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.
This patch improves handling of NULLs in textual fields in ColumnStore.
Previously empty strings were considered NULLs and it could be a problem
if data scheme allows for empty strings. It was also one of major
reasons of behavior difference between ColumnStore and other engines in
MariaDB family.
Also, this patch fixes some other bugs and incorrect behavior, for
example, incorrect comparison for "column <= ''" which evaluates to
constant True for all purposes before this patch.
The idea is relatively simple - encode prefixes of collated strings as
integers and use them to compute extents' ranges. Then we can eliminate
extents with strings.
The actual patch does have all the code there but miss one important
step: we do not keep collation index, we keep charset index. Because of
this, some of the tests in the bugfix suite fail and thus main
functionality is turned off.
The reason of this patch to be put into PR at all is that it contains
changes that made CHAR/VARCHAR columns unsigned. This change is needed in
vectorization work.
* Fix clang warnings
* Remove vim tab guides
* initialize variables
* 'strncpy' output truncated before terminating nul copying as many bytes from a string as its length
* Fix ISO C++17 does not allow 'register' storage class specifier for outdated bison
* chars are unsigned on ARM, having if (ival < 0) always false
* chars are unsigned by default on ARM and comparison with -1 if always true
Part 1:
As part of MCOL-3776 to address synchronization issue while accessing
the fTimeZone member of the Func class, mutex locks were added to the
accessor and mutator methods. However, this slows down processing
of TIMESTAMP columns in PrimProc significantly as all threads across
all concurrently running queries would serialize on the mutex. This
is because PrimProc only has a single global object for the functor
class (class derived from Func in utils/funcexp/functor.h) for a given
function name. To fix this problem:
(1) We remove the fTimeZone as a member of the Func derived classes
(hence removing the mutexes) and instead use the fOperationType
member of the FunctionColumn class to propagate the timezone values
down to the individual functor processing functions such as
FunctionColumn::getStrVal(), FunctionColumn::getIntVal(), etc.
(2) To achieve (1), a timezone member is added to the
execplan::CalpontSystemCatalog::ColType class.
Part 2:
Several functors in the Funcexp code call dataconvert::gmtSecToMySQLTime()
and dataconvert::mySQLTimeToGmtSec() functions for conversion between seconds
since unix epoch and broken-down representation. These functions in turn call
the C library function localtime_r() which currently has a known bug of holding
a global lock via a call to __tz_convert. This significantly reduces performance
in multi-threaded applications where multiple threads concurrently call
localtime_r(). More details on the bug:
https://sourceware.org/bugzilla/show_bug.cgi?id=16145
This bug in localtime_r() caused processing of the Functors in PrimProc to
slowdown significantly since a query execution causes Functors code to be
processed in a multi-threaded manner.
As a fix, we remove the calls to localtime_r() from gmtSecToMySQLTime()
and mySQLTimeToGmtSec() by performing the timezone-to-offset conversion
(done in dataconvert::timeZoneToOffset()) during the execution plan
creation in the plugin. Note that localtime_r() is only called when the
time_zone system variable is set to "SYSTEM".
This fix also required changing the timezone type from a std::string to
a long across the system.
This patch:
1. Removes the option to declare uncompressed columns (set columnstore_compression_type = 0).
2. Ignores [COMMENT '[compression=0] option at table or column level (no error messages, just disregard).
3. Removes the option to set more than 2 extents per file (ExtentsPreSegmentFile).
4. Updates rebuildEM tool to support up to 10 dictionary extent per dictionary segment file.
5. Adds check for `DBRootStorageType` for rebuildEM tool.
6. Renamed rebuildEM to mcsRebuildEM.
Progress keep and test commit
Progress keep and test commit
Progress keep and test commit
Again, trying to pinpoint problematic part of a change
Revert "Again, trying to pinpoint problematic part of a change"
This reverts commit 71874e7c0d7e4eeed0c201b12d306b583c07b9e2.
Revert "Progress keep and test commit"
This reverts commit 63c7bc67ae55bdb81433ca58bbd239d6171a1031.
Revert "Progress keep and test commit"
This reverts commit 121c09febd78dacd37158caeab9ac70f65b493df.
Small steps - I walk minefield here
Propagating changes - now CPInfo in convertValArray
Progress keep commit
Restoring old functionality
Progress keep commit
Small steps to avoid/better locate old problem with the write engine.
Progress keeping commit
Thread the CPInfo up to convertValArray call in writeColumnRec
About to test changes - I should get no regression and no updates in
ranges either.
Testing out why I get a regression
Investigating source of regression
Debugging prints
Fix compile error
Debugging print - debug regression
I clearly see calls to writeColumnRec and prints there added to discern
between these.
Fix warning error
Possible culprit
Add forgotten default parameter for convertValArray
New logic to test
Max/min gets updated during value conversion
To test results of updates
Debug logs
Debug logs
An attempt to provide proper sequence index
Debug logs
An attempt to provide proper sequence index - now magic for resetting
Debug logs
Debug logs
Debug logs
Trying to perform correct updates
Trying to perform correct updates - seqNum woes fight
COMMIT after INSERT performs 'mark extent as invalid' operation - investigating
To test: cut setting of CPInfo upon commit from DML processor
It may be superfluous as write engine does that too
Debug logs
Debug logs
Better interface for CPMaxMin
Old interface forgot to set isBinaryColumn field
Possible fix for the problems
I forgot to reassign the value in cpinfoList
Debug logs
Computation of 'binary' column property
logs indicated that it was not set in getExtentCPMaxMin, and it was impossible to compute there so I had to move that into writeengine.
