Merge branch 'master' into #642

Author: jmeyers314

CHANGELOG.md

@@ -27,12 +27,12 @@ API Changes
   issue, and we can add them back. (#218)
 - Made all returned matrices consistently use numpy.array, rather than
   numpy.matrix.  We had been inconsistent with which type different functions
-  returned, so now all matrices are numpy.arrays.  If you rely on the 
+  returned, so now all matrices are numpy.arrays.  If you rely on the
   numpy.matrix behavior, you should recast with numpy.asmatrix(M). (#218)
-- Deprecated CorrelatedNoise.calculateCovarianceMatrix, since it is not used 
+- Deprecated CorrelatedNoise.calculateCovarianceMatrix, since it is not used
   anywhere, and we think no one really has a need for it. (#630)
 - Officially deprecated the methods and functions that had been described as
-  having been removed or changed to a different name.  In fact, many of them 
+  having been removed or changed to a different name.  In fact, many of them
   had been still valid, but no longer documented.  This was intentional to
   allow people time to change their code.  Now these methods are officially
   deprecated and will emit a warning message if used. (#643)
@@ -48,14 +48,14 @@ New Features
 
 - Made all GalSim objects picklable unless they use fundamentally unpicklable
   things such as lambda expressions.  Also gave objects better str and repr
-  representations (str(obj) should be descriptive, but relatively succinct, 
+  representations (str(obj) should be descriptive, but relatively succinct,
   and repr(obj) should be unambiguous).  Also made __eq__, __ne__, and __hash__
   work better. (#218)
 - Added ability to set the zeropoint of a bandpass on construction.  (Only
   a numeric value; more complicated calculations still need to use the method
   `bandpass.withZeropoint()`.) (#218)
 - Added ability to set the redshift of an SED on construction. (#218)
-- Updated CorrelatedNoise to work with images that have a non-trivial WCS. 
+- Updated CorrelatedNoise to work with images that have a non-trivial WCS.
   (#501)
 - Added new methods of the image class to simulate detector effects:
   inter-pixel capacitance (#555) and image quantization (#558).
@@ -79,7 +79,7 @@ New Features
   rendering process compared to brute force evaluation for chromatic objects
   with basic properties that are wavelength-dependent. (#618)
 - Added new `ChromaticAiry` and `ChromaticOpticalPSF` classes for representing
-  optical PSFs.  These new classes allow the diffraction effects and (in the 
+  optical PSFs.  These new classes allow the diffraction effects and (in the
   latter case) aberrations to be wavelength-dependent. (#618)
 - Enable initializing a DES_PSFEx object using a pyfits HDU directly instead
   of a filename. (#626)
@@ -100,7 +100,7 @@ Bug Fixes and Improvements
 - Fixed a bug in UncorrelatedNoise where the variance was set incorrectly.
   (#630)
 - Changed the implementation of drawing Box and Pixel profiles in real space
-  (i.e. without being convolved by anything) to actually draw the surface 
+  (i.e. without being convolved by anything) to actually draw the surface
   brightness at the center of each pixel.  This is what all other profiles do,
   but had not been what a Box or Pixel did. (#639)
 - Fixed a bug where InterpolatedImage and Box profiles were not correctly
@@ -116,4 +116,3 @@ Updates to config options
 - Added Spergel type. (#616)
 - Added lam, diam, scale_units options to Airy and OpticalPSF types. (#618)
 - Added TopHat type. (#639)
-

galsim/chromatic.py

@@ -212,7 +212,7 @@ def drawImage(self, bandpass, image=None, integrator='trapezoidal', **kwargs):
 
         The task of drawImage() in a chromatic context is to integrate a chromatic surface
         brightness profile multiplied by the throughput of `bandpass`, over the wavelength interval
-        indicated by `bandpass`.  
+        indicated by `bandpass`.
 
         Several integrators are available in galsim.integ to do this integration when using the
         first method (non-interpolated integration).  By default,
@@ -520,8 +520,8 @@ def shear(self, *args, **kwargs):
         the appropriate change in area, either use shear() with magnify(), or use lens(), which
         combines both operations.
 
