Re: [amqp] Message filter syntax

On 18 May 2018 at 14:36, Clemens Vasters <clemensv@microsoft.com> wrote:

[... snip ...]

I'm fine with all the above, but note that footer, message-annotations and delivery-annotations have no restriction on the types that may be used for the values used in the map (e.g. the value may be an array, list, map, or even a described type), do we exclude the ability to filter on values of complex type, or do we want to define syntax for this?. There is also the case that theoretically the key in any of these sections can be a ulong, though that behaviour is reserved for use by AMQP definitions and we so far have not used that.

A further consideration is that AMQP does not make the same restrictions on the form of strings/symbols used in the keys of these (or the application properties) map. In particular spaces and operator characters may be used, e.g. a valid key might be "foo > 100". As such I think we need to define a way of quoting identifiers (as SQL does).

[Clemens Vasters] Yes, we will need to define a syntax to navigate into complex types, and we’ll have to pick up the quoted identifier rules from newer versions of SQL. There’s some 21^st century renovation needed on the JMS baseline 😉

It's a fine line to tread on the navigating complex types: if we wanted to be precise we'd not only have to define a complete rendering of the AMQP type system into text, but also define equality over complex types (e.g. if the key in a map is itself a map, then we need to define the notion of equality on maps...) I don't think we want to go overboard in trying to cover every scenario, but likely want to define enough for the common scenarios to work.

[... snip ...]

-- Rob

_____________________________________________________________________________

Red Hat GmbH, www.de.redhat.com,
Registered seat: Grasbrunn, Commercial register: Amtsgericht Muenchen, HRB 153243,
Managing Directors: Paul Argiry, Charles Cachera, Michael Cunningham, Michael O'Neill

amqp message