|
@param[in] ptr remaining part in update undo log */
|
|
|
|
void
|
|
|
|
row_upd_replace_vcol(
|
|
|
|
dtuple_t* row,
|
|
|
|
const dict_table_t* table,
|
|
|
|
const upd_t* update,
|
|
|
|
bool upd_new,
|
|
|
|
dtuple_t* undo_row,
|
|
|
|
const byte* ptr);
|
|
|
|
|
2014-02-26 19:11:54 +01:00
|
|
|
/***********************************************************//**
|
|
|
|
Checks if an update vector changes an ordering field of an index record.
|
|
|
|
|
|
|
|
This function is fast if the update vector is short or the number of ordering
|
|
|
|
fields in the index is small. Otherwise, this can be quadratic.
|
|
|
|
NOTE: we compare the fields as binary strings!
|
|
|
|
@return TRUE if update vector changes an ordering field in the index record */
|
|
|
|
ibool
|
|
|
|
row_upd_changes_ord_field_binary_func(
|
|
|
|
/*==================================*/
|
|
|
|
dict_index_t* index, /*!< in: index of the record */
|
|
|
|
const upd_t* update, /*!< in: update vector for the row; NOTE: the
|
|
|
|
field numbers in this MUST be clustered index
|
|
|
|
positions! */
|
|
|
|
#ifdef UNIV_DEBUG
|
|
|
|
const que_thr_t*thr, /*!< in: query thread */
|
|
|
|
#endif /* UNIV_DEBUG */
|
|
|
|
const dtuple_t* row, /*!< in: old value of row, or NULL if the
|
|
|
|
row and the data values in update are not
|
|
|
|
known when this function is called, e.g., at
|
|
|
|
compile time */
|
2016-08-12 11:17:45 +03:00
|
|
|
const row_ext_t*ext, /*!< NULL, or prefixes of the externally
|
2014-02-26 19:11:54 +01:00
|
|
|
stored columns in the old row */
|
2016-08-12 11:17:45 +03:00
|
|
|
ulint flag) /*!< in: ROW_BUILD_NORMAL,
|
|
|
|
ROW_BUILD_FOR_PURGE or ROW_BUILD_FOR_UNDO */
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((nonnull(1,2), warn_unused_result));
|
2014-02-26 19:11:54 +01:00
|
|
|
#ifdef UNIV_DEBUG
|
|
|
|
# define row_upd_changes_ord_field_binary(index,update,thr,row,ext) \
|
2016-08-12 11:17:45 +03:00
|
|
|
row_upd_changes_ord_field_binary_func(index,update,thr,row,ext,0)
|
2014-02-26 19:11:54 +01:00
|
|
|
#else /* UNIV_DEBUG */
|
|
|
|
# define row_upd_changes_ord_field_binary(index,update,thr,row,ext) \
|
2016-08-12 11:17:45 +03:00
|
|
|
row_upd_changes_ord_field_binary_func(index,update,row,ext,0)
|
2014-02-26 19:11:54 +01:00
|
|
|
#endif /* UNIV_DEBUG */
|
|
|
|
/***********************************************************//**
|
|
|
|
Checks if an FTS indexed column is affected by an UPDATE.
|
|
|
|
@return offset within fts_t::indexes if FTS indexed column updated else
|
|
|
|
ULINT_UNDEFINED */
|
|
|
|
ulint
|
|
|
|
row_upd_changes_fts_column(
|
|
|
|
/*=======================*/
|
|
|
|
dict_table_t* table, /*!< in: table */
|
|
|
|
upd_field_t* upd_field); /*!< in: field to check */
|
|
|
|
/***********************************************************//**
|
|
|
|
Checks if an FTS Doc ID column is affected by an UPDATE.
|
|
|
|
@return whether Doc ID column is affected */
|
|
|
|
bool
|
|
|
|
row_upd_changes_doc_id(
|
|
|
|
/*===================*/
|
|
|
|
dict_table_t* table, /*!< in: table */
|
|
|
|
upd_field_t* upd_field) /*!< in: field to check */
|
2016-06-21 14:21:03 +02:00
|
|
|
MY_ATTRIBUTE((nonnull, warn_unused_result));
|
2014-02-26 19:11:54 +01:00
|
|
|
/***********************************************************//**
|
|
|
|
Checks if an update vector changes an ordering field of an index record.
