IssueTracker

So you can do this now by doing something like:

@Suppress("UNCHECKED_CAST")
class DelegateWorker: Worker() {
    companion object {
        const val DELEGATE_WORKER_NAME = "DELEGATE_WORKER_NAME"
        const val PERIOD = "PERIOD" // in minutes.
    }

    override fun doWork(): Result {
        val delegateName = inputData.getString(DELEGATE_WORKER_NAME) ?: return Result.FAILURE
        val delegateKlass = Class.forName(delegateName) as? Class<out Worker> ?: return Result.FAILURE
        val period = inputData.getLong(PERIOD, 15)
        // build your worker with the constraints you want.
        val request = PeriodicWorkRequest.Builder(delegateKlass, period, TimeUnit.MINUTES)
                .setInputData(inputData)
                .build()
        WorkManager.getInstance()
                .enqueue(request)

        return Result.SUCCESS
    }
}

class PeriodicWorker: Worker() {
    override fun doWork(): Result {
        // Do some periodic work
        return Result.SUCCESS
    }
}

// You want to delay the PeriodicWorker. You can use initial delay on the DelegateWorker, which in
// turn enqueue's PeriodicWorker.

val data = mapOf(DelegateWorker.DELEGATE_WORKER_NAME to PeriodicWorker::class.java.name).toWorkData()
val delayedPeriodicWorkRequest = OneTimeWorkRequestBuilder<DelegateWorker>()
        .setInitialDelay(10, TimeUnit.SECONDS)
        .build()

WorkManager.getInstance()
    .enqueue(delayedPeriodicWorkRequest)

The idea is to create a OneTimeWorkRequest with an initial delay, which in turn calls an enqueue() on the PeriodicWorkReqeust.

This looks like a hack, instead of just adding setInitialDelay method on the periodic worker that will just run once.

I have tried going the route of the OneTimeWorkRequest with an initial delay that triggers a PeriodicWorkRequest but it doesn't seem to work. It would be much nicer to be able to add a simple delay to the first occurence.

#3: This is exactly what an internal implementation would look like because the backing JobScheduler does not support initial delays on periodic work.  I realize that WorkManager doing it for you is nicer, which is why this feature request is still open, but it may take us some time to get to it.

#4: Please file a separate bug if that doesn't work for you.

#5: How could I debug the behaviour of this chain of Work?