Bug Hunting
I spent the day hunting for the bug that is clobbering pixmap. More
Maybe Found
I think I know the cause of the bug, an error in the pixmap dimensions. But I still need to come up with a fix for the problem. More
Frustration
I edited the map_lookup function to use the standard Python array macros to access the pixmap. And I am passing the pixmap around as a Python object, because it contains its shape information and I don't need to pass that separately. The problem is that I can't seem to create a Python array in my ... More
Rewrite
I rewrote my unit testing code to get the pixmap from python and am debugging the unit tests. More
Almost There
Fixed a problem with the x-y orderingof coordinates in the code. All the unit tests pass, except one, which will require a deeper analysis. More
Next Problem
I resolved the last failing unit test. It was caused by not initializing p->no_over. No of the regression tests take an error exit, but they still do not match the old values. In particular, in three cases (one acs, two wfpc2) I get no overlap warnings from the blotting step, which as far ... More
Overlap Fixed
The no overlap problem in doblot turned out to be a mixup in the x-y coordinate order plus a typo in calc_pixmap.py. So now there are no obvious errors in the output, but the output still does not match the results of the previous version of the code and I am going to have to figure out why. More
A New Plan
My new plan is to compare the results from the current version of drizzlepac to my new version. I spent the day modifying the code and my unit tests so they would work together. Still have got a successful link. More
Backtracking
I spent the morning solving a link problem with the old version of drizzlepac linking with the unit tests. I spent some time trying to work with the unit tests in old drizzlepac. I came to the conclusion that the python callback object made writing true unit tests impossible. What I will have to ... More
A New Hope
I spent the day writing a test case that runs do_driz in the old version of drizzlepac. I am now trying to get it to run to completion -- it throws a segmentation error. More
Sturm und Drang
I got the old drizzlepac test code to run to completion and produce an output file. Naturally it did not match the result I got from the new drizzlepac code. Tomorrow I will try to figure out why. More
Test and Fail
I got the unit tests for do_driz and do_blot to run. They are close to, but do not match the old results. It looks like there is a subpixel difference in the coordinate mapping. I have edited the old C source so it produces a pixmap and I can compare the two. More
Half Fixed
I figured out last night that the problems with drizzlepac were caused by an off by one problem. I fixed the magnitude of the result by shifting the input pixel map, but the result is also shifted. I need to find where there needs to be a compensatory shift in the output.I also fixed a bug in the ... More
More Fixed
I found the other half of the bug I've been hunting for. All the unit tests pass. I ran the regression tests. Some things seem to work ok, but there are some issues I will need to look at more closely. More
Approximately Right
All the results from the regression tests approximately match the previous results. However, it is an open question whether the match is good enough. Warren suggests using tweakreg as a test and I am waiting on instructions on how to do this. I did another merge from subversion to get the latest ... More
Not Quite Yet
I wrote the first version of the code to replace linear interpolation with polynomial interpolation in map_value. I did a comparison of the old regression tests to the results from the new code. The results are not quite good enough, so I will have to do more testing. More
More Tests
I backed out the polynomial interpolation code and wrote some new tests where the input image is a grid. I've had some successful results, but still need to get some tests to run and write other tests. More
Still Testing
I finished some tests, ran them, and reported the results. I went back to coding the polynomial interpolator. More
Centroid
I worked on adding a different MOE to the drizzlepac code. It calculates the centroids of objects on the output image and compares them to the reference image. I'm still debugging the change. More
Good. but good enough?
I finished the centroid changes and reported the results. The big surprise was how the flux changes from kernel to kernel. More
More Debugging
I added the code to drizzlepac to a new github repository. I added some debugging prints to see if I can figure out the reason for the difference in the flux in the new and old versions of drizzlepac. More
Chasing Errors
I spent the day trying to determine the reason for the difference between the old and new code. I added some debugging prints and compared the intermediate results. I blamed the problem on ill conditioning in the formula for the total pixel area, but that did not turn out to be the case. More
Ghosts
I cleaned up the drizzlepac code from the recent testing and checked the changes into github. I added debugging prints to the Lanczos kernel and looked for the discrepancy between the old and new code. There appears to be an off by one error. Possibly the same error is in the gaussian kernel, as ... More