CCP4I2 Developers: Project Based Testing


Creating Source Projects
Running Tests Non-Graphically


Project based testing provides automatic testing of the non-graphical aspects of tasks and also provide test data and indications of the normal use for manual testing of the gui. Ideally we need one test-project per task that exercises all the main functionality of task. Some test projects that run through a series of tasks in a realistic structure solution would also be useful for testing and as examples for demonstrations and tutorials.

The automatic mechanism tests plugin scripts by rerunning all of the jobs from a source project and comparing the output from the test jobs with that from the source jobs. The source project can be set up graphically in CCP4i2 but the test procedure is normally run non-graphically by the testi2sys script. When the test jobs are run they take their input files from the source project directory even where the job input could be taken from a preceeding job in the test project (but sub-jobs in pipelines do pass files through the pipeline in the expected fashion).

After a test job has finished an automatic comparison is made of the outputs. For each job run the test system calls CPluginScript.assertSame() to do a comparison of the output data of the source and the test job as taken from the params.xml file of the two job runs. This method iterates over the data objects in the outputData container and calls the assertSame() method passing it the 'other' job run data value. This method returns a CErrorReport which will contain a list of difference between the two values. Only the assertSame() for file objects and performance indicator objects do meaningful tests.

There is usually one performance indicator parameter per task whose main function is a simple indication of progress shown in the Job list (see here). The same parameter turns out to be useful for comparison in testing. The implementations of CPerformanceIndicator.assertSame() usually compares the float,int or string members of the class - there is an allowed tolerance and it can be specified that a changed value may be an improvement (e.g. a decrease in test Rfactor compared to source is reported as a warning not an error). It is possible to implement a performance indicator class so that it serves as a test class but does not put anything in the database or Job list (eg. CAtomCountPerformance).

The test on output files is that they exist and then dependent on file type:

CDataFilecompare file checksum and/or optionally file size
CPdbDataFilecompare file checksum with ad hoc editing out of problem meta data
CMtzDataFilecompare cell,spaceGroup,resolutionRange, number of reflections and column types/labels

File comparison is tricky - for example a checksum test on PDB files output by Molrep failed because of the date in a HEADER line - this has been adressed but more, similar gotchas are expected. This is a work in progress - please talk to Liz about any false calls and suggestions.

The com.txt input file for the program run by each task is also compared.

The results are listed to the terminal and saved to a file in the test project called CCP4_test-year-month-day-hour-minute.log.

Creating Source Projects

A source project can be created as a normal project within CCP4i2. All jobs should be reasonably quick to run and complete successfully. The source project is exported by the Export option in the Manage projects window (accessed from the Projects menu). This creates a zip file which can be sent to Liz for inclusion in the automatic test suite. Before exporting the test project please test it from the Manage projects window by selecting the appropriate project and using the Rerun test project option - you will be required to specify a directory for output. Note that Rerun test project is only available if you are running in DEVELOPER mode (if necessary edit .CCP4I2/configs/ccp4i2_cofig.params.xml and set developer to True). A test directory will be created and the test jobs will then proceed automatically one at a time. The results from the comparison of the source tasks and the test tasks are shown in real time in a web browser window. You can also open the project window for the test project to view progress. The test results are also saved to a file in the test project directory called CCP4_test-year-month-day-hour-minute.log. The log file should list each job that is run (only the top level jobs) and the output data from the job looking like:

Job: 1 molrep_mr
XYZOUT               -OK- CPdbDataFile:300 Compared objects are the same
Job: 35 buccaneer_build_refine
XYZOUT               -OK- CPdbDataFile:300 Compared objects are the same
FPHIOUT              -ERROR- CMapCoeffsDataFile:308 Files failed checksum comparison/Users/lizp/Desktop/testing/demo8/CCP4_JOBS/job_35/FPHIOUT.mtz : /Users/lizp/Desktop/testing/demo8_rerun/CCP4_JOBS/job_35/FPHIOUT.mtz

The items on the line for each item of data are: the parameter name, the test status (OK,WARNING or ERROR), the data class, the error code (error code 300 is 'OK') and further details. If this test run fails to run of gives false fails please discuss with Liz.

Note that if you look at the test project in the Project Viewer the imported files will be absent.

Running Tests Non-Graphically

A test project or a suite of test projects can be run using the ccp4i2/bin/testi2sys script. This script works with a temporary database and unpacks the source projects into the outputDirectory and loads them into the database. It then reruns the source projects creating a test project. Each test project is given the name of the source project with the date and time and '_rerun' appended. The test project directories created in the outputDirectory are given the test project name. The jobs from the source project are then run one at a time with progress and errors listed to the terminal and saved to file CCP4_test.....log in the outputDirectory. The command line arguments:
-p pathpath of a compressed source project file or a directory containing them
-d directorydirectory where the compressed source project is unpacked and temporary database and test projects are created
-oOverwrite any existing temporary database
-runMode flag 0 :no testing, 1 :run jobs, 2:run tests, 3:run jobs and tests. Default: 3
-verbosity flagReport 0:fails only, 1:jobs and parameters tested, 3:jobs, parameters and parameter values. Default: 1
-testSubJobs flag0:no, 1:for failed jobs, 2:for all jobs. Default:1

Last modified: Thu Jan 29 12:22:01 GMT 2015