DM-9615: Add dcrAssembleCoadd task #173

isullivan · 2017-12-21T22:26:09Z

This pull request is being created early to allow comments on the changes. It is not expected to be merged for a while.

yalsayyad · 2017-12-22T17:29:12Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+__all__ = ["DcrAssembleCoaddTask"]
+
+
+class DcrAssembleCoaddConfig(CompareWarpAssembleCoaddConfig):


What extra methods/configs do you get from inheriting from CompareWarp vs vanilla Assemble? Oh I see, the templateCoadd.

I'm imagining you do want asteroids/CRs/ghosts clipped... but this skeleton overrides all the artifact-finding functionality of CompareWarp?

yalsayyad · 2017-12-22T17:30:08Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        doc="Maximum number of iterations of forward modeling.",
+        default=0.001,
+    )
+    assembleStaticSkyModel = pexConfig.ConfigurableField(


Keep something like this if you inherit from Assemble, but you don't need it if you inherit from CompareWarp

yalsayyad · 2017-12-22T17:31:25Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            " naive/first-iteration model of the static sky.",
+    )
+
+    def setDefaults(self):


I assume this is here because you're expecting to set some Assemble defaults by the time its time to merge.

yalsayyad · 2017-12-22T17:37:14Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+
+class DcrAssembleCoaddConfig(CompareWarpAssembleCoaddConfig):
+    dcrBufferSize = pexConfig.Field(


Probably don't need the word dcr in these configs.

yalsayyad · 2017-12-22T17:40:40Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        CompareWarpAssembleCoaddTask.__init__(self, *args, **kwargs)
+
+    @pipeBase.timeMethod
+    def run(self, dataRef, selectDataList=[]):


I'm optimistic that we can change AssembleCoaddTask.run so that you don't need to override the entirety of run here.

yalsayyad · 2017-12-22T17:42:44Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                self.log.info("Persisting dcrCoadd")
+                dataRef.put(retStruct.coaddExposure, "dcrCoadd", subfilter=retStruct.subFilter)
+
+            if coaddExp is None:


Oh, does this mean that this task will generate the the regular coadds too?

yalsayyad · 2017-12-22T17:43:18Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            else:
+                mimage = coaddExp.getMaskedImage()
+                mimage += retStruct.coaddExposure.getMaskedImage()
+        if self.config.doWrite:


Indent more?

yalsayyad · 2017-12-22T17:49:37Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+            if self.config.doWrite:
+                self.log.info("Persisting dcrCoadd")
+                dataRef.put(retStruct.coaddExposure, "dcrCoadd", subfilter=retStruct.subFilter)


We could make a AssembleCoaddTask.write that you can override.

yalsayyad · 2017-12-22T17:54:37Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            supplementaryData=supplementaryData,
+        )
+        coaddExp = None
+        for retStruct in retStructGen:


I think our ability to make a generic run() for both dcrAssemble and the other classes depends on how others feel about always iterating over the outputs of assemble. In most cases there will be one. It's fine with me. @r-owen? @TallJimbo?

PaulPrice

I'm not done, but I'm about to go to bed. Will pick up again later.

PaulPrice · 2018-03-14T01:40:49Z

bin.src/dcrAssembleCoadd.py

@@ -0,0 +1,26 @@
+#!/Users/sullivan/LSST/code/lsstsw/miniconda/bin/python


Fix the path.

PaulPrice · 2018-03-14T01:42:41Z

bin.src/dcrAssembleCoadd.py

+
+#
+# LSST Data Management System
+# Copyright 2008, 2009, 2010 LSST Corporation.


I think the new standard is to put the copyleft in a COPYRIGHT file and an AUTHORS file.
The copyright holder should be your institution, and the year should be updated.

