pyiron_base.storage.hdfstub module

Convenience class to lazily read values from HDF.

class pyiron_base.storage.hdfstub.HDFStub(hdf, group_name)

Bases: object

Provides lazy loading of data from HDF.

Instead of accessing an HDF group directly

>>> hdf[group_name]
...

you can wrap this with this class

>>> stub = HDFStub(hdf, group_name)

and then later perform this lookup with :method:`.load`

>>> stub.load() == hdf[group_name]
True

For simple datatypes there’s not a big advantages to this, but DataContainer uses this to load its contents lazily and ensure that nested containers are also lazily loaded. This is done by customizing what happend on :method:`.load` via :method:`.register`. This class method adds a callback to the class that will be called when the specified type name is found in the hdf group that is to be loaded.

>>> hdf['mytype/NAME']
MyType
>>> hdf['mytype/TYPE']
<class 'my.module.MyType'>
>>> HDFStub.register(MyType, lambda hdf, group: print(42) or hdf[group].to_object())
>>> my = HDFStub(hdf, 'mytype').load()
42
>>> my
MyType(...)

This is intended to allow classes that want to be lazily loaded in a certain way to customize what arguments they pass to_object() (and therefore to their own initializers).

load()

Read value from HDF.

If group_name is a node in HDF, simply its value will be returned. If it is a group in HDF and the ‘NAME’ node matches any of the types registered with :method:`.register`, it will be loaded with the provided callback. Otherwise it will be loaded with :method:`.ProjectHDFio.to_object()`.

classmethod register(type, load)

Register call back for a new type.

Parameters:

• type (type) – class to be registered

• load (function) – callback that is called on :method:`.load` when the type matches type_name, must accept hdf and group_name corresponding to the init parameters of this class and return (lazily) loaded object

pyiron_base.storage.hdfstub.to_object(hdf_group)