Should more classes have __eq__ and/or __repr__

[schwehr]

Not all classes have __eq__ and __repr__. This was originally mentioned in #218 when I was writing tests for the Link class. It's okay to not have one or both of these in a class, but it's worth a quick discussion. __eq__ is mostly useful for testing. Can/should the __repr__ be something that can be pasted to create an instance of that thing? That seems tough with the structure of pystac.

find pystac -name \*.py | xargs grep '__eq__'
pystac/serialization/identify.py:    def __eq__(self, other):
pystac/serialization/identify.py:        return not self.__eq__(other)

and

find pystac -name \*.py | wc -l
28

find pystac -name \*.py | xargs grep --files-without-match '__repr__' | wc -l
20

find pystac -name \*.py | xargs grep '__repr__'
pystac/collection.py:    def __repr__(self):
pystac/extensions/pointcloud.py:    def __repr__(self):
pystac/extensions/pointcloud.py:    def __repr__(self):
pystac/extensions/label.py:    def __repr__(self):
pystac/extensions/label.py:    def __repr__(self):
pystac/extensions/eo.py:    def __repr__(self):
pystac/extensions/eo.py:    def __repr__(self):
pystac/link.py:    def __repr__(self):
pystac/catalog.py:    def __repr__(self):
pystac/item.py:    def __repr__(self):
pystac/item.py:    def __repr__(self):
pystac/serialization/identify.py:    def __repr__(self):
pystac/serialization/identify.py:    def __repr__(self):

[duckontheweb]

__eq__ is mostly useful for testing.

Related to #380.

This is also useful for checking whether some kind of collection contains an object. Right now, the only case of this that I know if is in ItemCollection, but it's possible that in the future we would want to define __contains__ on Collection and Catalog instances as well so we can do tests like:

item: Item = ...
collection: Collection = ...

if item in collection:
    # Do something interesting...
    ...

Can/should the __repr__ be something that can be pasted to create an instance of that thing? That seems tough with the structure of pystac.

Agreed, I think this would end up in a __repr__ that is too overwhelming. See the discussion in this comment in #331 for some options. Aligning the outputs of __repr__ with our definition of equality seems like a good way to go.