My personal and professional life

2020-01-17

MySQL 8.0.19 suite released

MySQL 8.0.19 suite including Community Server, Router and Shell, was released on Monday (01/13). The focus of these releases is the new InnoDB ReplicaSet functionality that simplifies management of classic MySQL replication with single primary and multiple secondaries in the same way as setting up Group Replication with InnoDB Cluster. You could read more about it in the article The all new MySQL InnoDB ReplicaSet. Apart from this there are some other improvements for which you could read below.

SQL syntax

The server now understands the more general and standard ALTER TABLE … DROP/ALTER CONSTRAINT syntax for which I tweeted earlier. It's a feature request since MySQL 4.1 from more than 15 years ago. You could find more in Foreign Keys and Other Constraints section of ALTER TABLE article in documentation.

Table value constructors are another improvement towards the standard allowing generation of rows with the VALUES statement (not to be confused with VALUES keyword for INSERT statements). Due to name conflict with the VALUES function it's necessary to use the verbose form with ROW constructor. This basically means the standalone statement VALUES ROW(1), ROW(2), ROW(3) is now properly understood by the server. VALUES statement could be used also in unions, joins, as derived table and other cases where you could employ SELECT. See the VALUES statement article in documentation for more details.

The new TABLE statement implements another standard feature - explicit table clauses. It could be used like SELECT, because TABLE t is just like SELECT * FROM t, so it could be used in unions, with INTO, as IN subquery and elsewhere where SELECT could be employed.

An interesting new addition is the LIMIT clause for recursive CTEs, which allows the debugging of expressions generating too many results. The idea is simply to limit the result from a recursive CTE to some reasonable number (e.g. 10, 100, etc.), so you could see the generated output and be able to understand how and why it's generated like this. This new features is covered in the article A new, simple way to figure out why your recursive CTE is running away and in Limiting Common Table Expression Recursion section of WITH (Common Table Expressions) in documentation.

It's is now possible to use an alias for the new row and optionally columns with INSERT ... ON DUPLICATE KEY UPDATE statement following the VALUES or SET clause with the use of AS keyword. I guess this is the first step to removing VALUES function (see above) for better standards support. You could find more in the article for the ON DUPLICATE KEY UPDATE Statement.

Datatype changes

The TIMESTAMP and DATETIME types could now include timezone information as offset (e.g. +02:00 or -08:00) appended to the value. The details are in the article The DATE, DATETIME, and TIMESTAMP Types in documentation.

The YEAR(4) data type (i.e. with explicit display width) and the (undocumented) UNSIGNED attribute for YEAR are now deprecated and would be removed in future. This follows other deprecations of display widths and other type attributes from 8.0.17 for which I wrote before, so developers should check and correct any SQL scripts using deprecated types and attributes. With this release statements that print data type definitions (e.g. SHOW CREATE TABLE) no longer show display width for integers with exception for TINYINT(1) and ZEROFILL attribute.

Other changes in server

User-defined functions (UDFs) can now determine character set and collation of string argument as I noticed earlier. You could find more about this in the article A Tale of UDFs with Character Sets.

There are now new standard INFORMATION_SCHEMA views related to roles like ADMINISTRABLE_ROLE_AUTHORIZATION, APPLICABLE_ROLES, ENABLED_ROLES, ROLE_COLUMN_GRANTS, ROLE_ROUTINE_GRANTS and ROLE_TABLE_GRANTS.

InnoDB engine now supports efficient sampling of data for histogram statistics. Sampling no longer requires full table scan, which could be costly for large tables, but samples data pages from the clustered index instead. I should try this on first opportunity.

Function JSON_SCHEMA_VALID now provides detailed information about the reasons for validation failures by using SHOW WARNINGS.

Other notable changes is the possibility to lock user accounts after FAILED_LOGIN_ATTEMPTS login attempts for PASSWORD_LOCK_TIME days (see Failed-Login Tracking and Temporary Account Locking), compression in X protocol (see Connection Compression with X Plugin) and many more as you could find from the release notes.

Bugs fixed in server

The previously reported problem with EXPLAIN ANALYZE not executing and reporting on subqueries in the select list has been fixed (see bug 97296).

Workbench

As I read in the release notes of the new version that "The following MySQL server language features are now supported: value references in INSERT statements, new options (TLS version, cipher suite, compression, and privilege check) in CHANGE MASTER TO replication statements, random passwords in CREATE USER and ALTER USER statements, EXPLAIN ANALYZE, and binary collation names.", I was hopeful that the application is finally catching up with the server, but that was until I reviewed the new SQL syntax (see above).

Two of the bugs I reported previously were fixed - 97416 (for EXPLAIN ANALYZE) and 97281 (for ANALYZE TABLE ... UPDATE HISTOGRAM), but I opened two new for TABLE (see bug 98263) and VALUES (see bug 98266) statements. Such problems should be obvious, so it still strange to me how they continue appearing.

There is a new menu item "Copy Row (with names, tab separated)" for copying headers and rows of queried data with tabs as separators that makes it easy to move results into Excel or LibreOffice Calc.

Happy using and experimenting with the new MySQL releases!

No comments: