provide dependency injection for schema serialization about plone.restapi

[jone]

Should this maybe be a two-way solution so that we can prepare data for updating / creating?

[tisto]

Please note that the schema serialization was just a proof-of-concept. That was never meant to be the final code.

Your questions:

1. Not that I would know.
2. Could go into plone.restapi first, could be refactored into a separate package as well.
3. Looks ok to me. Adapters always work. ;)

[jone]

Ok, thanks. I'll make a PR 😉

[lukasgraf]

Looks good to me :)

Is there already some serialization mechanism which produces JSON-compatible output (!) for dexterity?

I wouldn't know of one that does exactly this, but:

• plone.supermodel provides XML import and export for schema interfaces based on zope.schema fields
• plone.rfc822 does marshalling of content objects described by zope.schema fields from and to RFC (2)822 messages (E-Mail / HTTP-Messages).

So even though they do something different, it still could be worth a look for some tricky cases, maybe we can pick up some odd bits and pieces from these packages.

[lukasgraf]

I think the IFieldSerializer adapter and your choice of discriminators is the way to go, but at some later stage we might need another layer of abstraction:

We'll have use cases where we'll want to serialize objects that aren't Plone content types (think SQL backed content for example). When using SQLAlchemy the Column objects could play a similar role to fields, in that the describe the data type, but I could imagine cases where you don't even have a field to adapt - you just know you're dealing with a datetime, and you want to be able to re-use the conversion to a JSON-compatible ISO format instead of writing it yourself.

So maybe at some later stage, we'd need some concept like plain Converters, where a DateTimeFieldSerializer would delegate to a DateTimeConverter instead of doing the conversion itself.

[buchi]

I've got two suggestions of cosmetic nature:

• IFieldSerializer should adapt the field first as it is the main thing that is adapted:
  def __init__(self, field, context, request)
• Do serialization in the __call__() method instead of a serialize()method. This reduces verbosity in code serializer.serialize() vs. serializer()

I've started implementing deserialization in #41 following this approach.