Why the function add_dynamic_partitions can't handle a Partition object list? Dynamic partitions can handle every kind of partition using deprected param partition_fn, but only static string partitions using the experimental name param.

def get_partitions():
    ps = blah blah
    return [Partition(name=p.n, value=p.v) for p in ps]

partition_def = DynamicPartitionsDefinition(get_partitions) # OK

def get_partitions():
    ps = blah blah
    return [Partition(name=p.n, value=p.v) for p in ps]

partition_def = DynamicPartitionsDefinition(name="name")
context.instance.add_dynamic_partitions(partition_def.name, get_partitions()) # NOT OK

Hi Cedric. We're moving away from supporting the

Partition

object and instead just use string partition keys. For that reason the new dynamic partitions only accept string values

Ok thanks ! Why remove Partition object ? In my case, I want to build a run_config with more complex information than just the key, there is something for this use case ? I will add a cutom local key-value storage and map my partitions key to the config, but it looks like a hack.

I believe we removed it because we weren't actually using fields on

Partition

object anywhere in the system, only using the partition key. So development was sometimes cumbersome having to build/fetch this

Partition

object throughout the system. Have you seen our docs on partitioned config? https://docs.dagster.io/concepts/partitions-schedules-sensors/partitions#date-partitioned-job If you want to pass in additional fields to run config other than just the partition key, I would recommend using partitioned config.