Merge branch 'adding-new-tests' into 'master'

Add a new test which runs through the examples

See merge request Monash/tupak!44

.gitlab-ci.yml

@@ -21,11 +21,13 @@ exitcode-jessie:
     - pip install -r requirements.txt
     - pip install coverage
     - pip install coverage-badge
+    - python setup.py install
     - coverage erase
     - coverage run --include=tupak/* -a test/prior_tests.py
     - coverage run --include=tupak/* -a test/tests.py
     - coverage run --include=tupak/* -a test/waveform_generator_tests.py
     - coverage run --include=tupak/* -a test/noise_realisation_tests.py
+    - coverage run --include=tupak/* -a test/example_tests.py
     - coverage html
     - coverage-badge -o coverage_badge.svg -f
   artifacts:

examples/injection_examples/create_your_own_time_domain_source_model.py

@@ -64,7 +64,7 @@ prior['phase'] = tupak.prior.Uniform(-np.pi/2, np.pi/2, r'$\phi$')
 
 
 # define likelihood
-likelihood = tupak.likelihood.Likelihood(IFOs, waveform)
+likelihood = tupak.likelihood.GravitationalWaveTransient(IFOs, waveform)
 
 # launch sampler
 result = tupak.sampler.run_sampler(likelihood, prior, sampler='dynesty', npoints=1000,

examples/other_examples/alternative_likelihoods.py

@@ -20,7 +20,7 @@ outdir = 'outdir'
 # use of the `tupak` waveform_generator to make the signal (more on this later)
 # But, one could make this work without the waveform generator.
 
-class GaussianLikelihood():
+class GaussianLikelihood(tupak.likelihood.Likelihood):
     def __init__(self, x, y, waveform_generator):
         self.x = x
         self.y = y

requirements.txt

@@ -2,7 +2,7 @@ future
 dynesty
 corner
 numpy
-matplotlib>=1.4
+matplotlib>=2.0
 scipy
 gwpy
 pandas

test/example_tests.py

+import unittest
+import os
+import shutil
+import logging
+import subprocess
+
+
+class Test(unittest.TestCase):
+    outdir = 'outdir'
+    dir_path = os.path.dirname(os.path.realpath(__file__))
+    dir_path = os.path.abspath(os.path.join(dir_path, os.path.pardir))
+
+    @classmethod
+    def setUpClass(self):
+        if os.path.isdir(self.outdir):
+            try:
+                shutil.rmtree(self.outdir)
+            except OSError:
+                logging.warning(
+                    "{} not removed prior to tests".format(self.outdir))
+
+    @classmethod
+    def tearDownClass(self):
+        if os.path.isdir(self.outdir):
+            try:
+                shutil.rmtree(self.outdir)
+            except OSError:
+                logging.warning(
+                    "{} not removed prior to tests".format(self.outdir))
+
+    def test_examples(self):
+        """ Loop over examples to check they run """
+        examples = ['examples/injection_examples/basic_tutorial.py',
+                    'examples/injection_examples/change_sampled_parameters.py',
+                    'examples/injection_examples/marginalized_likelihood.py',
+                    'examples/injection_examples/create_your_own_time_domain_source_model.py',
+                    'examples/other_examples/alternative_likelihoods.py',
+                    ]
+        for filename in examples:
+            print("Testing {}".format(filename))
+            subprocess.check_call(['python', filename, '--test'])
+
+
+if __name__ == '__main__':
+    unittest.main()
+
+

tupak/result.py

@@ -3,14 +3,14 @@ import os
 import numpy as np
 import deepdish
 import pandas as pd
-import tupak
 
 try:
     from chainconsumer import ChainConsumer
 except ImportError:
     def ChainConsumer():
-        raise ImportError(
-            "You do not have the optional module chainconsumer installed")
+        logging.warning(
+            "You do not have the optional module chainconsumer installed"
+            " unable to generate a corner plot")
 
 
 def result_file_name(outdir, label):
@@ -142,10 +142,11 @@ class Result(dict):
             if label is None:
                 self.parameter_labels[i] = 'Unknown'
         c = ChainConsumer()
-        c.add_chain(self.samples, parameters=self.parameter_labels,
-                    name=self.label)
-        fig = c.plotter.plot(**kwargs)
-        return fig
+        if c:
+            c.add_chain(self.samples, parameters=self.parameter_labels,
+                        name=self.label)
+            fig = c.plotter.plot(**kwargs)
+            return fig
 
