This avoids a problem exposed by netcdf saves in Iris 1.10,
but which is actually due to a bug fixed by the major indexing rewrite in numpy 1.9.
See : https://github.com/SciTools/iris/issues/2120
I've patched all indexing to convert tuples (where the problem was) to lists instead.
Notes:
replaces #188, except that this converts tuples to lists instead of arrays
converting _all_ 'vector_key' types did not work, as making arrays into lists upsets the special case of indexing with a boolean array
Checked : the Iris 1.10 tests still pass with these changes.
This avoids a problem exposed by netcdf saves in Iris 1.10, but which is actually due to a bug fixed by the major indexing rewrite in numpy 1.9. See : https://github.com/SciTools/iris/issues/2120
I've patched all indexing to convert tuples (where the problem was) to lists instead.
Notes:
Checked : the Iris 1.10 tests still pass with these changes.