General information
- Applicant: COPIL developpement TMMT
Users and administrators needs
A QGIS project could / shoud be tested to ensure:
- Tables in the project exist in the datamodel
- Other ?
Feasibility study
OpenGIS notice
-
In the CI, run tests with demo data as well (currently, we are doing this on an empty database).
It doesn't cost much, but it won't generate any revenue either. I would say that for this point, we could perhaps add a little more testing on the pum side.
-
In the CI, test the project: open the project with Docker Compose QGIS, and check that the layers are valid.
Given that we already have a QGIS running for testing the plugin with Docker Compose, it won't slow down the CI (which is already slow), it doesn't cost much to do, and I think it's worthwhile.
2b) Go a little further and do integration testing:
use the QGIS API to add/modify objects. Here I'm a little more torn about the complexity/benefit ratio, but it's doable.
- On Desktop, run the tests on the instance installed by oQtopus.
This would obviously be the best in terms of reliability. There is a bit of complexity: dependency on pytest, we will have to write code around the management of test results and discuss an approach if the tests fail (after an installation? after an update?) because these tests will probably (I would say almost certainly) have to be run after a commit, so no revert possible.
General information
Users and administrators needs
A QGIS project could / shoud be tested to ensure:
Feasibility study
OpenGIS notice
In the CI, run tests with demo data as well (currently, we are doing this on an empty database).
It doesn't cost much, but it won't generate any revenue either. I would say that for this point, we could perhaps add a little more testing on the pum side.
In the CI, test the project: open the project with Docker Compose QGIS, and check that the layers are valid.
Given that we already have a QGIS running for testing the plugin with Docker Compose, it won't slow down the CI (which is already slow), it doesn't cost much to do, and I think it's worthwhile.
2b) Go a little further and do integration testing:
use the QGIS API to add/modify objects. Here I'm a little more torn about the complexity/benefit ratio, but it's doable.
This would obviously be the best in terms of reliability. There is a bit of complexity: dependency on pytest, we will have to write code around the management of test results and discuss an approach if the tests fail (after an installation? after an update?) because these tests will probably (I would say almost certainly) have to be run after a commit, so no revert possible.