|
|
|
|
This function is fast if the update vector is short or the number of ordering
|
|
|
|
fields in the index is small. Otherwise, this can be quadratic.
|
|
|
|
NOTE: we compare the fields as binary strings!
|
|
|
|
@return TRUE if update vector may change an ordering field in an index
|
|
|
|
record */
|
|
|
|
ibool
|
|
|
|
row_upd_changes_some_index_ord_field_binary(
|
|
|
|
/*========================================*/
|
|
|
|
const dict_table_t* table, /*!< in: table */
|
|
|
|
const upd_t* update);/*!< in: update vector for the row */
|
2016-09-06 09:43:16 +03:00
|
|
|
/** Stores to the heap the row on which the node->pcur is positioned.
|
|
|
|
@param[in] node row update node
|
|
|
|
@param[in] thd mysql thread handle
|
|
|
|
@param[in,out] mysql_table NULL, or mysql table object when
|
|
|
|
user thread invokes dml */
|
2016-08-12 11:17:45 +03:00
|
|
|
void
|
|
|
|
row_upd_store_row(
|
2016-09-06 09:43:16 +03:00
|
|
|
upd_node_t* node,
|
|
|
|
THD* thd,
|
|
|
|
TABLE* mysql_table);
|
2016-08-12 11:17:45 +03:00
|
|
|
/***********************************************************//**
|
2014-02-26 19:11:54 +01:00
|
|
|
Updates a row in a table. This is a high-level function used
|
|
|
|
in SQL execution graphs.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return query thread to run next or NULL */
|
2014-02-26 19:11:54 +01:00
|
|
|
que_thr_t*
|
|
|
|
row_upd_step(
|
|
|
|
/*=========*/
|
|
|
|
que_thr_t* thr); /*!< in: query thread */
|
|
|
|
/*********************************************************************//**
|
|
|
|
Parses the log data of system field values.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return log data end or NULL */
|
2014-02-26 19:11:54 +01:00
|
|
|
byte*
|
|
|
|
row_upd_parse_sys_vals(
|
|
|
|
/*===================*/
|
2016-08-12 11:17:45 +03:00
|
|
|
const byte* ptr, /*!< in: buffer */
|
|
|
|
const byte* end_ptr,/*!< in: buffer end */
|
2014-02-26 19:11:54 +01:00
|
|
|
ulint* pos, /*!< out: TRX_ID position in record */
|
|
|
|
trx_id_t* trx_id, /*!< out: trx id */
|
|
|
|
roll_ptr_t* roll_ptr);/*!< out: roll ptr */
|
|
|
|
/*********************************************************************//**
|
|
|
|
Updates the trx id and roll ptr field in a clustered index record in database
|
|
|
|
recovery. */
|
|
|
|
void
|
|
|
|
row_upd_rec_sys_fields_in_recovery(
|
|
|
|
/*===============================*/
|
|
|
|
rec_t* rec, /*!< in/out: record */
|
|
|
|
page_zip_des_t* page_zip,/*!< in/out: compressed page, or NULL */
|
|
|
|
const ulint* offsets,/*!< in: array returned by rec_get_offsets() */
|
|
|
|
ulint pos, /*!< in: TRX_ID position in rec */
|
|
|
|
trx_id_t trx_id, /*!< in: transaction id */
|
|
|
|
roll_ptr_t roll_ptr);/*!< in: roll ptr of the undo log record */
|
|
|
|
/*********************************************************************//**
|
|
|
|
Parses the log data written by row_upd_index_write_log.
|
2016-08-12 11:17:45 +03:00
|
|
|
@return log data end or NULL */
|
2014-02-26 19:11:54 +01:00
|
|
|
byte*
|
|
|
|
row_upd_index_parse(
|
|
|
|
/*================*/
|
2016-08-12 11:17:45 +03:00
|
|
|
const byte* ptr, /*!< in: buffer */
|
|
|
|
const byte* end_ptr,/*!< in: buffer end */
|
2014-02-26 19:11:54 +01:00
|
|
|
mem_heap_t* heap, /*!< in: memory heap where update vector is
|
|
|
|
built */
|
|
|
|
upd_t** update_out);/*!< out: update vector */
|
|
|
|
|
|
|
|
|
|
|
|
/* Update vector field */
|
|
|
|
struct upd_field_t{
|
|
|
|
unsigned field_no:16; /*!< field number in an index, usually
|
|
|
|
the clustered index, but in updating
|
|
|
|
a secondary index record in btr0cur.cc
|
|
|
|
this is the position in the secondary
|
2016-09-06 09:43:16 +03:00
|
|
|
index. If this field is a virtual
|
|
|
|
column, then field_no represents
|
|
|
|
the nth virtual column in the table */
|
2014-02-26 19:11:54 +01:00
|
|
|
unsigned orig_len:16; /*!< original length of the locally
|
|
|
|
stored part of an externally stored
|
|
|
|
column, or 0 */
|
|
|
|
que_node_t* exp; /*!< expression for calculating a new
|
|
|
|
value: it refers to column values and
|
|
|
|
constants in the symbol table of the
|
|
|
|
query graph */
|
|
|
|
dfield_t new_val; /*!< new value for the column */
|
2016-08-12 11:17:45 +03:00
|
|
|
dfield_t* old_v_val; /*!< old value for the virtual column */
|
2014-02-26 19:11:54 +01:00
|
|
|
};
|
|
|
|
|
2016-08-12 11:17:45 +03:00
|
|
|
|
|
|
|
/* check whether an update field is on virtual column */
|
|
|
|
#define upd_fld_is_virtual_col(upd_fld) \
|
|
|
|
(((upd_fld)->new_val.type.prtype & DATA_VIRTUAL) == DATA_VIRTUAL)
|
|
|
|
|
|
|
|
/* set DATA_VIRTUAL bit on update field to show it is a virtual column */
|
|
|
|
#define upd_fld_set_virtual_col(upd_fld) \
|
|
|
|
((upd_fld)->new_val.type.prtype |= DATA_VIRTUAL)
|
|
|
|
|
2014-02-26 19:11:54 +01:00
|
|
|
/* Update vector structure */
|
|
|
|
struct upd_t{
|
2016-08-12 11:17:45 +03:00
|
|
|
mem_heap_t* heap; /*!< heap from which memory allocated */
|
2014-02-26 19:11:54 +01:00
|
|
|
ulint info_bits; /*!< new value of info bits to record;
|
|
|
|
default is 0 */
|
2016-08-12 11:17:45 +03:00
|
|
|
dtuple_t* old_vrow; /*!< pointer to old row, used for
|
|
|
|
virtual column update now */
|
2014-02-26 19:11:54 +01:00
|
|
|
ulint n_fields; /*!< number of update fields */
|
|
|
|
upd_field_t* fields; /*!< array of update fields */
|
2017-12-03 17:57:32 +03:00
|
|
|
byte vers_sys_value[8]; /*!< buffer for updating system fields */
|
2016-08-12 11:17:45 +03:00
|
|
|
|
|
|
|
/** Append an update field to the end of array
|
|
|
|
@param[in] field an update field */
|
|
|
|
void append(const upd_field_t& field)
|
|
|
|
{
|
|
|
|
fields[n_fields++] = field;
|
|
|
|
}
|
|
|
|
|
|
|
|
/** Determine if the given field_no is modified.
|
|
|
|
@return true if modified, false otherwise. */
|
|
|
|
bool is_modified(const ulint field_no) const
|
|
|
|
{
|
|
|
|
for (ulint i = 0; i < n_fields; ++i) {
|
|
|
|
if (field_no == fields[i].field_no) {
|
|
|
|
return(true);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return(false);
|
|
|
|
}
|
|
|
|
|
2018-02-22 19:14:58 +03:00
|
|
|
/** Determine if the update affects a system versioned column or row_end. */
|
Remove the flag vers_update_trt
THD::vers_update_trt, trx_t::vers_update_trt, trx_savept_t::vers_update_trt:
Remove. Instead, determine from trx_t::mod_tables whether versioned
columns were affected by the transaction.
handlerton::prepare_commit_versioned: Replaces vers_get_trt_data.
Return the transaction start ID and also the commit ID, in case
the transaction modified any system-versioned columns (0 if not).
