Arvados

Using the Arvados API from Python should require much less boilerplate than it does now. We should offer an API that is more convenient and looks more familiar to bioinformaticians and other Python programmers.

For example, this

from arvados.models import Specimen
bar = Specimen.all()
foo = Specimen.find(want_uuid)
foo.properties['baz'] = 'baz'
foo.save()

is more friendly than this

import arvados
bar = arvados.api().specimens().list().execute()['items']
foo = arvados.api().specimens().get(uuid=want_uuid).execute()
foo['properties']['baz'] = 'baz'
arvados.api().specimens().update(uuid=foo['uuid'], body=foo).execute()

Django seems like an obvious interface to (loosely) mimic: it's likely to be familiar to Python developers.

• Django docs → Managers (sorta like Active Relation in Rails)
• Django docs → Models (sorta like Active Record in Rails)