PaulPrice · 2018-03-14T01:46:16Z

python/lsst/pipe/tasks/assembleCoadd.py

@@ -471,6 +471,30 @@ def prepareInputs(self, refList):
        return pipeBase.Struct(tempExpRefList=tempExpRefList, weightList=weightList,
                               imageScalerList=imageScalerList)

+    def prepareStats(self, mask=None):
+        """!
+        \brief Prepare the statistics for coadding images.


You need to change all this to use numpydoc before merging.

This hasn't been converted to numpydoc yet?

PaulPrice · 2018-03-14T01:51:36Z

python/lsst/pipe/tasks/assembleCoadd.py

+            bit = afwImage.Mask.getMaskPlane(plane)
+            statsCtrl.setMaskPropagationThreshold(bit, threshold)
+        statsFlags = afwMath.stringToStatisticsProperty(self.config.statistic)
+        return (statsCtrl, statsFlags)


Prefer a Struct to a tuple.

PaulPrice · 2018-03-14T01:55:00Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+#
+# LSST Data Management System
+# Copyright 2008-2016 AURA/LSST.


Follow modern instructions for copyleft.

PaulPrice · 2018-03-14T02:43:32Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        """
+        retStruct = AssembleCoaddTask.run(self, dataRef, selectDataList=selectDataList)
+        for subfilter, coaddExposure in enumerate(retStruct.dcrCoadds):
+            if self.config.doWrite:


Put this before the for loop.

PaulPrice · 2018-03-14T02:43:52Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        for subfilter, coaddExposure in enumerate(retStruct.dcrCoadds):
+            if self.config.doWrite:
+                self.writeDcrCoadd(dataRef, coaddExposure, subfilter)
+        return pipeBase.Struct(coaddExposure=retStruct.coaddExposure, nImage=retStruct.nImage)


You should return the DCR coadds too.

PaulPrice · 2018-03-14T02:52:23Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            self.setBrightObjectMasks(coaddExposure, dataRef.dataId, brightObjectMasks)
+
+        self.log.info("Persisting dcrCoadd")
+        dataRef.put(coaddExposure, "dcrCoadd", subfilter=subfilter)


I think there's a problem with the data model: if you change the configuration of the reader to use a different number of sub-bands (why are they called "sub-bands" elsewhere and "subfilters" here?), then you would get the wrong data and wouldn't be aware of it. I think you either want to write these as data cubes (which I don't think we can do in afw) or include the total number of sub-bands in the data model (maybe the template should be ...%(subfilter)dof%(numSubfilters)d...).

PaulPrice · 2018-03-14T02:54:42Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        if self.config.doMaskBrightObjects:
+            brightObjectMasks = self.readBrightObjectMasks(dataRef)
+            self.setBrightObjectMasks(coaddExposure, dataRef.dataId, brightObjectMasks)


The above operations (both interpolation and masking bright objects) don't belong in a method that writes data. They should be done before calling writeDcrCoadd so the results can be returned from the run method. It would be best if you didn't copy the implementations but just called a method to do them.

PaulPrice · 2018-03-14T02:56:24Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        @param[in] dataRef: Data reference defining the patch for coaddition and the reference Warp
+                        - [out] "dcrCoadd"
+        @param[in] coaddExposure: pipeBase.Struct with coaddExposure


Are you sure this is a Struct? I think it's actually an Exposure based on how you're treating it below.

PaulPrice

Still more to do...

PaulPrice · 2018-03-14T16:46:01Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        Find artifacts and apply them to the warps' masks creating a list of alternative masks with a
+        new "CLIPPED" plane and updated "NO_DATA" plane.
+        Then pass these alternative masks to the base class's assemble method.


You've gotta update this description for DCR.

