Semantic MediaWiki and related extensions
|
Benchmark tests are to use PHPUnit
as integration platform and do not always represent the best tool for a performance comparison (as it depends on environmental factors such as hardware and software constraints which might not be under the control of the tester) but it can help to identify performance regressions among newly introduced features that run with the same environmental specification.
Benchmarks are not performed in isolation and therefore run in concert with the MediaWiki
infrastructure to determine the overall performance impact during execution.
When using git
, it is relatively easy to run tests and see if a change introduces a significant regression or improvement in terms of performance over the existing master
branch by comparing test results of the master
against a feature
branch.
The definition of what benchmarks are executed is specified by a JSONScript
found in the TestCases
directory. Supported types are:
import
to import data from an external sourcecontentCopy
copy content from an internal sourceeditCopy
edit content from an internal sourcejob
running selected jobsmaintenance
running selected maintenance scriptsquery
executing #ask
queriesRunning composer benchmark
from the Semantic MediaWiki base directory should output something similar to what can be seen below.