To test: code to allow cross-extent insertion
To test: removed another assertion for probable cause of errors
Debug logs
Dropped excessive logs
Better reset code
Again, trying to fix ordering
Fixing order of rowids for LBID computation
Debug logs
Remove update of second LBID in split insert
I have to validate incorrect behaviour for this test
Restoring the case where everything almost worked
Tracking changes in newly created extents
Progress keeping commit
Fixing build errors with recent server
An ability to get old values from blocks we update
Progress keeping commit
Adding analysis of old values to write engine code.
It is needed for updates and deletes.
Progress keeping commit
Moving max/min range update from convertValArray into separate function with simpler logic.
To test and debug - logic is there
Fix build errors
Update logic to debug
There is a suspicious write engine method updateColumnRecs which
receives a vector of column types but does not iterate over them
(otherwise it will be identical to updateColumnRec in logic).
Other than that, the updateColumnRec looks like the center of all
updates - deleteRow calls it, for example, dml processor also calls it.
Debug logs for insert bookkeeping regression
Set up operation type in externally-callable interface
Internal operations depend on the operation type and consistency is what matters there.
Debug logs
Fix for extent range update failure during update operation
Fix build error
Debug logs
Fix for update on deletion
I am not completely sure in it - to debug.
Debug log
writeColumnRec cannot set m_opType to UPDATE unconditionally
It is called from deleteRow
Better diagnostics
Debug logs
Fixed search condition
Debug logs
Debugging invalid LBID appearance
Debug logs - fixed condition
Fix problems with std::vector reallocation during growth
Fix growing std::vector data dangling access error
Still fixing indexing errors
Make in-range update to work
Correct sequence numbers
Debug logs
Debug logs
Remove range drop from DML part of write engine
A hack to test the culprit of range non-keeping
Tests - no results for now
MTR-style comments
Empty test results
To be filled with actual results.
Special database and result selects for all tests
Pleasing MTR with better folder name
Pleasing MTR - testing test result comparison
Pleasing MTR by disabling warnings
All test results
Cleaning up result files
Reset ranges before update
Remove comments from results - point of failure in MTR
Remove empty line from result - another MTR failure point
Probably fix for deletes
Possible fix for remaining failed delete test
Fix a bug in writeRows
It should not affect delete-with-range test case, yet it is a bug.
Debug logs
Debug logs
Tests reorganization and description
Support for unsigned integer for new tests
Fix type omission
Fix test failure due to warnings on clean installation
Support for bigint to test
Fix for failed signed bigint test
Set proper unsignedness flag
Removed that assignment during refactoring.
Tests for types with column width 1 and 2
Support for types in new tests
Remove trailing empty lines from results
Tests had failed because of extra empty lines.
Remove debug logs
Update README with info about new tests
Move tests for easier testing
Add task tag to tests
Fix invalid unsaigned range check
Fix for signed types
Fix regressions - progress keeping commit
Do not set invalid ranges into valid state
A possible fix for mcs81_self_join test
MCOL 2044 test database cleanup
Missing expected results
Delete extraneous assignment to m_opType
nullptr instead of NULL
Refactor extended CPInfo with TypeHandler
Better handling of ranges - safer types, less copy-paste
Fix logic error related to typo
Fix logic error related to typo
Trying to figure out why invalid ranges aren't displayed as NULL..NULL
Debug logs
Debug logs
Debug logs
Debug logs for worker node
Debug logs for worker node in extent map
Debugging virtual table fill operation
Debugging virtual table fill operation
Fix for invalid range computation
Remove debug logs
Change handling of invalid ranges
They are also set, but to invalid state.
Complete change
Fix typo
Remove unused code
"Fix" for tests - -1..0 instead of NULL..NULL for invalid unsigned ranges
Not a good change, yet I cannot do better for now.
MTR output requires tabs instead of spaces
Debug logs
Debug logs
Debug logs - fix build
Debug logs and logic error fix
Fix for clearly incorrect firstLBID in CPInfo being set - to test
Fix for system catalog operations suppot
Better interface to fix build errors
Delete tests we cannot satisfy due to extent rescan due to WHERE
Tests for wide decimals
Testing support for wide decimals
Fix for wide decimals tests
Fix for delete within range
Memory leak fix and, possible, double free fix
Dispatch on CalpontSystemCatalog::ColDataType is more robust
Add support for forgotten MEDINT type
Add forgottent BIGINT
empty() instead of size() > 0
Better layout
Remove confusing comment
Sensible names for special values of seqNum field
Tests for wide decimal support
Addressing concerns of drrtuy
Remove test we cannot satisfy
Final touches for PR
Remove unused result file
Create tables and schemas with lower case name only if the flag is set.
During operations, convert to lowercase in plugin. Byt the time a query gets to ExeMgr, DDLProc etc., everything must be lower case if the flag is set, and undisturbed if not.
* TEXT and BLOB now have separate identifiers internally
* TEXT columns are identified as such in system catalog
* cpimport only requires hex input for BLOB, not TEXT
* DML writes for multi-block dictionary (blob) now works
* PrimProc fixed so that the first block in multi-block is read
correctly
* Performance optimisation (removed string copy into stack) for new
dictionary entries
This patch does the following:
* Fix a year storage issue in the datestamp for the columnstore tables
metadata table (note that any previous tables will still have the
incorrect year stored).
* Expose the table creation date in CalpontSystemCatalog::getTables()
* Add an INFORMATION_SCHEMA table listing the tables in ColumnStore
(similar to systable)
* Add an INFORMATION_SCHEMA table listing the columns in ColumnStore
(similar to syscolumn)
* Add an INFORMATION_SCHEMA table listing the extents in ColumnStore
(similar to the editem tool but with additional file information)
* Modifies the build system and scripts to support the new tables
Date limit of year 1400 was used due to Boost's limits.
This patch strips out the use of Boost for date handling and sets the
lower limit to year 1000.
