You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Adding support to marshallers to require modules not present and read back approximate versions of the data, as well as lazy loading of required modules
#49
Closed
frejanordsiek opened this issue
Sep 14, 2016
· 2 comments
Adding marshallers to support types in more modules of the main Python libraries as well as external packages would increase loading time due to loading more packages and add dependencies for the way marshallers work in the 0.1.x series.
Thus, the parent marshaller class, hdf5storage.Marshallers.TypeMarshaller is being changed to support lists of required modules that can be lazy loaded when the marshaller is actually needed, and if they are not available, a read_approximate method to read back an approximate version of the data using standard python types and numpy types to be able to at least read something. The MarshallerCollection class as well as other classes are being changed to support this feature.
This is an API change, but it doesn't break compatibility so long as hdf5storage.Marshallers.TypeMarshaller or its descendants are still inherited from. Completely custom marshallers will no longer be compatible.
The text was updated successfully, but these errors were encountered:
Changed in commit 260ff78 . examples/example_hdf5storage_marshaller_plugin/example_hdf5storage_marshaller_plugin.py has a good example of how it looks in addition to being an example of the plugin architecture (see Issue #65)
Adding marshallers to support types in more modules of the main Python libraries as well as external packages would increase loading time due to loading more packages and add dependencies for the way marshallers work in the 0.1.x series.
Thus, the parent marshaller class,
hdf5storage.Marshallers.TypeMarshaller
is being changed to support lists of required modules that can be lazy loaded when the marshaller is actually needed, and if they are not available, aread_approximate
method to read back an approximate version of the data using standard python types and numpy types to be able to at least read something. TheMarshallerCollection
class as well as other classes are being changed to support this feature.This is an API change, but it doesn't break compatibility so long as
hdf5storage.Marshallers.TypeMarshaller
or its descendants are still inherited from. Completely custom marshallers will no longer be compatible.The text was updated successfully, but these errors were encountered: