https://pfspipe.ipmu.jp/jira This file is an XML representation of an issue en-us 8.3.4 803005 13-09-2019 https://pfspipe.ipmu.jp/jira/browse/PIPE2D-771 DRP 2-D Pipeline <p>The 2021-03-07 weekly detected a regression in wavelength calibration:</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java"> ====================================================================== FAIL: testResiduals (__main__.ArcTestCase_brn_39) (arm=<span class="code-quote">'b'</span>, fiberId=2) Test that wavelength fit residuals are reasonable ---------------------------------------------------------------------- Traceback (most recent call last): File <span class="code-quote">"/scratch/pprice/jenkins/weekly/2021-03-07/build/stack/miniconda3-4.5.12-1172c30/Linux64/pfs_pipe2d/w.2021.10/python/pfs/pipe2d/weekly/test_weekly.py"</span>, line 127, in testResiduals self.assertFloatsAlmostEqual(0.741*(uq - lq), 0.0, atol=2.0e-2) File <span class="code-quote">"/scratch/pprice/jenkins/weekly/2021-03-07/build/stack/miniconda3-4.5.12-1172c30/Linux64/utils/18.1.0/python/lsst/utils/tests.py"</span>, line 735, in assertFloatsAlmostEqual testCase.assertFalse(failed, msg=<span class="code-quote">"\n"</span>.join(errMsg)) AssertionError: True is not <span class="code-keyword">false</span> : 0.021484697109974946 != 0.0; diff=0.021484697109974946/0.021484697109974946=1.0 with rtol=2.220446049250313e-16, atol=0.02 </pre> </div></div> PIPE2D-771

Fix broken weekly

Story Normal Done Done price price Wed, 10 Mar 2021 19:06:49 +0000 Thu, 18 Mar 2021 19:37:22 +0000 Tue, 16 Mar 2021 21:15:36 +0000 0 3 <p>Reverting <tt>a4d7c14</tt> ("Don't casual mess with slit offsets") in obs&#95;pfs solved the problem. I suspect that the problem is due to the known 0.5 pixel offset in the detectorMap from the simulator. I think the thing to do is to introduce a bootstrap in the weekly to fix the 0.5 pixel offset.</p> <p>Fixing the simulator has to be a high priority. slit offsets should be considered pretty-much sacrosanct. Running an early bootstrap to fix this would be OK too, providing the weekly itself doesn't mess with them.</p> <p>I kind of like having the detectorMap from the simulator being off by a little: it mirrors reality! I therefore favour the bootstrap solution, which has the added advantage of exercising the bootstrap code to prevent bitrot.</p> <p>bootstrap alone wasn't sufficient to solve the problem. I tracked the problem to ambiguous line matches, which apparently happened sufficiently frequently that it inflated even the IQR-based RMS measurements that were failing the above test. Introducing an exclusion radius of 0.1 nm (about a pixel) to the reference lines solves this, and the result is a better fit than we even had before the weekly broke (softening 0.037 pixel --&gt; 0.022; IQR-based RMS wavelength residual per fiber ~ 0.01 nm --&gt; 0.008; fit lines/total lines 108656/142698 --&gt; 75229/97444; this is for the <tt>b</tt> arm).</p> <p>Updates reviewed. No additional comments raised. Weekly test run using ticket branch was successful.</p> <p>Merged to master.</p> Development Rank 0|02qpt6:508yl0chzzy0000039 Reviewers hassan Sprint 2DDRP-2021 A3