PaulPrice · 2018-03-14T16:47:15Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        statsCtrl, statsFlags = self.prepareStats(mask=badPixelMask)
+
+        subregionSizeArr = self.config.subregionSize
+        subregionSize = afwGeom.Extent2I(subregionSizeArr[0], subregionSizeArr[1])


subregionSize = afwGeom.Extent2I(*self.config.subregionSize)

PaulPrice · 2018-03-14T16:48:28Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            self.log.info("Initial convergence : %s", convergenceMetric)
+            convergenceList = [convergenceMetric]
+            convergenceCheck = 1.
+            while (convergenceCheck > self.config.convergenceThreshold) or (modelIter < self.config.minNIter):


Don't need parens here.

PaulPrice · 2018-03-14T16:51:26Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            self.log.info("Initial convergence : %s", convergenceMetric)
+            convergenceList = [convergenceMetric]
+            convergenceCheck = 1.
+            while (convergenceCheck > self.config.convergenceThreshold) or (modelIter < self.config.minNIter):


You're not respecting the useConvergence configuration parameter. In fact, I don't see that used anywhere.

PaulPrice · 2018-03-14T16:54:02Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                    convergenceCheck = (convergenceList[-1] - convergenceMetric)/convergenceMetric
+                    convergenceList.append(convergenceMetric)
+                except Exception as e:
+                    self.log.warn("Error during iteration %s while computing coadd %s: %s", subBBox, e)


More % outputs than there are inputs, so this will fail.

What's supposed to happen after this error? It looks like you simply break out of the convergence loop and proceed on your merry way. Surely that's a mistake!

PaulPrice · 2018-03-14T17:47:43Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                        mask &= ~mask.getPlaneBitMask(maskPlane)
+                    except Exception as e:
+                        self.log.warn("Unable to remove mask plane %s: %s", maskPlane, e.message)
+            maskedImage -= templateImage


Much of the above looks like boilerplate copied from somewhere else. Can you factor it out?

PaulPrice · 2018-03-14T17:48:23Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                    try:
+                        mask &= ~mask.getPlaneBitMask(maskPlane)
+                    except Exception as e:
+                        self.log.warn("Unable to remove mask plane %s: %s", maskPlane, e.message)


I don't think you should do e.message. Simply casting e to str as the %s does should be sufficient.

PaulPrice · 2018-03-14T17:48:52Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                    except Exception as e:
+                        self.log.warn("Unable to remove mask plane %s: %s", maskPlane, e.message)
+            maskedImage -= templateImage
+            obsWeight = self.calculateWeight(maskedImage, convergeMask)*1e3


Why the magic factor of 1000? If it's useful, shouldn't it be part of calculateWeight?

PaulPrice · 2018-03-14T18:34:29Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                residual.setXY0(bboxGrow.getBegin())
+                newModel = self.clampModel(residual, oldModel, bboxGrow,
+                                           useNonNegative=self.config.useNonNegative,
+                                           clamp=self.config.clampModel)


Why do you need to pass config parameters that are already readily accessible to the method?

PaulPrice · 2018-03-14T18:34:43Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                dcrSubModelOut.append(newModel)
+        self.regularizeModel(dcrSubModelOut, bboxGrow, baseMask,
+                             nSigma=self.config.regularizeSigma,
+                             clamp=self.config.clampFrequency)


Ditto on passing config parameters.

PaulPrice

Got through the main bulk of the code. Just the test left, I think.

PaulPrice · 2018-03-19T18:21:34Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        """! Calculate the weight of an exposure based on the goodness of fit of the matched template.
+
+        @param[in] residual: residual masked image after subtracting a DCR-matched template.
+        @param[in] goodMask: Mask plane to evaluate the goodness of fit of the residual over.


"Mask plane over which to evaluate..."

Also, it looks like it isn't a mask plane, but a bitmask.

PaulPrice · 2018-03-19T18:22:02Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        @return weight: Goodness of fit metric of the residual image.
+        """
+        residualVals = residual.getImage().getArray()
+        finiteInds = (numpy.isfinite(residualVals))


Outer parens not needed.

