Generalized retry function

[Alan Noble]

A helper function for flexible retrying is proposed, rather than just retrying every N seconds or M attempts.

For example, the following pattern appears frequently in our code base.

for {

err = before()

if err != nil {

    time.Sleep(retryTime)

    continue

}

   after()

}

Such code would be replaced with the following:

for {

   retry(before, after, policy)

}

The policy would describe the retry policy to be implemented and would permit more flexible retry policies than simply trying again after waiting for a period fo time, e.g.,

type RetryPolicy struct {

wait time.Duration // Wait time between retries.

retries int               // Maximum number of retries, or 0 for forever.

exponential bool   // True for exponential backoff.

}

The before/after functions could be anonymous functions.

[kortschak]

I think having a []func(<some meaningful set of parameters) error would probably be the approach I'd use here. Looking at the docs for iptables would likely be helpful for inspiration.