Add a WithKeyedResult option to the result pool

[zhongdai]

I have been using the result pool to fan-out some I/O operations, eg. fetch data from database. The code is something like,

p := pool.NewWithResults[[]record]().WithErrors().WithContext(ctx)

// queries is []string type
for _, query := range queries {
  p.Go(
    func(context.Context) ([]record, error) {
      return databaseCall(ctx, query)
    })
}

result, err := p.Wait()

In this example, the record is a struct to map to a database record, and each databaseCall() would return N records as a []record type.

If any of the query failed to run (eg. timeout), the result (typed [][]record) contains only the succeed returned []record. So the len(queries) != len(result).

In the case, I need to re-try the failed queries, I don't know which ones are failed. So if we can have a option to return a map[string][]record, we can check if which batch(es) are failing.

[camdencheek]

Hi @zhongdai! Thanks for the request. I have a draft PR open to make the set of results from a ResultPool match the order of the tasks were submitted in. In combination with p.CollectErrored(), that would mean that result[i] always corresponds to queries[i]. Does that solve your use case?

Side note: now that result order is deterministic, I'm considering making CollectErrored() default for the 1.0 release. However, that's a breaking change, so I didn't implement that in the original PR.