PaulPrice · 2018-03-19T18:22:45Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        @return weight: Goodness of fit metric of the residual image.
+        """
+        residualVals = residual.getImage().getArray()


residual.image.array

PaulPrice · 2018-03-19T18:24:51Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        """
+        residualVals = residual.getImage().getArray()
+        finiteInds = (numpy.isfinite(residualVals))
+        goodMaskInds = (residual.getMask().getArray() & goodMask) == goodMask


Why do you want == goodMask? Essentially it's saying that the pixels can only have DETECTED set, and nothing else. Why do you want to prohibit anything else? If you really do want to prohibit other mask planes from being set, why not specify a badMask value?

PaulPrice · 2018-03-19T18:29:36Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        residualVals = residual.getImage().getArray()
+        finiteInds = (numpy.isfinite(residualVals))
+        goodMaskInds = (residual.getMask().getArray() & goodMask) == goodMask
+        weight = 1./numpy.std(residualVals[finiteInds*goodMaskInds])


I would write finiteInds & goodMaskInds since they're booleans, but that works.

PaulPrice · 2018-03-19T20:12:07Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        """
+        parAngle = visitInfo.getBoresightParAngle().asRadians()
+        cd = wcs.getCdMatrix()
+        cdAngle = (numpy.arctan2(-cd[0, 1], cd[0, 0]) + numpy.arctan2(cd[1, 0], cd[1, 1]))/2.


Can you still do this with SkyWcs? If so, that's a bug. We should be hiding the CD matrix from the user. Not sure of the Proper Way to do this.

PaulPrice · 2018-03-19T20:12:22Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+        rotAngle = afwGeom.Angle(cdAngle + parAngle)
+        return rotAngle
+
+    @staticmethod


Drop staticmethod.

PaulPrice · 2018-03-19T20:13:31Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+                bbox.grow(afwGeom.Extent2I(-bufferXSize, -bufferYSize))
+                bbox.shift(afwGeom.Extent2I(dx0 + x0, dy0 + y0))
+                retMask[bbox, afwImage.PARENT] |= mask[bboxBase, afwImage.PARENT]
+        return retMask


Why aren't you implementing this with our own warping functions from afw?

PaulPrice · 2018-03-19T20:13:44Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+
+def wavelengthGenerator(lambdaEff, filterWidth, dcrNSubbands):
+    """! tIterate over the wavelength endpoints of subfilters.


Typo: "tIterate"

PaulPrice · 2018-03-19T20:14:35Z

tests/test_dcrAssembleCoadd.py

+from __future__ import absolute_import, division, print_function
+#
+# LSST Data Management System
+# Copyright 2008-2017 AURA/LSST.


Update copyleft to modern standards.

PaulPrice

Hooray, finally made it all the way through!

PaulPrice · 2018-03-19T23:55:53Z

tests/test_dcrAssembleCoadd.py

+# see <https://www.lsstcorp.org/LegalNotices/>.
+
+from collections import namedtuple
+from astropy import units as u


I really hate having something a module named simply u, but I understand it's standard practice...

PaulPrice · 2018-03-19T23:57:35Z

tests/test_dcrAssembleCoadd.py

+
+
+class DcrAssembleCoaddTestTask(lsst.utils.tests.TestCase):
+    """! A test case for the DCR-aware image coaddition algorithm.


Please use numpydoc (no leading !).

PaulPrice · 2018-03-19T23:58:57Z

tests/test_dcrAssembleCoadd.py

+    """! A test case for the DCR-aware image coaddition algorithm.
+    """
+
+    def setUp(self):


There's a lot of code in here. Can you summarise in the method docstring what you're going to set up?

PaulPrice · 2018-03-20T00:02:20Z

tests/test_dcrAssembleCoadd.py

+        psfSize = 2
+        nSrc = 5
+        seed = 5
+        self.randGen = np.random


Why would you put the module into self? It's not to save on typing. If you want something in self, you could do self.randGen = np.random.RandomState(seed). I would rename randGen as rng (random number generator).

