Closed dkrako closed 9 months ago
why do we want to specificly tests judo and gazebase, or will we do similar things for every dataset?
I think we should do that for every dataset.
In #492 we will implement a fixture to create GazeDataFrame instances from these files. These can be then used in acceptance / functional tests (instead of unit testing), i.e. we will define short user stories and check outcomes on a higher level. Column and schema checking should be mostly sufficient, as the specific validity of the results must be tested in unit tests.
As long as we don't solve the problem for caching public datasets, this is the best way I can think of to automatically test our pipeline on example data from public datasets.
This way I guess as a byproduct we will identify a test case that reproduces #517 in a functional test.
Description of the problem
A prerequisite for working on #492 is that we have test data available that is similar to the data in each dataset.
Description of a solution
Create small example files for each dataset.
Each file should contain the same header as in the original, but the rows should be limited to 10.
Minimum acceptance criteria
gazebase_example.csv
intests/gaze/io/files
gaze_on_faces_example.csv
intests/gaze/io/files
gazebase_vr_example.csv
intests/gaze/io/files
judo1000_example.csv
intests/gaze/io/files
hbn_example.csv
intests/gaze/io/files
sbsat_example.csv
intests/gaze/io/files
tests/gaze/io/csv_test.py