Explicit autobatching

[rahul1]

Background

Right now, Medplum supports time-based autobatching. This works great for front-end applications. For backend / server applications, developers have asked for more control over which requests are part of a batch request.

Proposal

add a startBatch() and endBatch() method to the MedplumClient, to be used as follows:

medplum.startBatch()
const patients = await Promise.all([ 
    medplum.readResource('Patient', '1'),
    medplum.readResource('Patient', '2'),
    medplum.readResource('Patient', '3'),
])
medplum.endBatch()

[dillonstreator]

@rahul1 how is this intended to work in the following example?

async function doThing() {
  medplum.startBatch()
  const patients = await Promise.all([ 
      medplum.readResource('Patient', '1'),
      medplum.readResource('Patient', '2'),
      medplum.readResource('Patient', '3'),
  ])
  medplum.endBatch()
}

doThing()
doThing()

Would subsequent startBatch() calls prior to an endBatch() call block until endBatch() is called?

Maybe a batch identifier could be provided or the startBatch call could return an identifier?

async function doThing(batchIdentifier: string) {
  medplum.startBatch(batchIdentifier)
  const patients = await Promise.all([ 
      medplum.readResource('Patient', '1', { batchIdentifier }),
      medplum.readResource('Patient', '2', { batchIdentifier }),
      medplum.readResource('Patient', '3', { batchIdentifier }),
  ])
  medplum.endBatch(batchIdentifier)
}

doThing("1")
doThing("2")
doThing("1") // This could block until the first call to `doThing("1")` finishes its batch. Ideally, unique batch ids would be provided so this shouldn't be encountered but this behavior would need to be documented.