# `PgFlow.Worker.TaskRow`
[🔗](https://github.com/agoodway/pgflow/blob/v0.3.1/lib/pgflow/worker/task_row.ex#L1)

Decodes one `pgflow.step_task_record` row into the fields the worker dispatches on.

The row shape is a positional list, so it is decoded in one place rather than
pattern-matched at the call site — that is what lets the worker straddle a
schema upgrade.

## Two shapes

`pgflow_helpers` v03 appends `attempts_count` to the composite type, so
`pgflow.start_tasks` returns eight columns instead of seven. A worker deployed
ahead of that migration — the order this change has to ship in, since the
reverse breaks every dispatch — still receives seven, and reports `attempt: 1`
for them. That is exactly the value the worker hardcoded before v03, so the
un-migrated path behaves as it always did.

    [flow_slug, run_id, step_slug, input, msg_id, task_index, flow_input]
    [flow_slug, run_id, step_slug, input, msg_id, task_index, flow_input, attempts_count]

`attempts_count` is nullable, and a null is read as the first attempt for the
same reason: a missing count must not present as a task that has never run.

# `t`

```elixir
@type t() :: %{
  flow_slug: String.t(),
  run_id: binary(),
  step_slug: String.t(),
  input: term(),
  msg_id: integer(),
  task_index: non_neg_integer(),
  flow_input: term() | nil,
  attempt: pos_integer()
}
```

# `decode`

```elixir
@spec decode([term()]) :: t()
```

Decodes a seven- or eight-column `step_task_record` row.

Returns the row's fields keyed by name, with `:attempt` always a positive
integer.

---

*Consult [api-reference.md](api-reference.md) for complete listing*