RE: #67
Working on allowing eumetsat readers to be passed a list of filenames. Current implementation idea is to stop overriding the _merge_data method from ChronFiles, but instead add a merge method to each of the file classes (AscatL1bEpsFile, etc) that the default ChronFiles._merge_data method already leverages.
This just needs two modifications:
1) The file reader classes' read methods need to be able to accept a list of strings and handle the merging before returning the data if necessary.
2) A modification to somehow allow the merge method to know what product type is being merged (previously, the _merge_data method on the ChronFiles subclass could get this from its self attributes). At the moment I've modified read_eps_l1b and read_eps_l2 to accept a return_ptype bool that would add a "product_type" field to the returned metadata if true. This way a merge method can get the product type of a dataset it's passed by inspecting its metadata. I'm not in love with this approach though and would appreciate other ideas.
Eventually this would of course be applicable to BUFR and NC readers as well
RE: #67 Working on allowing
eumetsat
readers to be passed a list of filenames. Current implementation idea is to stop overriding the_merge_data
method fromChronFiles
, but instead add amerge
method to each of the file classes (AscatL1bEpsFile
, etc) that the defaultChronFiles._merge_data
method already leverages.This just needs two modifications: 1) The file reader classes'
read
methods need to be able to accept a list of strings and handle the merging before returning the data if necessary.2) A modification to somehow allow the
merge
method to know what product type is being merged (previously, the_merge_data
method on theChronFiles
subclass could get this from itsself
attributes). At the moment I've modifiedread_eps_l1b
andread_eps_l2
to accept areturn_ptype
bool that would add a"product_type"
field to the returned metadata if true. This way amerge
method can get the product type of a dataset it's passed by inspecting its metadata. I'm not in love with this approach though and would appreciate other ideas.Eventually this would of course be applicable to BUFR and NC readers as well