The three PostgreSQL projects for this year’s Google Summer of Code are on track, and making good progress. All projects expect to finish on time.
Performance Farm
The data gathering for performance farm members is completed, as well as the new implementation for the JSON data transfer. The project iteratively updated it’s goals, and adjusted for newly identified UI issues.
Current work centers around making the website more pretty and useful, as well as reducing the number of used JavaScript libraries. The next step is presenting the work to the PostgreSQL Community for broader feedback.
PL/Java build system
The PL/Java project has just merged (PR #288) the first major pull request of new code from GSoC, creating a new plugin for the Maven build system that allows its actions to be guided by script snippets clearly exposed in the build files.
The same effect was formerly achieved by a workable but brittle combination of an existing Maven plugin that could handle most of the build requirements with another plugin that was able to run Ant, which was able to run scripts. That resulted in a non-ideal division of labor, where a good deal of build logic was hidden away inside plugins, while some parts were exposed in script out of necessity, rather than because they were interesting or likely to need adjustment.
This pull request proves the concept of a new plugin where the hardcoded Java portions are the uninteresting building blocks, and the overall logic of the build is clearly exposed in script.
For now, the new plugin is used to retire the maven-javadoc-plugin and remove the constraints it had imposed on the project’s javadocs (such as the need for absolute URLs for intermodule references, making the resulting tree hard to preview or relocate).
Work continues to reimplement the C native build and retire the nar-maven-plugin and maven-antrun-plugin, to be delivered in a future PR.
WAL-G Performance
We’ve just completed the decoupling of the complex WAL-G internal class. Thanks to it, the new functionality developed in July for a more intelligent backup creation process can now be safely integrated. This feature involves major changes so it requires time to verify that everything is working as expected. We plan to finish the integration in parallel with working on other features.
Currently, we are working on merging the new series of commands for the WAL archives that have been uploaded to storage. These commands will allow end users to analyze the storage for any missing WAL segments that may prevent performing a PITR. Also, Dan now is in the process of implementing the last feature and he expects to finish it on time.
Thanks to all mentors for the status update!