Implement unit tests

[tomchadwin]

@mtravis So the way this works is that you should write tests in Python, give them a filename starting test_, and put them in test. There are already some there which are provided out of the box by the Plugin Builder. You can then either run those tests individually to check they pass, or use nose to run them all at once.

@akbargumbira (I think) wrote a post describing how to run such tests on Windows, but generally, they only run on Linux/Mac.

Here's an initial list of possible tests, to expand as we think of more:

• xy_to_osgb gives a correct result
• xy_to_osgb outside BNG coverage behaves gracefully
• xy_to_osgb with different precision gives results with the correct precision
• Changing precision spinner changes precision setting
• Clicking with the tool writes to the clipboard

[tomchadwin]

I'm not sure what we want to test with this, as I've never used it. Who can think of some test cases?

[mtravis]

@sklencar any ideas?

[mtravis]

Essentially the plugin uses the clicked map co-ords and converts them using the xy_to_osgb function.
Could we feed a set of co-ords into that to test to see it is outputting correctly?

[tomchadwin]

Yes. The idea of unit testing is to break tests down into multiple smaller tests, each of which only tests one thing. So, off the top of my head:

1. Clicking a coordinate gets the right result
2. Clicking outside BNG coverage behaves gracefully
3. Clicking with different precision gives results with the correct precision

Have you had many bugs or problems? If you have, think of tests which would have spotted them. If you can cover most of the code with tests, it will make it easier to handle eg QGIS API changes.

The bad news is that my experience of writing QGIS tests is very specific - changing various settings, and ensuring qgis2web exports given code. In other words, my tests always end with comparing an output string to a control. Some of your tests will do that, but others might do something else. We'll see what we come up with.

[sklencar]

Tests above look reasonable. I would like to add:
4. "if value has changed in precision spin field, it has changed also the tool precision". Its a related to 3. test, but I guess that one should cover only the convert function.
5. Clicking a coordinate override clipboard with current converted value

[tomchadwin]

Is 5 the current behaviour? I think that should be configurable by checkbox, and default to off - people could lose data.

[sklencar]

Yes, so far it works like that. After opening bbox with info about point a user clicked on, the clipboard is always overwritten. Agree, that it should be configurable.

[tomchadwin]

OK, first test is there, runs on Travis, and passes. The test itself is the function test_xy_to_osgb:

https://github.com/mtravis/gridref-qgis/blob/master/test/test_gridref.py#L32

To write further tests, add more functions after it, within QGISTest, and use the different unittest assert methods to test for the desired results.

Hope this is comprehensible. I'll close this issue, but do ask question here if not.

[tomchadwin]

Right, I'll perhaps try to add one of those markdown checkbox lists to the first post here of possible tests, so we can keep track of what we've done, and add others if we think of them.

[mtravis]

We could run a test to check out the "point out bounds" works.

I think also that they want to add this tool to processing toolbox so we should have tests for batch processing but @sklencar will know more about that.