prototype of udp based job manager #1
Conversation
soxofaan
left a comment
soxofaan
left a comment
There was a problem hiding this comment.
a couple of notes
If I understand correctly, this PR adds two separate features to the existing job manager:
- producing jobs from a fixed (but parameterized) UDP and a user-provided dataframe of parameter
- Running the job manager in a thread
These features seems to be totally unrelated, so I wonder if they can't be separated.
For example:
- the producing of the jobs could be a factory for a standard job manager
- the threaded running could a method on the standard job manager
| if self.dataframe is None: | ||
| self.dataframe = jobs_dataframe | ||
| else: | ||
| raise ValueError("Jobs already added to the job manager.") |
There was a problem hiding this comment.
this if else raise pattern looks like this could have been a constructor argument
| p.get("schema", {}).get("subtype", "") == "geojson"] | ||
|
|
||
|
|
||
| output_file = Path("jobs.csv") |
There was a problem hiding this comment.
This static file reference should be an argument I guess
|
|
||
| cube = connection.datacube_from_process(row.udp_id,row.udp_namespace, **parameters) | ||
|
|
||
| title = row.get("title", f"Subjob {row.udp_id} - {str(parameters)}") |
There was a problem hiding this comment.
use row index instead of str(parameters) in title to avoid extremely large titles?
|
|
||
|
|
||
|
|
||
| import multiprocessing, time |
There was a problem hiding this comment.
these imports can be toplevel I think
|
Because of various changes to the "official" |
|
just merged Open-EO/openeo-python-client#644 |
Proposal for a UDP based variant of the job manager, created as part of APEx upscaling service:
https://jdries-vito.quarto.pub/apex-design/upscaling.html
Related issue is to support output to geoparquet: Open-EO/openeo-gfmap#107
The currently used csv format is limited in the sense that complex parameter types fail to deserialize correctly, requiring custom handling in this class. GeoParquet might improve this:
https://github.com/apache/parquet-format/blob/master/LogicalTypes.md#json