Running/Testing EnergyPlus – for Developers[LINK]
Any item mentioned in this section is available at no charge to collaborative or other developers – the documentation, however, may be rudimentary and use of the procedures require some knowledge of command line (Windows) or Linux scripts.
EnergyPlus is rigorously tested during each release cycle and prior to each release. Details on some of the test suites that have been used can be seen at:
Equally important is the testing done by each developer during feature development or changes. For example, on the core development team, developers are charged with executing the entire test suite (≥230 files) for their checkins. In addition, one of the core development team does run the entire test suite with each checkin and compares those results to previous results. Unexpected changes, and certainly crashes, should NOT occur.
Since most modules being developed are aimed at the HVAC or plant features, there is a standard 5-zone template geometry that can be used. This should form the basis of any new additions. The old 3-zone model should not be used. Of course, you may develop your own model.
Developers are also charged with making sure their input file runs an entire weather year, has minimal errors (including max simulation errors) and results compare exactly when design days (preferably winter-summer vs summer-winter) are reversed in consecutive runs (also known as ReverseDD). To assist in ReverseDD testing, each input file should have a “Run Control” object as well as (at least) two design days (winter-summer / summer-winter as the first two).
Input files should report Zone Air temperatures (Zone Mean Air Temperature or Zone/Sys Air Temperature) as well as meters such as electricity and gas (if applicable). Of course, reporting features being implemented should be done as well. These variables will help identify files that have proper “ReverseDD” requirements (failures usually indicate some initialization problems). Developers should try to minimize output file size – if you are running a full annual simulation (as required by your feature), you should NOT report variables at the timestep level.
To compare results, we have a python script (Mathdiff) that is run on the .csv files. It will report (by default) differences < = .001 or < = .5% as “within range” and outside those limits as “differences”. If they are exactly the same (from the .csv precision limits), they will be reported as such.
Developers in the core development team use several methods for running the entire test suite.
One method uses a list of input file names along with an indication of the proper weather file. A program reads this file and produces several batch files which help with not only running the file but comparing them to previous results, building the “composite error” (the .err files from each file run), and other utility features. (The same file can be used in Linux testing)
Another method uses a batch file with options that will allow running old vs. new exes as well as somewhat automating the reverse dd testing.
Still another method uses a simple batch procedure to “run” all files in a folder.
Finally, EP-Launch and “groups” can be used.
To facilitate testing, Environment Variables “values” have been implemented in EnergyPlus and/or script files. To use, one uses the “Set” command and the value as indicated. Environment variable value testing is inherent in F2003 compliant compilers; for others we have written a set of routines that can either be modified or used directly.