PaulPrice · 2018-03-20T00:04:10Z

tests/test_dcrAssembleCoadd.py

+        noiseLevel = 5
+        sourceSigma = 20.
+        fluxRange = 2.
+        edgeDist = self.config.bufferSize


Why edgeDist instead of bufferSize?

PaulPrice · 2018-03-20T00:44:43Z

tests/test_dcrAssembleCoadd.py

+                                                                    useInverse=False)
+        shift = (dcrShift[0].dy, dcrShift[0].dx)
+        refImage = scipy.ndimage.interpolation.shift(self.dcrModels[0].getImage().getArray(), shift)
+        refVariance = scipy.ndimage.interpolation.shift(self.dcrModels[0].getVariance().getArray(), shift)


All you're doing is rewriting the code in DcrAssembleCoaddTask.convolveDcrModelPlane --- that's not a real test. You should do a proper test: maybe centroid the sources on the model images to verify the shifts?

PaulPrice · 2018-03-20T00:52:20Z

tests/test_dcrAssembleCoadd.py

+        bboxClip.grow(afwGeom.Extent2I(-bufferXSize, -bufferYSize))
+
+        # Shifting the mask grows each mask plane by one pixel in the direction of the shift,
+        #  so a shift followed by the reverse shift should be the same as a dilation by one pixel.


Only if the shift is small. You should do a large (> 1 pixel) shift as well to ensure that your code is capable of doing more than just growing a mask, but also actually shifting.

PaulPrice · 2018-03-20T00:54:46Z

tests/test_dcrAssembleCoadd.py

+        # Shifting the mask grows each mask plane by one pixel in the direction of the shift,
+        #  so a shift followed by the reverse shift should be the same as a dilation by one pixel.
+        convolutionStruct = np.array([[True, True, True], [True, True, True], [True, True, True]])
+        maskRefCheck = dilate(model[bboxClip].getMask().getArray() == detectMask,


Again, why not use our own code (in this instance, lsst.afw.geom.SpanSet) instead of a third party's?

PaulPrice · 2018-03-20T00:56:17Z

tests/test_dcrAssembleCoadd.py

+        self.assertFloatsEqual(newMaskCheck, maskRefCheck)
+
+    def testRegularizationLargeClamp(self):
+        """! Frequency regularization should leave the models unchanged if all the clamp factor is large.


"all the clamp factor"

PaulPrice · 2018-03-20T00:58:36Z

tests/test_dcrAssembleCoadd.py

+            imageDiffHigh = model.getImage().getArray() - (templateImage*clamp + noiseLevel)
+            self.assertLessEqual(np.max(imageDiffHigh), 0.)
+            imageDiffLow = model.getImage().getArray() - (templateImage/clamp - noiseLevel)
+            self.assertGreaterEqual(np.max(imageDiffLow), 0.)


This looks like a reimplementation of the code --- that's not a real test.

PaulPrice

Last major concern is I/O.

PaulPrice · 2018-07-19T17:57:30Z

python/lsst/pipe/tasks/assembleCoadd.py

+        ----------
+        coaddExposure : `lsst.afw.image.Exposure`
+            The coadded exposure to process.
+        dataRef : `lsst.daf.persistence.butlerSubset.ButlerDataRef`


Please don't include the unnecessary butlerSubset.

PaulPrice · 2018-07-19T18:00:31Z

python/lsst/pipe/tasks/assembleCoadd.py

+        Parameters
+        ----------
+        mask : `int`, optional
+            Mask planes to exclude from coaddition.


"Mask planes" would have the type "list of str", but this is an int. Since you're plugging this into setAndMask, I think you mean the bit mask value.

PaulPrice · 2018-07-19T18:02:15Z

python/lsst/pipe/tasks/assembleCoadd.py

+            Statistics structure with the following fields:
+
+            - ``statsCtrl``: Statistics control object for coadd
+            - ``statsFlags``: afwMath.Property object for statistic for coadd


I really would like to see this follow the example in the developer guide, where the type is included in parens and backticks. And the type should be fully-qualified (e.g., lsst.afw.math.Property).

