One Out of Three
I found and fixed a bug in the Lanczos kernel. I am still trying to determine the cause of a difference between the results for the new and old code with the other kernels. It might be a more subtle off by one error. I still haven't figured the cause of the error in the stsdas help script. More
Patience Required
I worked on the bug with stsdas help, with no success. I then returned to trying to figure out the difference between the old and new drizzlepac coordinate mapping. It turns out the old drizzlepac was using its internal mapping code, so I forced it to use the wcs mapping code instead.But that did ... More
Snake Eyes
I did some more testing and code reading in drizzlepac in search of the discrepancy between the old an new version. I learned some more about the code, but otherwise did not move the ball forward. After lunch I worked on the stsdas help dbmopen problem. I wrote a test script which also does not ... More
One Fixed
In the morning I continued to debug drizzlepac by adding more prints. In the afternoon I rewrote the help database code to use ties instead of dbmopen. More
Fractional Zero
I installed yesterday's fix to the stsdas online help and closed the call. I kept tracking the cause between the old and new versions of drizzlepac. There appear to be two causes. The first is a difference in the results between the old and new version of the WCS librararies. This, thankfully, is ... More
Almost There
Changing the data type of pixmap from float to double made almost all the round off errors in test_driz go away. However, the errors in test_blot are still too large to pass the test (median error < 10^-6 maximum error < 10^-5.) Will have to find and fix that error before moving on. More
Got It
The problem with doblot has been resolved. I'll check in the code tomorrow and run the regression tests again. More
Next Change
I checked in the changes to drizzlepac to guthub. Then I created a new branch to change the code to use the numpy array structures in C instead of C pointers. I've converted the simpler cases in the code, but not the more complex context structure handling. I'm now working on updating the unit ... More
Debugging Time
I updated the unit tests and fixed all the syntax errors when compiling. I'm now trying to get the unit tests to pass. So far, none of the C tests pass. More
All Pass
I got all the unit tests to pass for the last code change. I added some new tests for some of the alternative blot interpolators, as they use some of the changes that would not otherwise be tested. More
More Fixes
I rewrote all the code that accesses the context image to use a new accessor. All the tests pass after the change. I also eliminated a lot of code that is no longer used. More
Keep Plugging On
I added a new test to see if the context_output is set correctly. I converted all the image sizes to get their values from the Python image dimensions. This is not yet debugged. I wrote most of a replacement for calculating the image overlap. I'm not planning to integrate this change until higher ... More
Next Task
I got the changes to access the dimensions of an image from the PyArray structure to pass all the unit tests. I checked the changes into git and merged the branch. I started changing the code to replace map_value with map_point. Map_value maps a line at a time and uses temporary arrays. Map_point ... More
Still Cranking
I finished the changes to convert all the calls in cdrizzlebox and cdrizzleblot to use either get_pixel if whole number inputs, or map_point if not. This allowed me to delete the temporary arrays used to hold a line of points, which simplified the code a lot. Most tests pass, but there are a few ... More