Ingestion vs. Handling

[dylanmcreynolds]

Currently, the setup.py in this uses the databroker.handler entrypoint to register ingestors. It should be switched to databroker.ingestor.

Along those lines though, talking with @ihumphrey , maybe this repo could serve dual-duty. The repo already exists as a place to put ingestors that have common dependencies that we don't really want in every xicam instance (as discussed with @ronpandolfi ). An ingestor and a handler are distinct, but for common file types, their dependencies will be almost identical. So, for a given type, maybe we can define an ingestor (a generator of documents) and a handler (a callable that provides a numpy array for a given spec.) Maybe a class with both a call and an ingest method? They would be related based on the spec field of the resource document. I ingest tiffs and set resource.spec='TIFF', then 'TIFF' is added to the list of specs that the related handler handles.

FWIW,

Ingestors vs handlers are a little confusing. I know of three distinct use

1. I have files I want in databroker / mongo. I therefore need a document stream. Create an ingestor, serialize documents. to mongo.
2. I have files I want to open in xicam, but I don't care about mongo. Create an ingestor to generate a document stream, xicam uses that generator when the user selects a file
3. I have things that have been ingested into databroker, and that has a spec like "TIFF" or "fccd". Now, I open a dataset in xicam. I registered a handler for that spec already,so it gets invoked when I need to produce a numpy array from the data file.