Can we set up a different dispatch predicate per data loader when using ScheduledDataLoaderRegistry?

[srinivasankavitha]

I'm trying to use the experimental ScheduledDataLoaderRegistry. We need to set up a different dispatch predicate per data loader. Is it possible to have multiple ScheduledDataLoaderRegistry instances, each with different dispatch predicates?
There seems to be only one data loader registry that we can set up with the execution input. Is there another way to have a different dispatch predicate per data loader within the ScheduledDataLoaderRegistry?

ExecutionInput executionInput = newExecutionInput()
        .query(getQuery())
        .dataLoaderRegistry(registry)
        .build();

[bbakerman]

in graphql-java - no - there is only 1 dataloader registry used

What do you want multiple for - certain dataloaders get different parameters??

You could write a delegating DataLoaderRegistry that some how finds them from a group of them maybe??

[bbakerman]

Maye something like

    class DelegatingDataLoaderRegistry extends DataLoaderRegistry {
        private final List<DataLoaderRegistry> registries;

        public DelegatingDataLoaderRegistry(List<DataLoaderRegistry> registries) {
            this.registries = registries;
        }

        @Override
        public List<DataLoader<?, ?>> getDataLoaders() {
            List<DataLoader<?, ?>> loaders = new ArrayList<>(super.getDataLoaders());
            registries.forEach( dlr -> loaders.addAll(dlr.getDataLoaders()));
            return loaders;
        }

        @Override
        public Map<String, DataLoader<?, ?>> getDataLoadersMap() {
            Map<String, DataLoader<?, ?>> loaders = new LinkedHashMap<>(super.getDataLoadersMap());
            registries.forEach( dlr -> loaders.putAll(dlr.getDataLoadersMap()));
            return loaders;
        }

        @Override
        public <K, V> DataLoader<K, V> getDataLoader(String key) {
            DataLoader<K, V> dataLoader = super.getDataLoader(key);
            if (dataLoader == null) {
                for (DataLoaderRegistry registry : registries) {
                    dataLoader = registry.getDataLoader(key);
                    if (dataLoader != null) {
                        return dataLoader;
                    }
                }
            }
            return null;
        }

        @Override
        public Set<String> getKeys() {
            return getDataLoadersMap().keySet();
        }

        @Override
        public void dispatchAll() {
            super.dispatchAll();
            for (DataLoaderRegistry registry : registries) {
                registry.dispatchAll();
            }
        }

        @Override
        public int dispatchAllWithCount() {
            int count = super.dispatchAllWithCount();
            for (DataLoaderRegistry registry : registries) {
                count += registry.dispatchAllWithCount();
            }
            return count;
        }

        @Override
        public int dispatchDepth() {
            int count = super.dispatchDepth();
            for (DataLoaderRegistry registry : registries) {
                count += registry.dispatchDepth();
            }
            return count;
        }

    }

where the main registry is really a list of others ??

This is a rough bit of code - eg should you be alllowed to "register new DL" into this combining one?? Or should computeIfAnsent work or not?

I am not sure - I just whipped this up as an idea

[srinivasankavitha]

Thanks so much. I think this makes the most sense for our use case. Will try this out. Do you think this would be something worth supporting in the future as part of the java-dataloader? Or is it too specific a use case?

[bbakerman]

Nah perhaps it would be a good idea

[LeiWangHoward]

Any specific reason we don't want to support per dataloader dispatch predicate configuration?

• The current design of DispatchPredicate is already testing the criteria/predicate per-dataLoader: https://github.com/graphql-java/java-dataloader/blob/master/src/main/java/org/dataloader/registries/DispatchPredicate.java#L21
• Use case for the per-dataLoader predicate configuration: different batch function may have different limit based on upstream API (e.g. myBatchGetUsers might have maximum user limit 5, thus I want to configure queue depth predicate to 5 for myBatchGetUsers dataLoader, while have queue depth predicate set to 10 for other dataLoaders.
• Netflix DGS supports this feature: https://netflix.github.io/dgs/data-loaders/#scheduled-data-loaders-with-dispatch-predicates Potentially we can have DispatchPredicate as a parameter of BatchLoader

Any plan to support this feature, or it's okay to submit community pull request?

[bbakerman]

This is now supported in 3.2.1

[bbakerman]

See https://github.com/graphql-java/java-dataloader/pull/130/files

[srinivasankavitha]

This is great, thank you!