TUW-GEO / pytesmo / 978 / 3

85%

master: 88%

push

travis-ci

Val Framework compares dataset matches that do not include the temporal reference

This changes the behavior of the validation framework.

Before this a validation run with `(n, k)` where `k < n` only gave results for
combinations of datasets which included the temporal reference.

e.g. If 3 dataset (DS1, DS2 and DS3) where given and a metric calculator was run
with `(3, 2)` with the temporal reference DS1. Then only the combinations of
`DS1 - DS2` and `DS1 - DS3` gave results.

This rectifies this behavior to also return results for the combination `DS2 -
DS3` in the above example.

Tests still need to be adapted but I want to discuss this first to see if this
is what people expect.

There are also some edge cases. e.g. If 4 datasets are given but the
configuration of the metric calculator would be `(3, 2)`. I think we should
raise an error in these cases because that would mean that the temporal
reference is not included in all matches.

628 of 794 branches covered (79.09%)

1703 of 1996 relevant lines covered (85.32%)

0.85 hits per line