PaulPrice · 2018-07-19T18:03:27Z

python/lsst/pipe/tasks/assembleCoadd.py

+
+        Parameters
+        ----------
+        maskedImage : `lsst.afw.image.maskedImage`


Typo: type is lsst.afw.image.MaskedImage.

PaulPrice · 2018-07-19T18:06:05Z

python/lsst/pipe/tasks/assembleCoadd.py

+        ------
+        subBBox : `lsst.afw.geom.Box2I`
+            Next sub-bounding box of size subregionSize or smaller; each subBBox
+            is contained within bbox, so it may be smaller than subregionSize at


Put parameter names in double-backticks, e.g.,

is contained within ``bbox``, so it may be smaller than ``subregionSize`` at

PaulPrice · 2018-07-19T19:27:47Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        Returns
+        -------
+        `float`


Needs a name.

PaulPrice · 2018-07-19T19:27:57Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        Returns
+        -------
+        `lsst.afw.image.ExposureF`


Needs a name.

PaulPrice · 2018-07-19T19:28:07Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+
+        Returns
+        -------
+        `list` of `lsst.afw.image.ExposureF`


Needs a name.

PaulPrice · 2018-07-19T19:28:48Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            Best fit model of the true sky after correcting chromatic effects.
+        skyInfo : `lsst.pipe.base.Struct`
+            Patch geometry information, from getSkyInfo
+        tempExpRefList : `list` of `ButlerDataRef`


Use fully-qualified namespaces, lsst.daf.persistence.ButlerDataRef.

PaulPrice · 2018-07-19T19:29:22Z

python/lsst/pipe/tasks/dcrAssembleCoadd.py

+            Best fit model of the true sky after correcting chromatic effects.
+        bbox : `lsst.afw.geom.box.Box2I`
+            Sub-region to coadd
+        tempExpRefList : `list` of `ButlerDataRef`


lsst.daf.persistence.ButlerDataRef.

PaulPrice · 2018-07-19T19:51:18Z

@isullivan points out that saving the components of the DcrModel in separate FITS files makes it easy to do multiband measurements (by setting coaddName=dcr and using --id subfilter=0..2). I'm not perfectly happy about this, but it seems like a reasonable solution for the time being. In that case, the existing solution for I/O is reasonable, so I withdraw my objections.

RobertLuptonTheGood · 2018-07-19T20:29:44Z

In [1]: import lsst.afw.image as afwImagepythonstartup/rhl is not in your load path In [1]: import lsst.afw.image as afwImage In [2]: mi = afwImage.MaskedImageF In [3]: mi = afwImage.MaskedImageF(10, 20); mi.image[:] = 10; mi.variance[:] = 100 In [4]: mi /= 10 In [5]: print(mi.image[0,0]) 1.0 In [6]: print(mi.variance[0,0]) 1.0

yalsayyad reviewed Dec 22, 2017

View reviewed changes

isullivan force-pushed the tickets/DM-9615 branch from cc84fe6 to d4e215e Compare January 10, 2018 22:42

isullivan force-pushed the tickets/DM-9615 branch from ab30c4b to dc965d4 Compare January 17, 2018 23:45

isullivan force-pushed the tickets/DM-9615 branch from 8adf54d to c4786a2 Compare January 30, 2018 03:00

isullivan force-pushed the tickets/DM-9615 branch 3 times, most recently from 3cb5500 to fff3f55 Compare February 16, 2018 21:40

isullivan force-pushed the tickets/DM-9615 branch 3 times, most recently from bf7b7e6 to 31ade9b Compare March 5, 2018 05:37

isullivan force-pushed the tickets/DM-9615 branch 2 times, most recently from 4b91a17 to b2438d4 Compare March 9, 2018 00:02