-        Note that, while gravitational shear is monochromatic, the shear method may be used for 
-        many other use cases including some which may be wavelength-dependent, such as 
+        Note that, while gravitational shear is monochromatic, the shear method may be used for
+        many other use cases including some which may be wavelength-dependent, such as
         intrinsic galaxy shape, telescope dilation, atmospheric PSF shape, etc.  Thus, the
         shear argument is allowed to be a function of wavelength like other transformations.
 
@@ -566,7 +566,7 @@ def lens(self, g1, g2, mu):
 
         While gravitational lensing is achromatic, we do allow the parameters `g1`, `g2`, and `mu`
         to be callable functions to be parallel to all the other transformations of chromatic
-        objects.  In this case, the functions should take the wavelength in nanometers as the 
+        objects.  In this case, the functions should take the wavelength in nanometers as the
         argument, and the return values are the corresponding value at that wavelength.
 
         @param g1       First component of lensing (reduced) shear to apply to the object.
@@ -721,7 +721,7 @@ class InterpolatedChromaticObject(ChromaticObject):
     Any ChromaticObject can be used, although the interpolation procedure is most effective
     for non-separable objects, which can sometimes be very slow to render.
 
-    Normally, you would not create an InterpolatedChromaticObject directly.  It is the 
+    Normally, you would not create an InterpolatedChromaticObject directly.  It is the
     return type from `chrom_obj.interpolate()`.  See the description of that function
     for more details.
 
@@ -739,7 +739,7 @@ class InterpolatedChromaticObject(ChromaticObject):
                             whichever wavelength has the highest Nyquist frequency.
                             `oversample_fac`>1 results in higher accuracy but costlier
                             pre-computations (more memory and time). [default: 1]
-    """ 
+    """
     def __init__(self, obj, waves, oversample_fac=1.0):
 
         self.separable = obj.separable
@@ -926,7 +926,7 @@ def drawImage(self, bandpass, image=None, integrator='trapezoidal', **kwargs):
         images at the specified wavelengths.
 
         This integration will take place using interpolation between stored images that were
-        setup when the object was constructed.  (See interpolate() for more details.) 
+        setup when the object was constructed.  (See interpolate() for more details.)
 
         @param bandpass         A Bandpass object representing the filter against which to
                                 integrate.
@@ -1110,7 +1110,7 @@ def drawImage(self, bandpass, image=None, integrator='trapezoidal', **kwargs):
         @returns the drawn Image.
         """
         return self.build_obj().drawImage(bandpass, image, integrator, **kwargs)
- 
+
 
 class Chromatic(ChromaticObject):
     """Construct chromatic versions of galsim GSObjects.
@@ -1191,7 +1191,7 @@ class ChromaticTransformation(ChromaticObject):
     Typically, you do not need to construct a ChromaticTransformation object explicitly.
     This is the type returned by the various transformation methods of ChromaticObject such as
     shear(), rotate(), shift(), transform(), etc.  All the various transformations can be described
-    as a combination of transform() and shift(), which are described by (dudx,dudy,dvdx,dvdy) and 
+    as a combination of transform() and shift(), which are described by (dudx,dudy,dvdx,dvdy) and
     (dx,dy) respectively.
 
     @param obj              The object to be transformed.
@@ -1230,7 +1230,7 @@ def __init__(self, obj, jac=np.identity(2), offset=(0,0), flux_ratio=1., gsparam
             self._jac = jac
             self._offset = offset
             self._flux_ratio = flux_ratio
-            
+
         elif isinstance(obj, ChromaticTransformation):
             self.original = obj.original
             if hasattr(jac, '__call__'):
@@ -2014,7 +2014,7 @@ class ChromaticOpticalPSF(ChromaticObject):
                            [default: galsim.arcsec]
     @param   **kwargs      Any other keyword arguments to be passed to OpticalPSF, for example,
                            related to struts, obscuration, oversampling, etc.  See OpticalPSF
-                           docstring for a complete list of options. 
+                           docstring for a complete list of options.
     """
     def __init__(self, lam, diam=None, lam_over_diam=None, aberrations=None,
                            scale_unit=galsim.arcsec, **kwargs):

galsim/utilities.py

@@ -35,14 +35,14 @@ def roll2d(image, (iroll, jroll)):
 def kxky(array_shape=(256, 256)):
     """Return the tuple `(kx, ky)` corresponding to the DFT of a unit integer-sampled array of input
     shape.
-    
+
     Uses the SBProfile conventions for Fourier space, so `k` varies in approximate range (-pi, pi].
     Uses the most common DFT element ordering conventions (and those of FFTW), so that `(0, 0)`
     array element corresponds to `(kx, ky) = (0, 0)`.
 
