Skip to content

Latest commit

 

History

History
executable file
·
81 lines (62 loc) · 3.71 KB

CONTRIBUTING.md

File metadata and controls

executable file
·
81 lines (62 loc) · 3.71 KB
Raw

SAS DLPy Developer How-To

Developing DLPy is just like developing any other project on GitHub. You clone the project, do your work, and submit a pull request.

Submitting a Pull Request

Submitting a pull request uses the standard process at GitHub. Note that in the submitted changes, there must always be a unit test for the code being contributed. Pull requests that do not have a unit test will not be accepted.

You also must include the text from the ContributerAgreement.txt file along with your sign-off verifying that the change originated from you.

Testing

For the most part, testing the SAS DLPy package is just like testing any other Python package. Tests are written using the standard unittest package. All test cases are subclasses of TestCase.

SAS DLPy is centered around SAS' deep learning functionalities. Therefore most of the tests include creating models and training some data. A number of resources are located under tests/datasources. Locate these files and folders under your DLPY_DATA_DIR. If you are using a Windows client then also locate the object detection folder under DLPY_DATA_DIR_LOCAL. These two (DLPY_DATA_DIR and DLPY_DATA_DIR_LOCAL) are environment variables and should be set before running the unittests. Further, there are a number of tests that require model files (ending with .h5). They are not located as they do take some space. That said, those files can easily be downloaded online.

The below part is the configurations for the CAS Server.

Since CAS is a network resource and requires authentication, there is some extra setup involved in getting your tests configured to run against your CAS server. Normally this involves setting the following environment variables.

  • CASHOST - the hostname or IP address of your CAS server (Default: None)

  • CASPORT - the port of your CAS server (Default: None)

  • CASPROTOCOL - the protocol being using ('cas', 'http', 'https' or 'auto'; Default: 'cas')

  • CASUSER - the CAS account username (Default: None)

  • CASPASSWORD - the CAS account password (Default: None)

  • CASDATALIB - the CASLib where data sources are found (Default: CASTestTmp)

  • CASMPPDATALIB - the CASLib on MPP servers where the data sources are found (Default: HPS)

  • CASOUTLIB - the CASLib to use for output CAS tables (Default: CASUSER)

  • CASMPPOUTLIB - the CASLib to use for output CAS tables on MPP servers (Default: CASUSER)

Some of these can alternatively be specified using configuration files. The CASHOST, CASPORT, and CASPROTOCOL variables can be specified in a .casrc in your home directory (or in any directory from the directory you are running from all the way up to your home directory). It is actually written in Lua, but the most basic form is as follows:

cashost = 'myhost.com'
casport = 5570
casprotocol = 'cas'

The CASUSER and CASPASSWORD variables are usually extracted from your ~/.authinfo file automatically. The only reason you should use environment variables is if you have a generalized test running account that is shared across various tools.

Finally, the CASDATALIB and CASOUTLIB variables configured where your data sources and output tables reside. Using the CASDATALIB location will make your tests run more efficiently since the test cases can load the data from a server location. If you don't specify a CASDATALIB (or the specified one doesn't exist), the data files will be uploaded to the server for each test (which will result in hundreds of uploads). Most people will likely set them all to CASUSER and create a directory called datasources in their home directory with the contents of the dlpy/tests/datasources/ directory.

Once you have these setup, you can use tools like nosetest to run the suite:

nosetests -v dlpy.tests