Testing¶
By default, your top level devenv/bin directory will contain a script called openmdao_test that uses a Python package called nose to run all of the unit tests for any package that you specify. For example, to run all of the unit tests in the openmdao namespace package, do the following:
bin/openmdao_test openmdao
which should generate output something like this:
......................................................................
.................................................
----------------------------------------------------------------------
Ran 119 tests in 0.888s
OK
To run the full test suite, which includes the openmdao namespace package and other packages from contrib, type:
bin/openmdao_test --all
To get a list of options available with bin/openmdao_test, type bin/openmdao_test --help from the devenv directory. The bin/openmdao_test script uses the nose testing framework internally, so all options available when running nosetests should also be available when using bin/openmdao_test.
Test Coverage¶
There is a Python package called coverage that is accessible through bin/test that makes it easy to determine if your tests cover every line of code in your source files. To get a coverage report for the openmdao package, do the following from the devenv directory:
bin/openmdao_test openmdao --with-coverage --cover-package=openmdao
The report should look something like this:
................................................................................
..
Name Stmts Exec Cover Missing
-----------------------------------------------------------------
openmdao 5 0 0% 2-6
openmdao.lib 0 0 100%
openmdao.lib.components 0 0 100%
openmdao.lib.drivers 0 0 100%
openmdao.lib.drivers.conmindriver 183 179 97% 149, 233-234, 271
openmdao.lib.factories 0 0 100%
openmdao.lib.variables 0 0 100%
openmdao.main 6 3 50% 5-7
openmdao.main.arrayvar 48 47 97% 32
openmdao.main.assembly 103 101 98% 95, 129
openmdao.main.component 47 41 87% 58, 92, 99, 106, 121, 142
openmdao.main.constants 4 4 100%
openmdao.main.constraint 44 43 97% 24
openmdao.main.container 201 185 92% 22-24, 138, 156, 166, 253-254, 276-277, 337, 340, 356, 359, 367-368
openmdao.main.containervar 50 29 58% 29, 38-43, 49-55, 66-72, 82
openmdao.main.driver 18 15 83% 35, 40-41
openmdao.main.exceptions 5 5 100%
openmdao.main.expreval 122 115 94% 27, 32, 36, 40, 85, 177, 222
openmdao.main.factory 6 5 83% 25
openmdao.main.factorymanager 21 16 76% 28, 33-37
openmdao.main.float 70 54 77% 38-41, 49-53, 58-61, 69-73, 105, 120
openmdao.main.hierarchy 49 46 93% 34, 40, 59
openmdao.main.importfactory 28 25 89% 47-49
openmdao.main.int 42 24 57% 31-34, 39-46, 51-54, 59-66
openmdao.main.interfaces 54 54 100%
openmdao.main.logger 9 9 100%
openmdao.main.pkg_res_factory 61 59 96% 88, 114
openmdao.main.string 42 28 66% 31-34, 42-46, 51-54, 62-66
openmdao.main.stringlist 56 40 71% 31-34, 42-46, 51-54, 62-66, 92, 95
openmdao.main.tarjan 58 26 44% 52-71, 78-96, 100
openmdao.main.variable 138 113 81% 22, 54, 65, 73, 101-104, 112, 117, 129, 141, 184, 202, 227, 263, 265-270, 276, 282-285, 289-290
openmdao.main.vartypemap 19 17 89% 42-45
openmdao.main.workflow 56 35 62% 30, 43, 56, 61-75, 79, 86-88, 92
-----------------------------------------------------------------
TOTAL 1545 1318 85%
----------------------------------------------------------------------
Ran 82 tests in 5.678s
OK
The numbers in the Missing column indicate lines or ranges of lines that are not covered by the current set of tests.
If you edit source code, the coverage data may become inaccurate, so you should clear the coverage database by issuing the following command:
bin/openmdao_test openmdao --cover-erase
Adding New Tests¶
Generally, you should write your tests using Python’s unittest framework if possible, although the nose package is able to discover and run tests that do not use unittest.
The following is a simple example of a unit test written using the unittest framework.
import unittest
class TestSomeFunctions(unittest.TestCase):
def setUp(self):
# put setup code here. It will run at the beginning of each
# test function (function with name that starts with 'test')
def tearDown(self):
# put code here that you want to be run after each test function
# is completed
def testfunct1(self):
# a test function
def test_some_stuff(self):
# another test function
if __name__ == '__main__':
unittest.main()
The unittest.TestCase class provides a number of functions to test output during a test function. For example:
- self.assertTrue(expr[,msg])
- Test will fail if expr does not evaluate to True.
- self.assertEqual(val1,val2)
- Test will fail if val1 != val2
- self.assertNotEqual(val1,val2)
- Test will fail if val1 == val2
- self.assertAlmostEqual(val1,val2[,places=7])
- Test will fail if val1 differs from val2 by more than a small number of decimal places.
- self.fail([msg])
- Test will fail and display the given message.
Often in a test you will want to make sure that a specific exception is raised when a certain thing happens, and usually you want to verify that the error message contains certain information. The unittest framework provides an assertRaises function that does part of this job, but it does not allow you to check the error message. So the preferred way to test exceptions is shown in the code below. In this example, we will assume that the exception we are interested in is a ValueError. Note that we would place our test function inside of our unittest.TestCase derived class.
def test_myexception(self):
try:
# perform action here that should raise exception
except ValueError, err:
self.assertEqual(str(err), "this should be my expected error message")
else:
self.fail('expected a ValueError')
Note that the else block after the except is important because we want the test to fail if no exception is raised at all. Without the else block, the test would pass if no exception were raised.
