There were numerous memory leaks in plugin's code and associated code.
During typical run of MTR tests it leaked around 65 megabytes of
objects. As a result they may severely affect long-lived connections.
This patch fixes (almost) all leaks found in the plugin. The exceptions
are two leaks associated with SHOW CREATE TABLE columnstore_table and
getting information of columns of columnstore-handled table. These
should be fixed on the server side and work is on the way.
The most important fix here is the fix of possible buffer overrun in
DATEFORMAT() function. A "%W" format, repeated enough times, would
overflow the 256-bytes buffer for result. Now we use ostringstream to
construct result and we are safe.
Changes in date/time projection functions made me fix difference between
us and server behavior. The new, better behavior is reflected in changes
in tests' results.
Also, there was incorrect logic in TRUNCATE() and ROUND() functions in
computing the decimal "shift."
* fix(rowgroup): RGData now uses uint64_t counter for the fixed sizes columns data buf.
The buffer can utilize > 4GB RAM that is necessary for PM side join.
RGData ctor uses uint32_t allocating data buffer.
This fact causes implicit heap overflow.
* feat(bytestream,serdes): BS buffer size type is uint64_t
This necessary to handle 64bit RGData, that comes as
a separate patch. The pair of patches would allow to
have PM joins when SmallSide size > 4GB.
* feat(bytestream,serdes): Distribute BS buf size data type change to avoid implicit data type narrowing
* feat(rowgroup): this returns bits lost during cherry-pick. The bits lost caused the first RGData::serialize to crash a process
fix(client): Fix columnstore.cnf file
This fix changes option file to apply '--quick' option only for 'mariadb' and 'mysql' clients instead of all MariaDB clients.
Otherwise 'mysqladmin' uses this option, but it doesn't exist. As a result broken CI multinode MTR stage.
This changeset enables quick (mariadb -q) mode when columnstore is
installed. Quick mode precludes client CLI program from storing too
much data in memory, preventing out of memory conditions.
This fixes discrepance with the server, which assigns DATE type to
last_day()'s result.
Now we also assigns DATE result type and, also, use proper
dataconvert::Day data structure to return date.
Tests agree with InnoDB.
Also, this patch includes test for MCOL-5669, to show we fixed it.
When ORDER BY column is not in GROUP BY, is not an aggregate and there
is a SELECT column that is also not an aggregate, there was a problem:
ordering happened on the SELECTed column, not ORDERed one.
This patch fixes that particular problem and also performs some tidying
around newly added aggregate.
This patch fixes the problem in MCOL-4234 and also generally improves
behavior of GROUP BY.
It does so by introducing a "dummy" aggregate and by wrapping columns
into it. This allows for columns that are not in GROUP BY to be used
more freely, for example, in SELECT * FROM tbl GROUP BY col - all
columns that are not "col" will be wrapped into an aggregate and query
will proceed to execution.
The dummy aggregate itself does nothing more than remember last value
passed into it.
There also an additional error message that tries to explain what types
of expressions can be wrapped into an aggregate.
Server expands ut8_XXX aliases to utf8mb3_XXX or utf8mb4_XXX depending
on the UTF8_IS_UTF8MB3 setting in the OLD_MODE environment variable.
Server already has the necessary code implemented in the get_utf8_flag()
method of class THD. There are several uses of this flag and all we have
to do to be in line with server is to use it.
This patch does that for DDL as work on MCOL-5705 uncovered some
problems in that area.
This changeset contains fixes in Oracle mode tests and for the
implementation of the CONCAT_ORACLE. Also, we harmonise our
translation process with the recent changes in the server.
Due to changed behavior of the server, some CREATE VIEW/EXPLAIN
statements' results begun to output unexpected results and need to be
fixed.
Also, concatenation operation's name also changed. This lead to
disabled func_concat_oracle test to be enabled to test it and it
turned out that our implementation of this function was broken
and need to be fixed too.
This changeset contains fixes in Oracle mode tests and for the
implementation of the CONCAT_ORACLE. Also, we harmonise our
translation process with the recent changes in the server.
Due to changed behavior of the server, some CREATE VIEW/EXPLAIN
statements' results begun to output unexpected results and need to be
fixed.
Also, concatenation operation's name also changed. This lead to
disabled func_concat_oracle test to be enabled to test it and it
turned out that our implementation of this function was broken
and need to be fixed too.
Fixes MCOL-5643.
The problem was that different views with same column names in GROUP BY
and on the SELECT clause produced an error about "projection column is
not an aggergate neither in GROUP BY list."
This was due to incorrect search in expressions's list that lead to
duplicate columns in GROUP BY list.
For UPDATEs involving a single table, the server call to handler::direct_update_rows() is used to correctly set the count for the number of updated rows in the UPDATE statement.
However, for UPDATEs involving multi-tables, the server does not call handler::direct_update_rows(). This patch adds support to correctly report the number of updated rows to the client by setting
multi_update::updated and multi_update::found in handler::rnd_end().
We add intermediate calculations in int128_t when target is UBIGINT and
check for overflow before converting into the UBIGINT. This is so
because we can overflow on addition and multiplication, with (some)
signed operands or both unsigned.
