feat: .get(<cid array>) from the DAG API

[daviddias]

So that the DHT can handle multiple block requests without blasting hundreds of requests. Needed for ipfs/js-ipfs#856

[daviddias]

Quick correction, this module uses the DAG API and therefore needs and function call there that can return multiple IPLD nodes.

This is intertwined with the GraphSync effort. @vmx do you already have code for this?

[vmx]

No, I haven't any code for this yet.

[daviddias]

@achingbrain now that @vmx shipped getMany on IPLD, can you look into shipping this one? It is needed for the DHT effort ipfs/js-ipfs#856

[achingbrain]

Just checking what the requirement is here - after chatting with @vasco-santos it doesn't sound like there's much to do from a DHT Endeavour perspective?

We can expose ipld.getMany as ipfs.dag.getMany (or perhaps just chain through if an array of CIDs is passed to ipfs.dag.get), and then this module can use it when exporting a DAGNodes worth of DAGLinks which should be a nice performance improvement, but it doesn't sound like that would hold up the DHT effort.

@diasdavid have I missed something? What changes did you have in mind that would resolve this issue?

[daviddias]

It doesn't need to block the DHT effort but it will quickly be a issue if we do a DHT query for every single block we need to fetch. Using the .getMany is a first step towards a maintainable solution.

[achingbrain]

I've implemented this in the shiny new decomposed ipfs/js-ipfs-unixfs-exporter repo.