Thank you, pieter.
I suppose you will not mind if I quote your response below.
>Hi,
> Your proposal is very close to how we are using the CallStep in Sqlg.
> The CallStep itself is always just a placeholder and is replaced in the
> strategies and the CallStep.serviceName is the marker. Here is the current
> code in the strategy which replaces the CallStep.
> private void handleCallStep(ListIterator<Step<?, ?>> stepIterator,
> CallStep<?, ?> callStep, MutableInt pathCount) {
> String serviceName = callStep.getServiceName();
> switch (serviceName) {
> case SqlgPGRoutingFactory.NAME:
> handlePgRouting(serviceName, callStep, pathCount);
> break;
> case SqlgPGVectorFactory.NAME:
> handlePGVector(serviceName, callStep, pathCount);
> break;
> case SqlgFunctionFactory.NAME:
> handleFunction(serviceName, callStep, pathCount);
> break;
> default:
>
> throw new IllegalStateException("Unknown serviceName: " +
> serviceName);
> }
> }
> I think your approach is very much the same as this so I am onboard with your
> proposal.
> I certianly felt that the grammar should also be generated from the
> @GremlinDsl, but alas having no experience with ANTLR I parked that for when
> I have more time to investigate. For now back to groovy was the quick fix.
My intention right now is to collect all feedback that I can from the
dev list and create a PR for TP that any vendor can check and provide
feedback on, alongside the TP team itself, of course.
I look forward to seeing you involved and probably even using our fork
for testing your extensions.
On Thu, Aug 28, 2025 at 8:00 AM pieter <[email protected]> wrote:
