On Thu, 19 Dec 2019 at 10:11, Ryan Skraba <r...@skraba.com> wrote:
> Hello! You might be interested in this short discussion on the dev@
> mailing list:
> https://lists.apache.org/x/thread.html/dd7a23c303ef045c124050d7eac13356b20551a6a663a79cb8807f41@%3Cdev.avro.apache.org%3E
>
> In short, it appears that the record name is already ignored in
> record-to-record matching (at least outside of unions) as an
> implementation detail in Java. I never *did* get around to verifying
> the behaviour of the other language implementations, but if this is
> what is being done in practice, it's worth clarifying in the
> specification.
>
That's really interesting, thanks! I think this should indeed be clarified
in the specification, as I'm aware of at least two implementations that
follow the letter of the spec in this respect (one is actually too strict
and compares the fully-qualified name).
It's also really useful to know that the most-used Java implementation is
relaxed in this way, because it means that (presumably) most tooling around
Avro will be similarly relaxed. Maybe my cunning plan is a possibility
after all :)
It does seems like a very pragmatic thing to do, and would help with
> the CloudEvents Avro use case. It would be a nice recipe to share in
> the docs: the right way to read an envelope from a custom message when
> you don't need the payload.
>
> I'm not sure I understand the third strategy, however! There aren't
> any names in binary data when writing - what would the alias do?
>
The names that I'm thinking of there are the names in the writer's schema.
When you're reading a message, you have access to both the writer schema
and the reader schema (including all the names). However, for comparison
purposes, writer schemas are often canonicalised, which includes stripping
out all redundant fields, including aliases. If this is done, then although
you'll have access to type names, you won't have access to the alias
information.
tbh, I'm less keen on the above option anyway. I like the idea of a
structural subtype relationship without the need to explicitly predeclare a
type relationship. To me that's somewhat reminiscent of Go's approach to
interface types, which works well in practice.
(Also, I largely prefer your avro version with explicitly typed
> metadata fields and names as well!)
>
Thanks again for your feedback. I'll try making a proposal for a different
CloudEvent format, and try to get some implementations to relax their rules
a bit.
cheers,
rog.
All my best, Ryan
>
> On Wed, Dec 18, 2019 at 5:49 PM roger peppe <rogpe...@gmail.com> wrote:
> >
> > Hi,
> >
> > Background: I've been contemplating the proposed Avro format in the
> CloudEvent specification, which defines standard metadata for events. It
> defines a very generic format for an event that allows storage of almost
> any data. It seems to me that by going in that direction it's losing almost
> all the advantages of using Avro in the first place. It feels like it's
> trying to shoehorn a dynamic message format like JSON into the Avro format,
> where using Avro itself could do so much better.
> >
> > I'm hoping to propose something better. I had what I thought was a nice
> idea, but it doesn't quite work, and I thought I'd bring up the subject
> here and see if anyone had some better ideas.
> >
> > The schema resolution part of the spec allows a reader to read a schema
> that was written with extra fields. So, theoretically, we could define a
> CloudEvent something like this:
> >
> > { "name": "CloudEvent", "type": "record", "fields": [{ "name":
> "Metadata", "type": { "type": "record", "name": "CloudEvent", "namespace": "
> avro.apache.org", "fields": [{ "name": "id", "type": "string" }, {
> "name": "source", "type": "string" }, { "name": "time", "type": "long",
> "logicalType": "timestamp-micros" }] } }] }
> >
> > Theoretically, this could enable any event that's a record that has at
> least a Metadata field with the above fields to be read generically. The
> CloudEvent type above could be seen as a structural supertype of all
> possible more-specific CloudEvent-compatible records that have such a
> compatible field.
> >
> > This has a few nice advantages:
> > - there's no need for any wrapping of payload data.
> > - the CloudEvent type can evolve over time like any other Avro type.
> > - all the data message fields are immediately available alongside the
> metadata.
> > - there's still exactly one schema for a topic, encapsulating both the
> metadata and the payload.
> >
> > However, this idea fails because of one problem - this schema resolution
> rule: "both schemas are records with the same (unqualified) name". This
> means that unless everyone names all their CloudEvent-compatible records
> "CloudEvent", they can't be read like this.
> >
> > I don't think people will be willing to name all their records
> "CloudEvent", so we have a problem.
> >
> > I can see a few possible workarounds:
> >
> > when reading the record as a CloudEvent, read it with a schema that's
> the same as CloudEvent, but with the top level record name changed to the
> top level name of the schema that was used to write the record.
> > ignore record names when matching schema record types.
> > allow aliases to be specified when writing data as well as reading it.
> When defining a CloudEvent-compatible event, you'd add a CloudEvent alias
> to your record.
> >
> > None of the options are particularly nice. 1 is probably the easiest to
> do, although means you'd still need some custom logic when decoding
> records, meaning you couldn't use stock decoders.
> >
> > I like the idea of 2, although it gets a bit tricky when dealing with
> union types. You could define the matching such that it ignores names only
> when the two matched types are unambiguous (i.e. only one record in both).
> This could be implemented as an option ("use structural typing") when
> decoding.
> >
> > 3 is probably cleanest but interacts significantly with the spec (for
> example, the canonical schema transformation strips aliases out, but they'd
> need to be retained).
> >
> > Any thoughts? Is this a silly thing to be contemplating? Is there a
> better way?
> >
> > cheers,
> > rog.
> >
>
