martin53
11th February 2015, 21:25
After I made the 1st plugin, and before I write the doc and publish it, I felt the need for a test suite.
So here I'd like to hear suggestions, see attached files.
The test is implemented in a batch file (Test.cmd, run it to execute the test), which sets an environment variable in a test loop.
In the corresponding test script, the RT_Stats plugin can make the variable available to the script.
This variable controls the essential variations of the plugin.
I opened all variations once, and when I found the output OK, I saved the frame as the future automated test reference, 0.png to xx.png (this does not work with noise or random numbers in scripts, I stepped into that trap :o)
For the evaulation, the plugin output is compared to the resp. reference frame with the RGBDifference function in a runtime environment. And depending on the result, either "pass" or "fail" files are outout for each test step.
For a fail, the actual output is also written for easy comparision.
As the host to open the script, I chose AvsMeter.
EDIT
avs and cmd files have evolved quite a bit during the last weeks, so find actual version here (http://www.mediafire.com/download/3496gbme5pq678c/Test_20150308_m53.zip). It is a test for the TfNLMeans fork i'm playing with.
So here I'd like to hear suggestions, see attached files.
The test is implemented in a batch file (Test.cmd, run it to execute the test), which sets an environment variable in a test loop.
In the corresponding test script, the RT_Stats plugin can make the variable available to the script.
This variable controls the essential variations of the plugin.
I opened all variations once, and when I found the output OK, I saved the frame as the future automated test reference, 0.png to xx.png (this does not work with noise or random numbers in scripts, I stepped into that trap :o)
For the evaulation, the plugin output is compared to the resp. reference frame with the RGBDifference function in a runtime environment. And depending on the result, either "pass" or "fail" files are outout for each test step.
For a fail, the actual output is also written for easy comparision.
As the host to open the script, I chose AvsMeter.
EDIT
avs and cmd files have evolved quite a bit during the last weeks, so find actual version here (http://www.mediafire.com/download/3496gbme5pq678c/Test_20150308_m53.zip). It is a test for the TfNLMeans fork i'm playing with.