Backport annotations cause gotcha when diffing round-tripped files

[bbartley]

Our current mechanism for converter tests involves testing the original file with the round-tripped file for an exact match of triples. This creates a complication, since round-tripping adds backport annotations– thus, the original file and its round-trip never exactly match. To get around this, we have to artificially add backport annotations to the original file to ensure it matches the round-tripped file. This is a subtle and annoying gotcha when trying to generate test files for new converter tests.

At the moment, I'm uncertain what is the best solution, but likely it involves some special manipulation of backport annotations by the handle_3to2_conversion and handle_2to3_conversion methods. And the solution in each of these methods will need to be slightly different, since round-tripping starting from SBOL2 may end up with something different than round-tripping starting from SBOL3.

[jakebeal]

One good way to do this would be to add a "compare, ignoring backport annotations" method, which would strip out the backport annotation triples before serializing and comparing.

That will mean that we don't have to modify the original files, and we should always be safe to ignore backport materials, because the only things that will use them are the converters, for which the backport annotation integrity is being implicitly tested by the round-trip.

[jakebeal]

Note that a good source of triple manipulation code that can be adapted for this is the non-native "javascript wrapper" converters, which have to do a bunch of "triple grep" operations in order to deal with known bugs in the non-native converter.