     def plot_walks(self, save=True, **kwargs):
         """ Plot the chain walks using chain-consumer
@@ -165,12 +166,13 @@ class Result(dict):
         if save:
             kwargs['filename'] = '{}/{}_walks.png'.format(self.outdir, self.label)
             logging.info('Saving walker plot to {}'.format(kwargs['filename']))
-        if self.injection_parameters is not None:
+        if getattr(self, 'injection_parameters', None) is not None:
             kwargs['truth'] = [self.injection_parameters[key] for key in self.search_parameter_keys]
         c = ChainConsumer()
-        c.add_chain(self.samples, parameters=self.parameter_labels)
-        fig = c.plotter.plot_walks(**kwargs)
-        return fig
+        if c:
+            c.add_chain(self.samples, parameters=self.parameter_labels)
+            fig = c.plotter.plot_walks(**kwargs)
+            return fig
 
     def plot_distributions(self, save=True, **kwargs):
         """ Plot the chain walks using chain-consumer
@@ -190,12 +192,13 @@ class Result(dict):
         if save:
             kwargs['filename'] = '{}/{}_distributions.png'.format(self.outdir, self.label)
             logging.info('Saving distributions plot to {}'.format(kwargs['filename']))
-        if self.injection_parameters is not None:
+        if getattr(self, 'injection_parameters', None) is not None:
             kwargs['truth'] = [self.injection_parameters[key] for key in self.search_parameter_keys]
         c = ChainConsumer()
-        c.add_chain(self.samples, parameters=self.parameter_labels)
-        fig = c.plotter.plot_distributions(**kwargs)
-        return fig
+        if c:
+            c.add_chain(self.samples, parameters=self.parameter_labels)
+            fig = c.plotter.plot_distributions(**kwargs)
+            return fig
 
     def write_prior_to_file(self, outdir):
         """

tupak/sampler.py

@@ -190,6 +190,9 @@ class Sampler(object):
     def run_sampler(self):
         pass
 
+    def _run_test(self):
+        raise ValueError("Method not yet implemented")
+
     def check_cached_result(self):
         """ Check if the cached data file exists and can be used """
 
@@ -263,6 +266,18 @@ class Nestle(Sampler):
         self.result.logzerr = out.logzerr
         return self.result
 
+    def _run_test(self):
+        nestle = self.external_sampler
+        self.external_sampler_function = nestle.sample
+        self.external_sampler_function(
+            loglikelihood=self.log_likelihood,
+            prior_transform=self.prior_transform,
+            ndim=self.ndim, maxiter=10, **self.kwargs)
+        self.result.samples = np.random.uniform(0, 1, (100, self.ndim))
+        self.result.logz = np.nan
+        self.result.logzerr = np.nan
+        return self.result
+
 
 class Dynesty(Sampler):
 
@@ -312,6 +327,22 @@ class Dynesty(Sampler):
         self.result.logzerr = out.logzerr[-1]
         return self.result
 
+    def _run_test(self):
+        dynesty = self.external_sampler
+        nested_sampler = dynesty.NestedSampler(
+            loglikelihood=self.log_likelihood,
+            prior_transform=self.prior_transform,
+            ndim=self.ndim, **self.kwargs)
+        nested_sampler.run_nested(
+            dlogz=self.kwargs['dlogz'],
+            print_progress=self.kwargs['verbose'],
+            maxiter=10)
+
+        self.result.samples = np.random.uniform(0, 1, (100, self.ndim))
+        self.result.logz = np.nan
+        self.result.logzerr = np.nan
+        return self.result
+
 
 class Pymultinest(Sampler):
 
@@ -458,7 +489,11 @@ def run_sampler(likelihood, priors=None, label='label', outdir='outdir',
             logging.info("Using cached result")
             return sampler.cached_result
 
-        result = sampler.run_sampler()
+        if utils.command_line_args.test:
+            result = sampler._run_test()
+        else:
+            result = sampler.run_sampler()
+
         result.noise_logz = likelihood.noise_log_likelihood()
         if use_ratio:
             result.log_bayes_factor = result.logz

tupak/utils.py

@@ -530,6 +530,9 @@ def set_up_command_line_arguments():
                         default=['H1', 'L1', 'V1'],
                         help=("List of detectors to use in open data calls, "
                               "e.g. -d H1 L1 for H1 and L1"))
+    parser.add_argument("-t", "--test", action="store_true",
+                        help=("Used for testing only: don't run full PE, but"
+                              " just check nothing breaks"))
     args, _ = parser.parse_known_args()
 
     if args.quite: