Status: Closed (View Workflow)
Squash is off the table for now, there are two strong arguments against it:
1. Using Squash requires changing code in all the storage engines using compression libraries, including 3rd party storage engines and this is problematic because it is very unlikely we will be able to propagate these changes upstream.
2. In the unified API it provides, Squash does not support all the bits and parts our current storage engines use from compression libraries. Tweaking the storage engines code to get rid of API calls that Squash doesn't support is too complicated and amplifies the problem stated at point 1.
Given the above, we should implement this with MariaDB services using an approach like the one below:
1. Add new `--use-compression=snappy,bzip2` option in the server.
Code place: sql/sys_vars.cc
2. Choose one compression library and create a new mariadb service that contains dummy functions for all corresponding functions within the compression library (only functions that are used by the storage engines we support). Make sure the service is built correctly, test it by calling one of its functions from a dynamic plugin’s init function and see it printing something.
For creating services, the instructions are in the libservices/HOWTO file
3. Alter the service so that it works with static plugins as well (we need this for storage engines such as InnoDB). Quickly test that it works with a dummy print as well, from InnoDB for example.
The `#ifdef MYSQL_DYNAMIC_PLUGIN` used within a service is the key to make static plugins have access to the service structure. See how the `debug_sync` service does it to understand how the changes in our service should look like.
4. We need to have the ability to trick storage engines into calling our custom service API instead of the compression library code. Thus, we need to tweak the include paths so that storage engines using this library will include our own header, with our own functions, and this without actually changing any storage engine source code.
After this step is done, we should easily see in a debugger that calls, from a storage engine to the compression library we’ve picked, go through this service.
Code place: cmake/plugin.cmake
5. The server should try to load the compression libraries based on the values passed within the `use-compression` option. If a library is loaded correctly, then the function pointers within the corresponding compression service should be set to point within that compression library.
Once this will be done, we can safely move to writing some proper tests for this service and onto the next compression library
Code place: To be debated when we get there, but a new file ‘sql/compression_libs.cc’ that contains 1 function (called from mysqld.cc) for dinamically binding compression functions within the service structure and all the dummy functions returning
6. Tests should be written for each service that takes care of one compression library.
As per Sergei G suggestion, one test `/mysql-test/suite/compression_libs/bzip2.test` might look something like:
`set global innodb_compression_algorithm = ABC;
create table innodb_compressed(c1 int, b char(20)) engine=innodb row_format=compressed key_block_size=8;
This test should be run with both `--use-compression=bzip2` flag and without
- is blocked by
MDEV-11068 Review which innodb_compression_algorithm to support in binary packages
- In Progress
- is part of
MDEV-12933 sort out the compression library chaos