TR_table::store_data(): Remove (merge with update() below).
TR_table::update(): Add the parameters start_id, end_id.
ha_commit_trans(): Remove a condition on SQLCOM_ALTER_TABLE.
If we need something special for ALTER TABLE...ALGORITHM=INPLACE,
that can be done inside InnoDB by modifying trx_t::mod_tables.
innodb_prepare_commit_versioned(): Renamed from innodb_get_trt_data().
Check trx_t::mod_tables to see if any changes to versioned columns
are present.
trx_mod_table_time_t: A pair of logical timestamps, replacing the
undo_no_t in trx_mod_tables_t. Keep track of not only the first
modification to a persistent table in each transaction, but also
the first modification of a versioned column in a table.
dtype_t, dict_col_t: Add the accessor is_any_versioned(), to check
if the type refers to a system-versioned user or system column.
upd_t::affects_versioned(): Check if an update affects a versioned
column.
trx_undo_report_row_operation(): If a versioned column is affected
by the update, invoke trx_mod_table_time_t::set_versioned().
trx_rollback_to_savepoint_low(): If all changes to versioned columns
were rolled back, invoke trx_mod_table_time_t::rollback_versioned(),
so that trx_mod_table_time_t::is_versioned() will no longer hold.
2017-11-24 12:03:24 +02:00
|
|
|
bool affects_versioned() const
|
|
|
|
{
|
|
|
|
for (ulint i = 0; i < n_fields; i++) {
|
2018-02-22 19:14:58 +03:00
|
|
|
dtype_t type = fields[i].new_val.type;
|
|
|
|
if (type.is_versioned()) {
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
// versioned DELETE is UPDATE SET row_end=NOW
|
|
|
|
if (type.vers_sys_end()) {
|
Remove the flag vers_update_trt
THD::vers_update_trt, trx_t::vers_update_trt, trx_savept_t::vers_update_trt:
Remove. Instead, determine from trx_t::mod_tables whether versioned
columns were affected by the transaction.
handlerton::prepare_commit_versioned: Replaces vers_get_trt_data.
Return the transaction start ID and also the commit ID, in case
the transaction modified any system-versioned columns (0 if not).
TR_table::store_data(): Remove (merge with update() below).
TR_table::update(): Add the parameters start_id, end_id.
ha_commit_trans(): Remove a condition on SQLCOM_ALTER_TABLE.
If we need something special for ALTER TABLE...ALGORITHM=INPLACE,
that can be done inside InnoDB by modifying trx_t::mod_tables.
innodb_prepare_commit_versioned(): Renamed from innodb_get_trt_data().
Check trx_t::mod_tables to see if any changes to versioned columns
are present.
trx_mod_table_time_t: A pair of logical timestamps, replacing the
undo_no_t in trx_mod_tables_t. Keep track of not only the first
modification to a persistent table in each transaction, but also
the first modification of a versioned column in a table.
dtype_t, dict_col_t: Add the accessor is_any_versioned(), to check
if the type refers to a system-versioned user or system column.
upd_t::affects_versioned(): Check if an update affects a versioned
column.
trx_undo_report_row_operation(): If a versioned column is affected
by the update, invoke trx_mod_table_time_t::set_versioned().
trx_rollback_to_savepoint_low(): If all changes to versioned columns
were rolled back, invoke trx_mod_table_time_t::rollback_versioned(),
so that trx_mod_table_time_t::is_versioned() will no longer hold.