PaulPrice reviewed Mar 14, 2018

View reviewed changes

PaulPrice reviewed Mar 19, 2018

View reviewed changes

PaulPrice requested changes Mar 20, 2018

View reviewed changes

isullivan force-pushed the tickets/DM-9615 branch from cc4ba21 to 070aa27 Compare March 26, 2018 21:47

isullivan force-pushed the tickets/DM-9615 branch 3 times, most recently from 5cf9cb2 to 6336f2d Compare April 6, 2018 20:59

isullivan force-pushed the tickets/DM-9615 branch 2 times, most recently from 71c25b6 to 911ad6f Compare April 20, 2018 21:07

isullivan added 10 commits July 9, 2018 14:19

Add unit test of regularization between iterations.

68054d3

No need to use sub-images when calculating dcrNImages

8d17ad8

Minimum working version with the new DcrModel class

e400934

Fix unit tests to work with new DcrModel class

8136581

Rename DCR unit test

f5d2c8a

Move DcrModel class to its own module.

90dc764

Remove access to DcrModel internal parameters

df7b651

Remove whitespace

4c70764

Clean up the docs.

1c64350

Make DcrModel iterable.

962d7bf

isullivan force-pushed the tickets/DM-9615 branch 2 times, most recently from 2c22596 to e40ff4a Compare July 11, 2018 02:22

isullivan added 5 commits July 12, 2018 15:03

Clean up DCR docstrings.

76a9dde

Remove DcrModel.getImage() in favor of __getitem__

d3a0f51

afwImage.PARENT is now the default for bounding boxes.

c32d7ab

Add DcrModel helper methods

08e67b7

New DcrModel construction interface

1728bd8

isullivan force-pushed the tickets/DM-9615 branch from e40ff4a to 080c20e Compare July 12, 2018 22:04

Add method to load an existing DcrModel from a repository

2584eaa

isullivan force-pushed the tickets/DM-9615 branch from ea65e4c to 5dbe3f1 Compare July 18, 2018 18:44

PaulPrice reviewed Jul 19, 2018

View reviewed changes

PaulPrice approved these changes Jul 19, 2018

View reviewed changes

isullivan added 3 commits July 19, 2018 14:30

Rename calculateRotationAngle

53bf95b

Fix variance in conditionDcrModel

677be0f

Clean up DcrModel documentation

198144e

isullivan force-pushed the tickets/DM-9615 branch from fefdbc0 to 198144e Compare July 19, 2018 23:31

isullivan merged commit 81851de into master Jul 20, 2018

ktlim deleted the tickets/DM-9615 branch August 25, 2018 06:46

		__all__ = ["DcrAssembleCoaddTask"]


		class DcrAssembleCoaddConfig(CompareWarpAssembleCoaddConfig):



		class DcrAssembleCoaddConfig(CompareWarpAssembleCoaddConfig):
		dcrBufferSize = pexConfig.Field(

		@@ -0,0 +1,26 @@
		#!/Users/sullivan/LSST/code/lsstsw/miniconda/bin/python



		def wavelengthGenerator(lambdaEff, filterWidth, dcrNSubbands):
		"""! tIterate over the wavelength endpoints of subfilters.



		class DcrAssembleCoaddTestTask(lsst.utils.tests.TestCase):
		"""! A test case for the DCR-aware image coaddition algorithm.

DM-9615: Add dcrAssembleCoadd task #173

DM-9615: Add dcrAssembleCoadd task #173

Conversation

isullivan commented Dec 21, 2017

yalsayyad Dec 22, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PaulPrice left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PaulPrice left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PaulPrice left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PaulPrice left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PaulPrice left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PaulPrice commented Jul 19, 2018

RobertLuptonTheGood commented Jul 19, 2018 via email

yalsayyad Dec 22, 2017 •

edited