Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.stsci.edu/~bsimon/weblog/y2014/m09/index.html
Дата изменения: Wed Oct 1 00:25:03 2014
Дата индексирования: Mon Apr 11 11:26:40 2016
Кодировка:

Поисковые слова: п п п п п п п п п п п п п п п п п п п п п п
Archive for September

Archive for September

More

More

More

More

More

More

Next Up

I got all my tests for check_over to pass and moved on to writing the first test for do_kernel_square. Unfortunately, it dumps core. So that is the next problem to solve. More

Patience Is a Virtue

I solved the segmentation violations and got do_kernel_quare to run without crashing, but not much more than that. Trying to figure out the code which sums the image contributions and see why it isn't doing what I expect. More

Forward Motion

I got the first test case of do_kernel_square, which does no scaling, to work. Next step is to figure out all the scaling options and coordinate offsets so I can make sense of the code and write more tests. More

Reassessing

I took a step back and studied the code and asked Warren some questions about it. I believe I now have a good handle on how the square kernel works. There's a problem: some of the code assumes one-indexing instead of zero-indexing. I do not believe this is still valid after I changed the ... More

Fixing Something

I got all the tests working again and modified the pixmap code to create a float array instead of a double and use a half pixel grid. The latter change is going to slow things down when computing the mapping, so I'm not sure if I'll keep it. More

Back Tracking

I dumped the half pixel grid, as it took too long. I got all the test cases to run and then modified the cdrizzlebox code to be zero based instead of one based. This is necessary because we now have a pixel to pixel mapping instead of a WCS to pixel mapping in the code and pixel coordinates are ... More

More Tests

I got the existing tests to pass the code after the one to zero code transition and wrote some more. There are still a few problems that need clearing up. More

Solved But Not Satisfied

I've written enough tests to satisfy myself that my change to the indexing in do_kernel_square is correct. But I still have questions about how the flux is summed in the code. This is code I haven't touched and should be unaffected by my changes. I will ask about that tomorrow. More

Another One Down

I wrote unit tests for dobox and go them to pass. These tests also tested the turbo kernel, which is the next highest priority after the square kernel. Dobox is the top of the hierarchy of code that needs to be tested for errors after rewriting map_value. But I still need to make parallel changes ... More

Tblot

I modified the code to doblot, similar to the way I modified dobox to fix the off by one problem with the new pixel mapping. I wrote the first unit test, but ran into a puzzle where the calling sequence of the interpolation routine doesn't seem to match how it is being called. More

Doblot works

I wrote two unit tests for doblot and got them to work. Then I went back and changed the calculated pixmap so that it calculates coordinates for the center of each pixel instead of the corners. And I got the tests to work again. The next step is to get the regression tests to run. More

Cleaning Up

I restored the ability to subset the image being drizzled even though the code does not currently use it. I cleaned up stuff for subversion, getting ready for the big merge. I checked the code against the Drizzle Handbook and it matches the formulas in there. But I still have some qualms about how ... More

Calculations

I did some more calculations using the drizzle handbook formulas to see if I could make sense of my results. Everything checks out against the code. I started setting up the regression tests. I copied the files over but did not get beyond that. I wrote down some of my notes for the fields in th p ... More

Phase 2

I revised my unit tests and convinced myself the problems were caused by edge effects. No code change needed. I started running the regression tests and fixing the problems reported. More

Stalled

Had trouble figuring out why drizzlepac wasn't found by the regression tests. Will ask Christine's help tomorrow. I went back to studying and writing some trial Cython and C interface code. More

Moving Again

With Christine's help I was able to continue regression testing and resolved three problems with the drizzlepac code. More

Hunting for Errors

I spent the day trying to figure out the rather cryptic error messages and find more information. All I know so far is that one error (-11) is generated in turbo mode and the other (-6) when writing files. More

More Error Hunting

I tried to characterize the errors by determining what tests caused which methods. After talking with Christine, I decided to ignore the official logging and add my own. I tracked the error to an attempt to access pixel [1023,1024] , which looks like the image bounds aren't being computed ... More

One Down, at Least One More

I found the error in the drizzle code, which was an attempt to access beyond the end of the input image: [1023,1024]. I still have a problem with blotting crashing in5 out of 15 test cases. This is odd because it doesn't happen near the edge of the imge, it's at pixel [999,575]. More

Bad Data?

I'm looking for the source of the error in doblot. It looks like the pixmap itself has bad data, which would be a result of a bad WCS in the test files. I will need to investigate this further. More

Back Filling

I modified the code to test for the case when two images do not overlap and then short circuit the execution. I'm testing the change. More