Adds a special column which helps to differentiate data and rollups of
various depts and a simple logic to row aggregation to add processing of
subtotals.
1. Extend the calpontsys.syscolumn system catalog table
with a new column, 'charsetnum'.
'charsetnum' field is set to the 'number' member of the
'charset_info_st' struct defined in the server in m_ctype.h.
For CHAR/VARCHAR/TEXT column types, 'charset_info_st' is
initialized to the charset/collation of the column, which
is set at the column-level or at the table-level in the DDL.
For BLOB/VARBINARY binary column types, 'charset_info_st' is
initialized to my_charset_bin (charsetnum=63).
For all other column types, charsetnum is set to 0.
2. Add support for the newly added 'charsetnum' column in the
automatic system catalog upgrade logic in dbbuilder.
For existing table definitions, charsetnum for the column is
defaulted to 0.
3. Add MTR test case that creates a few table definitions with
a range of charset/collation combinations and queries the
calpontsys.syscolumn system catalog table with the charsetnum
field for the columns in the table DDLs.
feat(charset)!: utf8 is a new charset default and utf8_general_ci is a new collation default in the engine configuration file shipped
---------
Co-authored-by: Leonid Fedorov <leonid.fedorov@mariadb.com>
Co-authored-by: mariadb-DanielLee <daniel.lee@mariadb.com>
This patch:
1. Properly processes situation when pm join result count is exceeded.
2. Adds session variable 'columnstore_max_pm_join_result_count` to control the limit.
This patch improves/fixes the existing handling of CHARSET and
COLLATION symbols in the ColumnStore DDL parser.
Also, add fCollate and fCharsetNum member variables to the
ddlpackage::ColumnType class.
This patch:
1. Handles corner case when the bucket exceeded the memory limit, but we cannot redistribute the data in this bucket into new buckets based on a hash algorithm, because the rows have the same values.
2. Adds force option for disk join step.
3. Add a option to contol the depth of the partition tree.
Internal memory representation of MEDIUMINT datatype uses 24 bits. This is
true for both MariaDB server as well as ColumnStore. MCS plugin code uses
TypeHandlerSInt24 and TypeHandlerUInt24 classes to respectively convert the
binary representation of the signed and unsigned MEDIUMINT values passed by
the server to the plugin. The plugin then outputs the text representation
of these values into an open file descriptor which is piped to cpimport
for the final load into the MCS db files.
The TypeHandlerXInt24 classes were earlier incorrectly using
WriteBatchField::ColWriteBatchXInt32() functions which operate on a 4 byte
buffer. This resulted in incorrect parsing of MEDIUMINT values. As a fix,
we implement WriteBatchField::ColWriteBatchXInt24() functions which
correctly handle the 24 bit input buffer used for MEDIUMINT datatype.
For the following query:
select item from (
select item from (select a as item from t1) tt
union all
select item from (select a as item from t1) tt
) ttt;
There is an if predicate in buildSimpleColFromDerivedTable() that compares
the outermost query field name (ttt.item) to the returned column list of
the inner query (tt.item) when building the returned column list of the
outer most query. In the above query example, the inner query field name
is an alias set in the inner most query and is set to "`tt`.`item`",
while the outermost query field name is set to "item". The use of
backticks "`" in the inner query alias is causing the execution to
not enter the if block which creates the SimpleColumn for the outermost
query field name. As a fix, we strip off the backticks from the inner
query alias.
1. Input and output RowGroup's used in GROUP_CONCAT classes
are currently allocating a raw memory buffer of size equal
to the actual width of the string datatype. As an example,
for the following query:
SELECT col1, GROUP_CONCAT(col2) FROM t GROUP BY col1;
If col2 is a TEXT field with default width, the input
RowGroup containing the target rows to be concatenated will
assign 64kb of memory for every input row in the RowGroup.
This is wasteful as actual field values in real workloads
would be much smaller. We fix this by enabling the
RowGroup to use the StringStore when the RowGroup contains
long strings.
2. RowAggregation::initialize() allocates a memory buffer
for a NULL row. The size of this buffer is equal to the
row size for the output RowGroup. For the above scenario,
using the default group_concat_max_len (which is a server
variable that sets the maximum length of the GROUP_CONCAT string)
value of 1mb, the buffer size would be
(1mb + 64kb + some additional metadata). If the user sets
group_concat_max_len to a higher value, say 3gb, this buffer
size would be ~3gb. Now if the runtime initiates several
instances of RowAggregation, total memory consumption by
PrimProc could exceed the hardware memory limits causing the
OS OOM to kill the process. We fix this problem by again
enabling the StringStore for the NULL row allocation.
3. In the plugin code in buildAggregateColumn(), there is
an integer overflow when the server group_concat_max_len
variable (which is an uint32_t) is set to a value > INT32_MAX
(such as 3gb) and is assigned to
CalpontSystemCatalog::ColType::colWidth (which is an int32_t).
As a short term fix, we saturate the assigned value to colWidth
to INT32_MAX. Proper fix would be to upgrade
CalpontSystemCatalog::ColType::colWidth to an uint32_t.