2017-11-24 12:03:24 +02:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2016-08-12 11:17:45 +03:00
|
|
|
#ifdef UNIV_DEBUG
|
|
|
|
bool validate() const
|
|
|
|
{
|
|
|
|
for (ulint i = 0; i < n_fields; ++i) {
|
|
|
|
dfield_t* field = &fields[i].new_val;
|
|
|
|
if (dfield_is_ext(field)) {
|
|
|
|
ut_ad(dfield_get_len(field)
|
|
|
|
>= BTR_EXTERN_FIELD_REF_SIZE);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return(true);
|
|
|
|
}
|
|
|
|
#endif // UNIV_DEBUG
|
|
|
|
|
2014-02-26 19:11:54 +01:00
|
|
|
};
|
|
|
|
|
2017-11-28 15:03:25 +03:00
|
|
|
/** Kinds of update operation */
|
|
|
|
enum delete_mode_t {
|
|
|
|
NO_DELETE = 0, /*!< this operation does not delete */
|
|
|
|
PLAIN_DELETE, /*!< ordinary delete */
|
|
|
|
VERSIONED_DELETE /*!< update old and insert a new row */
|
|
|
|
};
|
|
|
|
|
2014-02-26 19:11:54 +01:00
|
|
|
/* Update node structure which also implements the delete operation
|
|
|
|
of a row */
|
|
|
|
|
|
|
|
struct upd_node_t{
|
|
|
|
que_common_t common; /*!< node type: QUE_NODE_UPDATE */
|
2017-11-28 15:03:25 +03:00
|
|
|
delete_mode_t is_delete; /*!< kind of DELETE */
|
2014-02-26 19:11:54 +01:00
|
|
|
ibool searched_update;
|
|
|
|
/* TRUE if searched update, FALSE if
|
|
|
|
positioned */
|
2018-04-28 15:49:09 +03:00
|
|
|
bool in_mysql_interface;
|
|
|
|
/* whether the update node was created
|
2014-02-26 19:11:54 +01:00
|
|
|
for the MySQL interface */
|
2016-08-12 11:17:45 +03:00
|
|
|
dict_foreign_t* foreign;/* NULL or pointer to a foreign key
|
|
|
|
constraint if this update node is used in
|
|
|
|
doing an ON DELETE or ON UPDATE operation */
|
2018-02-06 13:53:20 +02:00
|
|
|
upd_node_t* cascade_node;/* NULL or an update node template which
|
|
|
|
is used to implement ON DELETE/UPDATE CASCADE
|
|
|
|
or ... SET NULL for foreign keys */
|
2016-09-06 09:43:16 +03:00
|
|
|
mem_heap_t* cascade_heap;
|
2018-02-06 13:53:20 +02:00
|
|
|
/*!< NULL or a mem heap where cascade
|
|
|
|
node is created.*/
|
2014-02-26 19:11:54 +01:00
|
|
|
sel_node_t* select; /*!< query graph subtree implementing a base
|
|
|
|
table cursor: the rows returned will be
|
|
|
|
updated */
|
|
|
|
btr_pcur_t* pcur; /*!< persistent cursor placed on the clustered
|
|
|
|
index record which should be updated or
|
|
|
|
deleted; the cursor is stored in the graph
|
|
|
|
of 'select' field above, except in the case
|
|
|
|
of the MySQL interface */
|
|
|
|
dict_table_t* table; /*!< table where updated */
|
|
|
|
upd_t* update; /*!< update vector for the row */
|
|
|
|
ulint update_n_fields;
|
|
|
|
/* when this struct is used to implement
|
|
|
|
a cascade operation for foreign keys, we store
|
|
|
|
here the size of the buffer allocated for use
|
|
|
|
as the update vector */
|
|
|
|
sym_node_list_t columns;/* symbol table nodes for the columns
|
|
|
|
to retrieve from the table */
|
|
|
|
ibool has_clust_rec_x_lock;
|
|
|
|
/* TRUE if the select which retrieves the
|
|
|
|
records to update already sets an x-lock on
|
|
|
|
the clustered record; note that it must always
|
|
|
|
set at least an s-lock */
|
|
|
|
ulint cmpl_info;/* information extracted during query
|
|
|
|
compilation; speeds up execution:
|
|
|
|
UPD_NODE_NO_ORD_CHANGE and
|
|
|
|
UPD_NODE_NO_SIZE_CHANGE, ORed */
|
|
|
|
/*----------------------*/
|
|
|
|
/* Local storage for this graph node */
|
|
|
|
ulint state; /*!< node execution state */
|
|
|
|
dict_index_t* index; /*!< NULL, or the next index whose record should
