update doc, test

.travis.yml

@@ -30,7 +30,6 @@ install:
   #- conda install --yes python=$TRAVIS_PYTHON_VERSION numpy matplotlib nose coverage cython scipy
   - conda install --yes --file requirements.txt
   - conda install --yes --file requirements-dev.txt
-  #- conda install --yes cython 
   - pip install .
 # # command to run tests, e.g. python setup.py test
 script:

doc/source/overview.txt

@@ -14,6 +14,10 @@
     * Multitapering is also available
 
 
-
+The targetted audience is diverse. Although the use of power spectrum of a
+signal is fundamental in electrical engineering (e.g. radio communications,
+radar), it has a wide range of applications from cosmology (e.g., detection of
+gravitational waves in 2016), to music (pattern detection) or biology (mass
+spectroscopy). 
 
 

doc/source/quickstart.rst

@@ -30,7 +30,7 @@ We can analyse this data using one of the Power Spectrum Estimation method provi
 
 Here, we have created an object Periodogram. No computation has been performed yet. To run the actual estimation, you can use either::
 
-    p() # now you run the estimation
+    p() 
 
 or::
 
@@ -40,14 +40,19 @@ and finally, you can plot the resulting PSD::
 
     p.plot(marker='o')  # standard matplotlib options are accepted
 
+.. warning::
+
+    .. versionchanged:: 0.6.7 you do not need to use p() or p.run() anymore. It will be
+       called automatically when using p.plot() or when you access to the *psd*
+       attribute.
+
 
 .. plot::
     :width: 80%
 
     from spectrum import *
     data = data_cosine(N=1024, A=0.1, sampling=1024, freq=200)
     p = Periodogram(data, sampling=1024) #here you just define the PSD estimate 
-    p() # now you run the estimation
     p.plot(marker='o')
 
 Since the data is purely real, the PSD (stored in p.psd) is a onesided PSD, with positive frequencies only. If the data were complex, the two-sided PSD would have been computed and plotted. For the real case, you can still plot a two-sided PSD by setting the sides option manually::
@@ -61,7 +66,6 @@ Since the data is purely real, the PSD (stored in p.psd) is a onesided PSD, with
     from spectrum import *
     data = data_cosine(N=1024, A=0.1, sampling=1024, freq=200)
     p = Periodogram(data, sampling=1024) #here you just define the PSD estimate 
-    p() # now you run the estimation
     p.plot(marker='o', sides='twosided')
 
 

setup.py

@@ -9,7 +9,7 @@
 
 _MAJOR               = 0
 _MINOR               = 6
-_MICRO               = 6
+_MICRO               = 7
 version              = '%d.%d.%d' % (_MAJOR, _MINOR, _MICRO)
 release              = '%d.%d' % (_MAJOR, _MINOR)
 

src/spectrum/psd.py

@@ -452,9 +452,9 @@ def _setData(self, data):
 
     def _getPSD(self):
         if self.__psd is None:
-             logging.warning('PSD not yet computed. call the object to estimate the PSD.')
-        else:
-            return self.__psd
+            logging.debug('Computing PSD.')
+            self()
+        return self.__psd
     def _setPSD(self, psd):
         # Reset the sides attribute depending on the datatype
         # if a user sets the PSD manually, the only check available is that

test/test_arma.py

@@ -19,7 +19,7 @@ def test_arma_values():
         3.22899406+6.39981425j,  3.16557650+5.92783737j,
         3.47120865+5.48246963j,  2.79508215+3.3238971j ,
         2.13174602+1.51034329j]), decimal=4)
-    
+
 
 def test_arma():
     """arma, check that rho is correct (appendix 10.A )and reproduce figure 10.2"""
@@ -73,7 +73,6 @@ def test_arma2psd():
 
 def test_parma():
     p = parma(marple_data, 4, 4, 30, NFFT=4096)
-    p()
     p.plot()
     print(p)
 

test/test_burg.py

@@ -66,7 +66,6 @@ def test_burg_criteria():
 
 def test_pburg():
     p = pburg(marple_data, 15, NFFT=4096)
-    p()
     p.plot(sides='centerdc')
     print(p)
     # test real case