     See also the docstring for np.fftfreq, which uses the same DFT convention, and is called here,
     but misses a factor of pi.
-    
+
     Adopts NumPy array index ordering so that the trailing axis corresponds to `kx`, rather than the
     leading axis as would be expected in IDL/Fortran.  See docstring for numpy.meshgrid which also
     uses this convention.
@@ -112,7 +112,7 @@ def parse_pos_args(args, kwargs, name1, name2, integer=False, others=[]):
         f(name1=x, name2=y)
 
     If the inputs must be integers, set `integer=True`.
-    If there are other args/kwargs to parse after these, then their names should be 
+    If there are other args/kwargs to parse after these, then their names should be
     be given as the parameter `others`, which are passed back in a tuple after the position.
     """
     def canindex(arg):
@@ -174,13 +174,13 @@ class SimpleGenerator:
     """
     def __init__(self, obj): self._obj = obj
     def __call__(self): return self._obj
-            
+
 class AttributeDict(object):
     """Dictionary class that allows for easy initialization and refs to key values via attributes.
 
     NOTE: Modified a little from Jim's bot.git AttributeDict class so that tab completion now works
     in ipython since attributes are actually added to __dict__.
-    
+
     HOWEVER this means the __dict__ attribute has been redefined to be a collections.defaultdict()
     so that Jim's previous default attribute behaviour is also replicated.
     """
@@ -258,7 +258,7 @@ def _convertPositions(pos, units, func):
                 np.array([pos.y], dtype='float') ]
 
     # Check for list of PositionD or PositionI:
-    # The only other options allow pos[0], so if this is invalid, an exception 
+    # The only other options allow pos[0], so if this is invalid, an exception
     # will be raised:
     elif isinstance(pos[0],galsim.PositionD) or isinstance(pos[0],galsim.PositionI):
         pos = [ np.array([p.x for p in pos], dtype='float'),
@@ -297,7 +297,7 @@ def _convertPositions(pos, units, func):
 
 def thin_tabulated_values(x, f, rel_err=1.e-4, preserve_range=False):
     """
-    Remove items from a set of tabulated f(x) values so that the error in the integral is still 
+    Remove items from a set of tabulated f(x) values so that the error in the integral is still
     accurate to a given relative accuracy.
 
     The input `x,f` values can be lists, NumPy arrays, or really anything that can be converted
@@ -308,7 +308,7 @@ def thin_tabulated_values(x, f, rel_err=1.e-4, preserve_range=False):
     @param rel_err          The maximum relative error to allow in the integral from the removal.
                             [default: 1.e-4]
     @param preserve_range   Should the original range of `x` be preserved? (True) Or should the ends
-                            be trimmed to include only the region where the integral is 
+                            be trimmed to include only the region where the integral is
                             significant? (False)  [default: False]
 
     @returns a tuple of lists `(x_new, y_new)` with the thinned tabulation.
@@ -417,4 +417,3 @@ def _gammafn(x):
                0.00000000000000122678, -0.00000000000000011813, 0.00000000000000000119,
                0.00000000000000000141, -0.00000000000000000023, 0.00000000000000000002
              )
-

galsim/wfirst/wfirst_psfs.py

@@ -202,8 +202,8 @@ def getPSF(SCAs=None, approximate_struts=False, n_waves=None, extra_aberrations=
                     pupil_plane_im=galsim.wfirst.pupil_plane_file,
                     oversampling=1.2, pad_factor=2.)
             if n_waves is not None:
-                PSF.setupInterpolation(waves=np.linspace(blue_limit, red_limit, n_waves),
-                                       oversample_fac=1.5)
+                PSF = PSF.interpolate(waves=np.linspace(blue_limit, red_limit, n_waves),
+                                      oversample_fac=1.5)
         else:
             tmp_aberrations = use_aberrations * zemax_wavelength / wavelength_nm
             if approximate_struts:
@@ -279,7 +279,8 @@ def storePSFImages(PSF_dict, filename, bandpass_list=None, clobber=False):
     SCA_index_list = []
     for SCA in PSF_dict.keys():
         PSF = PSF_dict[SCA]
-        if not isinstance(PSF, galsim.ChromaticOpticalPSF):
+        if not isinstance(PSF, galsim.ChromaticOpticalPSF) and \
+                not isinstance(PSF, galsim.InterpolatedChromaticObject):
             raise RuntimeError("Error, PSFs are not ChromaticOpticalPSFs.")
         star = galsim.Gaussian(sigma=1.e-8, flux=1.)
 