|
|
|
|
be updated */
|
|
|
|
dtuple_t* row; /*!< NULL, or a copy (also fields copied to
|
|
|
|
heap) of the row to update; this must be reset
|
|
|
|
to NULL after a successful update */
|
2018-03-19 17:25:58 +03:00
|
|
|
dtuple_t* historical_row; /*!< historical row used in
|
2018-04-06 17:20:45 +03:00
|
|
|
CASCADE UPDATE/SET NULL;
|
|
|
|
allocated from historical_heap */
|
|
|
|
mem_heap_t* historical_heap; /*!< heap for historical row insertion;
|
|
|
|
created when row to update is located;
|
|
|
|
freed right before row update */
|
2014-02-26 19:11:54 +01:00
|
|
|
row_ext_t* ext; /*!< NULL, or prefixes of the externally
|
|
|
|
stored columns in the old row */
|
|
|
|
dtuple_t* upd_row;/* NULL, or a copy of the updated row */
|
|
|
|
row_ext_t* upd_ext;/* NULL, or prefixes of the externally
|
|
|
|
stored columns in upd_row */
|
|
|
|
mem_heap_t* heap; /*!< memory heap used as auxiliary storage;
|
|
|
|
this must be emptied after a successful
|
|
|
|
update */
|
|
|
|
/*----------------------*/
|
|
|
|
sym_node_t* table_sym;/* table node in symbol table */
|
|
|
|
que_node_t* col_assign_list;
|
|
|
|
/* column assignment list */
|
|
|
|
ulint magic_n;
|
2016-08-12 11:17:45 +03:00
|
|
|
|
2018-03-19 17:25:58 +03:00
|
|
|
/** Also set row_start = CURRENT_TIMESTAMP/trx->id
|
|
|
|
@param[in] trx transaction */
|
|
|
|
void make_versioned_update(const trx_t* trx);
|
|
|
|
/** Only set row_end = CURRENT_TIMESTAMP/trx->id.
|
|
|
|
Do not touch other fields at all.
|
|
|
|
@param[in] trx transaction */
|
|
|
|
void make_versioned_delete(const trx_t* trx);
|
|
|
|
|
|
|
|
private:
|
|
|
|
/** Appends row_start or row_end field to update vector and sets a
|
|
|
|
CURRENT_TIMESTAMP/trx->id value to it.
|
|
|
|
Supposed to be called only by make_versioned_update() and
|
|
|
|
make_versioned_delete().
|
|
|
|
@param[in] trx transaction
|
|
|
|
@param[in] vers_sys_idx table->row_start or table->row_end */
|
|
|
|
void make_versioned_helper(const trx_t* trx, ulint idx);
|
2014-02-26 19:11:54 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
#define UPD_NODE_MAGIC_N 1579975
|
|
|
|
|
|
|
|
/* Node execution states */
|
|
|
|
#define UPD_NODE_SET_IX_LOCK 1 /* execution came to the node from
|
|
|
|
a node above and if the field
|
|
|
|
has_clust_rec_x_lock is FALSE, we
|
|
|
|
should set an intention x-lock on
|
|
|
|
the table */
|
|
|
|
#define UPD_NODE_UPDATE_CLUSTERED 2 /* clustered index record should be
|
|
|
|
updated */
|
|
|
|
#define UPD_NODE_INSERT_CLUSTERED 3 /* clustered index record should be
|
|
|
|
inserted, old record is already delete
|
|
|
|
marked */
|
|
|
|
#define UPD_NODE_UPDATE_ALL_SEC 5 /* an ordering field of the clustered
|
|
|
|
index record was changed, or this is
|
|
|
|
a delete operation: should update
|
|
|
|
all the secondary index records */
|
|
|
|
#define UPD_NODE_UPDATE_SOME_SEC 6 /* secondary index entries should be
|
|
|
|
looked at and updated if an ordering
|
|
|
|
field changed */
|
|
|
|
|
|
|
|
/* Compilation info flags: these must fit within 3 bits; see trx0rec.h */
|
|
|
|
#define UPD_NODE_NO_ORD_CHANGE 1 /* no secondary index record will be
|
|
|
|
changed in the update and no ordering
|
|
|
|
field of the clustered index */
|
|
|
|
#define UPD_NODE_NO_SIZE_CHANGE 2 /* no record field size will be
|
|
|
|
changed in the update */
|
|
|
|
|
|
|
|
|
|
|
|
#include "row0upd.ic"
|
|
|
|
|
|
|
|
#endif
|