r.buffer: Added test script #4482
Conversation
github-actions
bot
added
raster
|
|
||
| self.assertRasterExists(output) | ||
|
|
||
| expected_categories = [1] + [i + 1 for i in range(len(distances) + 1)] |
There was a problem hiding this comment.
Why do you want 1 twice?
| expected_categories = [1] + [i + 1 for i in range(len(distances) + 1)] | |
| expected_categories = [i + 1 for i in range(len(distances) + 1)] |
There was a problem hiding this comment.
the first 1 is the default category of the road, then i start i from i+1 because i already have a default 1 in the expected category list. The last +1 in len(distances) + 1 is because categories start from index 1.
I will conduct further tests to understand this in a deeper context, and apply corrections accordingly.
There was a problem hiding this comment.
You can also look at the start and stop parameters of range if you'd feel it would make it simpler
There was a problem hiding this comment.
Added the suggested change
| category_values = gs.read_command( | ||
| "r.stats", flags="n", input=output | ||
| ).splitlines() | ||
| category_values = [int(line.split()[0]) for line in category_values] | ||
|
|
||
| print("Category values:", category_values) | ||
|
|
||
| unique_actual_categories = set(category_values) | ||
| self.assertTrue( | ||
| unique_actual_categories.issubset(set(expected_categories)), | ||
| msg=( | ||
| "Output categories should be a subset of expected categories: " | ||
| + str(expected_categories) | ||
| ), | ||
| ) |
There was a problem hiding this comment.
If you want to test the exact categories, than the previous min/max test is somewhat redundant. Why subset, the categories should be compared exactly, no?
There was a problem hiding this comment.
My thought process here was to first check if the categories are in the expected min max range. However, it doesn't verify if all expected categories are present or if there are extra categories.
Second check was to make sure there are no extra categories or abnormal categories that the user didn't want.
But I can see how it can be redundant. I will work on patching this and making sure a single test is sufficient.
There was a problem hiding this comment.
Removed the redundant test
|
|
||
| self.assertRasterExists(output) | ||
| stats = gs.read_command("r.univar", map=output, flags="g") | ||
| self.assertIn( |
There was a problem hiding this comment.
There was a problem hiding this comment.
In this test I use r.univar to ensure that the output map has no non-null cells when the input is entirely null. According to my understanding r.univar helps calculate univar stats and if it is a null output it gives out n=0 (number of non null cells). This verifies that the buffer operation correctly handles null input data without introducing any non-null values.
There was a problem hiding this comment.
Used assertRasterFitsUnivar instead of manually doing it
| print("Category values:", category_values) | ||
|
|
||
| self.assertNotIn( | ||
| 0, |
There was a problem hiding this comment.
r.buffer won't output 0 in any case, so it looks like it does not test what you want it to test.
There was a problem hiding this comment.
I will work on patching it and diving deeper to understand the exact expected outputs in a few edge cases.
There was a problem hiding this comment.
Compared with expected_categories = [1,2] instead of 0
| refmin=min(expected_categories), | ||
| refmax=max(expected_categories), | ||
| msg=( | ||
| "Buffer zones should have category values from 1 to " # Checking if there are no abnormal values in the output raster map |
There was a problem hiding this comment.
Please use f string here
|
|
||
| expected_categories = [1, 2] | ||
|
|
||
| self.assertNotIn( |
There was a problem hiding this comment.
This test is incorrect, it will pass even if category_values are empty. Could you better explain what exactly are you trying to test here? Also distances can be just 100, since given the input, it will never get to 200 meters...
There was a problem hiding this comment.
I apologize for the multiple changes, I will be more careful with them. New to the team and still learning.
Will recheck these and push the changes soon.
There was a problem hiding this comment.
The main idea here was to make a map with alternating rows of zero and one values. The first check is to see if it indeed made an output map and second is to verify there are no 0 values in the output as -z flag will help ignore zero values in output when creating buffers. This should verify that the -z flag works properly and confirms that buffer zones are not created around zero values and they are ignored.
I made a mistake in the assert. It should have been:
self.assertNotIn( 0, category_values, msg="Output should not contain buffer zones around zero values", )
Do you recommend keeping the test after making the above change and removing expected_categories?
There was a problem hiding this comment.
There will never be 0 in the output, the output values always start with category 1 ("distances calculated from these locations"). So this test is not very useful as it is now. Perhaps you can create a raster with zeros and then run it with -z and then you should get an output raster with nulls only, because the zeroes are ignored and there are no source cells for the buffer, hence nulls everywhere. I recommend running it with different scenarios to make sure you understand what it's doing.
This PR adds regression tests for the
r.buffermodule to ensure its functionality remains intact with future code changes. The following test cases are covered:test_buffer_creation(): Verifies the creation of buffer zones around raster features based on specified distances. It checks that the output raster exists, has expected category values starting from 1, and that the minimum and maximum values match the expected range.
test_no_non_null_values(): Confirms that the
r.buffermodule correctly handles an input raster that is entirely null, ensuring the output contains no non-null values.test_ignore_zero_values(): Ensures that zero values in the input raster are ignored when creating buffer zones, resulting in an output that does not include any zero categories.