Test File Locations¶
Unit tests are typically placed in a test subdirectory within the directory where the source code being tested is located. For example, the test files for openmdao.main are located in openmdao.main/src/openmdao/main/test.
Testing Code in the Documentation¶
The OpenMDAO documentation includes quite a few examples of Python code. These examples are used to explain how to use features of the OpenMDAO API as well as how to develop new components and plugins. Thus, it is imperative that any code examples included in the documentation (particularly the User’s Guide and the Developer’s Guide) be tested to ensure that the code is error-free.
Fortunately, there are tools built into the Sphinx reStructuredText processor that make the process of testing code samples much easier. Sphinx includes a builder called doctest as a plugin in its standard library. Instead of building a readable document as is done by the HTML builder, the doctest builder scans the documentation files for code blocks and tests them.
There are two types of code examples found in the documentation. The first type is a block of code as would be found in a Python script:
from openmdao.examples.enginedesign.engine import Engine
my_engine = Engine("new_engine")
The second type of code example is a copy of an interactive shell session:
>>> print "Hello!"
Hello!
Both of these types of code samples must be tested, although the way to accomplish this differs slightly. Unlike the doctest module built into Python, which can handle only shell session blocks, the doctest builder included with Sphinx can handle both of these code sample types.
Testing Code Blocks¶
The doctest builder in Sphinx provides a flexible environment to easily and effectively test blocks of code using a set of directives. The test code directive is used to mark a block of text that should be tested and treated as Python code. It is not always possible to execute a standalone block of Python code without first executing some preliminary code containing any prerequisites (e.g., imports.) The testsetup block makes it possible to run the preliminary code. This block is hidden by default, so it does not show up in the generated HTML documentation. Additionally, there is a testoutput block, where any output that is generated by the testcode block should be included so that it can be tested.
A simple example of how to implement these three blocks is shown here:
.. testsetup:: Group1
# Put any preliminary code that needs to run before the sample code.
# This block does not show up when Sphinx builds the HTML
.. testcode:: Group1
# This is the sample code that shows up in your docs
.. testoutput:: Group1
# If your code block outputs anything when executed, then that output
# needs to go in this block.
Group1 is a label that we’ve given this set of blocks. Note that you can have multiple labels in your documents. Note also that the testsetup and testoutput blocks are both optional; some code examples don’t need either. You can have multiple testcode blocks for a single testsetup block. The environment is preserved across all of the testcode blocks in a given group, so that the code executed in the first testcode block in Group1 affects all later blocks in Group1.
The label is optional, and defaults to default when not explicitly defined.
There is one other directive of note. The doctest directive is used to mark blocks of interactive shell Python code. Note that if the directive is omitted, the doctest builder can often find the Python blocks by itself, but it is still good form to include it.
.. doctest:: Group2
>>> # This code is tested
The doctest blocks share their workspace in a similar manner as the testcode blocks. There are other options that can be enabled for the doctest blocks, but so far the default ones have been fine.
More details on using the doctest builder can be found here: http://sphinx.pocoo.org/ext/doctest.html
Including Code Straight from the Source¶
There are times that it is more efficient to directly include code from a source file. There is a directive built into Sphinx that enables this – the literalinclude block:
.. literalinclude:: ../../openmdao.examples/openmdao/examples/enginedesign/engine_wrap_c.py
:start-after: engine_weight = 0.0
:end-before: # end engine.py
:language: python
The first line contains the relative path location of the file that is to be included. Since you rarely want to include an entire file, the options start-after and end-before can be used to define the bookends that bound the block of text to be included.
Sometimes, it makes more sense to grab specific lines from a file. This can also be done with the lines option.
.. literalinclude:: ../../openmdao.examples/openmdao/examples/enginedesign/engine_wrap_c.py
:lines: 3,7-12,45
:language: python
More details on the literalinclude directive can be found at http://sphinx.pocoo.org/markup/code.html.
Helpful Tips¶
Indentation is not preserved between code blocks in the same group. This means that all functions and class definitions effectively close when the block ends. If you need to show code from the middle of a function class, you may have to get creative in what you place in your testsetup block (e.g., defining self as something.)
Tracebacks don’t have to be accurately reproduced (and they can’t be anyway.) Handle these the same way they are in doctest, that is:
>>> my_engine.set("RPM",7500) Traceback (most recent call last): ... TraitError: Trait 'RPM' must be a float in the range [1000.0, 6000.0] but attempted value is 7500
Care should be taken to assure that each block of code is being tested. One way to do this is to purposefully introduce an error into a block to verify that it is caught.
Be wary of including code by line number. If the source file is changed, and lines are added or removed, then the included code might not be what was intended.
To include a numerical example in a doctest block, you can use elipses to match the output to a specific tolerance. For example, this block of text passes:
>>> import numpy >>> numpy.pi 3.14...
Running the Tests¶
The build procedure currently generates a script for testing the code in the documents. This can be run by going to the devenv folder, and running:
[unix_prompt]$ bin/testdocs
The following output indicates a successful test:
Doctest summary
===============
37 tests
0 failures in tests
0 failures in setup code
build succeeded.
If any tests fail, they will also be noted in this summary, and specific tracebacks will be given for each failure earlier in the output.
The document testing is also part of the full testing process and is executed as part of the normal bin/openmdao_test.
