add drop support to Sequence

[tisdall]

To deal with #224. Sequence/SequenceSchema now supports null and drop in the same way Mapping does.

[tisdall]

Okay, tests in place.

[tisdall]

I'm thinking this functionality should be moved to SchemaNode.serialize() and SchemaNode.deserialize() instead of being implemented in each separate type. Otherwise, independently created types won't properly handle colander.drop.

[tisdall]

I guess handling drop only makes sense in particular types that are collections of other SchemaNode. So, it'd require some sort of check in SchemaNode to determine if we're dealing with a collection type and then iterating over values looking for a colander.drop. Doing that with a cstruct would be easy, but doing it with an appstruct would be very tricky.

I'm leaning towards just implementing this in Sequence and Tuple and then modifying the documentation to indicate that the handling of missing=colander.drop is handled by the container type and any particular container may or may not implement that feature.

What do other people think about this?

[tisdall]

Looking at this some more... drop on a tuple element doesn't make much sense since the number of elements has to remain the same. So, I think this patch is good, the documentation just needs fixing up.

[pallix]

But should a tuple having (value, None) be dropped of its parent collection or not?

[tisdall]

@pallix - can you elaborate? I don't understand.

My point is as follows:

Let's say you have a TupleSchema like this:

class Coord(colander.TupleSchema):
    x = colander.SchemaNode(colander.Int(), missing=colander.drop)
    y = colander.SchemaNode(colander.Int(), missing=colander.drop)
    z = colander.SchemaNode(colander.Int(), missing=colander.drop)

The result of serialization should always be a 3-element tuple in this case. It doesn't make sense to drop an element as you would with a dict or list because there's no way to figure out which element was the one dropped. For example, if the serialized value was (3, 2), which coordinate axis is missing? Same issue with deserializing a tuple with not enough elements. If you really want to drop values then you should be using a Mapping in this case.

[pallix]

Ok I see your point ; I'm relatively new using this library so I will trust your competence.

[tisdall]

@pallix - new... old... if you see something wrong with my logic, let me know. :)

[tisdall]

I know right now there's no "validation" on missing values, but I'm wondering if an exception should be thrown when you try to using missing=colander.drop in situations it doesn't make sense (such as on a TupleSchema). Thoughts?

[pallix]

It's always nice to have as much validation as possible as earlier as possible.

[tisdall]

Okay, I think this is good to be merged now.