No, don't use __getattribute__ b/c it leads to infinite recursion. New idea: this is not a class that should be subclassing at all.

When #620 removed the inherited fsspec.spec.AbstractBufferedFile class from earthaccess.EarthAccessFile, it broke the ability of xarray.backends.list_engines()["h5netcdf"].guess_can_open to recognize these pointers. It comes back to MRO, and the fact that EarthAccessFile is no longer an instance of io.IOBase.

It would be very helpful if we had an integration test for the behavior that the EarthAccessFile class enables.

A change to repair damage done in #620 is to reintroduce a base class on EarthAccessFile (and I think io.IOBase might work), but it doesn't fix this issue.

Merged #832 in order to make way for a release. Note that the test surfacing this bug has been marked as an expected fail.

I would like to clarify a couple of things.

First of all, comparing two bound methods should be done with "==", not with "is" to check for method equivalence:

class Dummy:
    def read(self): pass

d = Dummy()
print(d.read is d.read)  # → False
print(d.read == d.read)  # → True

Now running

auth = earthaccess.login()
results = earthaccess.search_data(short_name="VIIRSJ1_L2_OC", count=1)
files = earthaccess.open(results)
with files[0] as g:
    print(f"Type of g: {type(g)}")
    print(f"Has g.read? {hasattr(g, 'read')}")
    print(f"g.read: {g.read}")
    print(f"Has g.f? {hasattr(g, 'f')}")
    if hasattr(g, 'f'):
        print(f"Type of g.f: {type(g.f)}")
        print(f"Has g.f.read? {hasattr(g.f, 'read')}")
        if hasattr(g.f, 'read'):
            print(f"g.f.read: {getattr(g.f, 'read')}")
            print(f"g.read is g.f.read: {g.read == getattr(g.f, 'read')}")
    else:
        print("g.f does not exist")
    assert g.read == getattr(g.f, "read")

gives the following output:

QUEUEING TASKS | : 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 1534.13it/s]
PROCESSING TASKS | : 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 2.14it/s]
COLLECTING RESULTS | : 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 7025.63it/s]
Type of g: <class 'earthaccess.store.EarthAccessFile'>
Has g.read? True
g.read: <bound method AbstractBufferedFile.read of <File-like object HTTPFileSystem, https://oceandata.sci.gsfc.nasa.gov/cmr/getfile/JPSS1_VIIRS.20171129T213001.L2.OC.nc>>
Has g.f? True
Type of g.f: <class 'fsspec.implementations.http.HTTPStreamFile'>
Has g.f.read? True
g.f.read: <bound method HTTPStreamFile._read of <File-like object HTTPFileSystem, https://oceandata.sci.gsfc.nasa.gov/cmr/getfile/JPSS1_VIIRS.20171129T213001.L2.OC.nc>>
g.read is g.f.read: False
Traceback (most recent call last):
File "/home/obpg/scripts/test_earthdata.py", line 26, in
assert g.read == getattr(g.f, "read")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError

Modifying the class as follows:

class EarthAccessFile(fsspec.spec.AbstractBufferedFile):
    """Handle for a file-like object pointing to an on-prem or Earthdata Cloud granule."""

    def __init__(
        self, f: fsspec.spec.AbstractBufferedFile, granule: DataGranule
    ) -> None:
        """EarthAccessFile connects an Earthdata search result with an open file-like object.

        No methods exist on the class, which passes all attribute and method calls
        directly to the file-like object given during initialization. An instance of
        this class can be treated like that file-like object itself.

        Parameters:
            f: a file-like object
            granule: a granule search result
        """
        import types
        import inspect
        self.f = f
        self.granule = granule
        # Automatically copy all methods from self.f
        for name, member in inspect.getmembers(f):
            if inspect.ismethod(member) or inspect.isfunction(member):
                setattr(self, name, types.MethodType(member.__func__, f))

    def __getattr__(self, method: str) -> Any:
        return getattr(self.f, method)

    def __reduce__(self) -> Any:
        return make_instance, (
            type(self.f),
            self.granule,
            earthaccess.__auth__,
            dumps(self.f),
        )

    def __repr__(self) -> str:
        return repr(self.f)

fixes the issue:

QUEUEING TASKS | : 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 1556.33it/s]
PROCESSING TASKS | : 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 2.55it/s]
COLLECTING RESULTS | : 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 5468.45it/s]
Type of g: <class 'earthaccess.store.EarthAccessFile'>
Has g.read? True
g.read: <bound method HTTPStreamFile._read of <File-like object HTTPFileSystem, https://oceandata.sci.gsfc.nasa.gov/cmr/getfile/JPSS1_VIIRS.20171129T213001.L2.OC.nc>>
Has g.f? True
Type of g.f: <class 'fsspec.implementations.http.HTTPStreamFile'>
Has g.f.read? True
g.f.read: <bound method HTTPStreamFile._read of <File-like object HTTPFileSystem, https://oceandata.sci.gsfc.nasa.gov/cmr/getfile/JPSS1_VIIRS.20171129T213001.L2.OC.nc>>
g.read is g.f.read: True

If we want to resolve the issue only for the read method, then

        # Automatically copy all methods from self.f
        for name, member in inspect.getmembers(f):
            if inspect.ismethod(member) or inspect.isfunction(member):
                setattr(self, name, types.MethodType(member.__func__, f))

should be replaced with

self.read = f.read

Important to note:
https://docs.python.org/3/reference/datamodel.html#object.__getattr__:

Called when the default attribute access fails with an AttributeError (either getattribute() raises an AttributeError because name is not an instance attribute or an attribute in the class tree for self; or get() of a name property raises AttributeError). This method should either return the (computed) attribute value or raise an AttributeError exception. The object class itself does not provide this method.

Note that if the attribute is found through the normal mechanism, getattr() is not called. (This is an intentional asymmetry between getattr() and setattr().) This is done both for efficiency reasons and because otherwise getattr() would have no way to access other attributes of the instance. Note that at least for instance variables, you can take total control by not inserting any values in the instance attribute dictionary (but instead inserting them in another object). See the getattribute() method below for a way to actually get total control over attribute access.

This means that :

    def __getattr__(self, method: str) -> Any:
        return getattr(self.f, method)

doesn't override existing methods like read.