About how processor to split metrics name into datapoint's key-value pair attributes

[mufiye]

Getting the statsD metrics data from airflow collected by our opentelemetry collector, I want to process the data by key-value pair attributes. But all metadata are in the name, such as "ti.finish.<dag_id>.<task_id>.", we want to add "dag_id" and "task_id" as the key-value pair attributes to datapoint.
I have tried to use the metrics transform processor, transform processor and attributes processor.
The metrics transform processor's combine function can process the counter and gauge data, but can not process summary and histogram data. It will encounter some problems when processing histogram data it will report errors in the terminal, and for summary data, the datapoint part of the result data is null.
For the transform processor, I could use functions like replace_pattern to make the metrics name concise, such as "ti.finish.<dag_id>.<task_id>." to "ti.finish", but I have no idea how to add the <dag_id> and <task_id> into the attributes, maybe the Split function to split name by "." and set function, but it does not work.
Do you have any solution or suggestion to my problems? Thank you very much.