tests/test_chromatic.py

@@ -533,7 +533,7 @@ def test_chromatic_flux():
         err_msg="Drawn ChromaticConvolve flux (interpolated) doesn't match analytic prediction")
     # As an aside, check for appropriate tests of 'integrator' argument.
     try:
-        np.testing.assert_raises(TypeError, final_int.drawImage, bandpass, 
+        np.testing.assert_raises(TypeError, final_int.drawImage, bandpass,
                                  integrator='midp') # minor misspelling
         np.testing.assert_raises(TypeError, final_int.drawImage, bandpass,
                                  integrator=galsim.integ.midpt)
@@ -741,7 +741,7 @@ def test_ChromaticObject_expand():
     np.testing.assert_almost_equal(myy / (sigma/pixel_scale)**2, 1.0, decimal=4)
     np.testing.assert_almost_equal(mxy / (sigma/pixel_scale)**2, 0, decimal=4)
 
-    # First a very simple case with no actual wavelength dependence, but using the 
+    # First a very simple case with no actual wavelength dependence, but using the
     # functional syntax.
     gal1 = galsim.ChromaticObject(gal).expand(lambda w: 1.2)
     # Use a simple bandpass so we can do the integral below analytically
@@ -863,7 +863,7 @@ def test_ChromaticObject_rotate():
     np.testing.assert_almost_equal(myy / (sigma/pixel_scale)**2, (1-0.3)/fact, decimal=4)
     np.testing.assert_almost_equal(mxy / (sigma/pixel_scale)**2, 0, decimal=4)
 
-    # First a very simple case with no actual wavelength dependence, but using the 
+    # First a very simple case with no actual wavelength dependence, but using the
     # functional syntax.
     gal1 = galsim.ChromaticObject(gal).rotate(lambda w: 0.4 * galsim.radians)
     bp = galsim.Bandpass(lambda w: 1. - 0.12*(w-600)**2/100**2, 500, 700)
@@ -963,7 +963,7 @@ def test_ChromaticObject_shear():
     np.testing.assert_almost_equal(myy / (sigma/pixel_scale)**2, 1.0, decimal=4)
     np.testing.assert_almost_equal(mxy / (sigma/pixel_scale)**2, 0.0, decimal=4)
 
-    # First a very simple case with no actual wavelength dependence, but using the 
+    # First a very simple case with no actual wavelength dependence, but using the
     # functional syntax.
     gal1 = galsim.ChromaticObject(gal).shear(lambda w: galsim.Shear(e1=0.23, e2=0.13))
     bp = galsim.Bandpass(lambda w: 1. - 0.12*(w-600)**2/100**2, 500, 700)
@@ -1519,7 +1519,7 @@ def test_ChromaticOpticalPSF():
     t1 = time.time()
 
     # For ChromaticOpticalPSF, exact evaluation is too slow for routine unit tests.  So, for
-    # this unit test, we use an interpolated version only.  The tests of 
+    # this unit test, we use an interpolated version only.  The tests of
     # ChromaticObject in the previous unit test should be enough to ensure that exact
     # and interpolated evaluation match in general (given reasonable settings).
 
@@ -1547,7 +1547,7 @@ def test_ChromaticOpticalPSF():
     # obscuration = 0.18
     # nstruts = 2
     # scale = 0.02
-    # 
+    #
     # psf = galsim.ChromaticOpticalPSF(lam=lam, diam=diam, aberrations=aberrations,
     #                                  obscuration=obscuration, nstruts=nstruts)
     # obj = galsim.Convolve(psf, star)
@@ -1643,7 +1643,7 @@ def test_ChromaticAiry():
     # diam = 3.1 # meters
     # obscuration = 0.11
     # scale = 0.02
-    # 
+    #
     # psf = galsim.ChromaticAiry(lam=lam, diam=diam, obscuration=obscuration)
     # obj = galsim.Convolve(psf, star)
     # im_r = obj.drawImage(bandpass, scale=scale)