DataFile behaves differently inside with statement

[johnomotani]

I noticed while looking at #1241 that DataFile has unexpected (to me) behaviour inside a with statement.

BOUT-dev/tools/pylib/boututils/datafile.py

Lines 166 to 167 in 5dcf404

	def __enter__(self):
	return self.impl.__enter__()

because __enter__ returns self.impl.__enter__(), in a statement like

with DataFile('BOUT.dmp.0.nc') as f:
    # do stuff here

f is a DataFile_netCDF object, not a DataFile object. In practice this doesn't make much difference because the interfaces are very similar, but DataFile.read does some work

BOUT-dev/tools/pylib/boututils/datafile.py

Lines 197 to 201 in 5dcf404

	if ranges is not None:
	for x in ranges:
	if isinstance(x, (list, tuple)):
	x = slice(*x)
	return self.impl.read(name, ranges=ranges, asBoutArray=asBoutArray)

(although I think both netCDF and hdf5 can handle converting lists and tuples to slices) and bout_type(self, varname) which returns self.attributes(varname)["bout_type"] would be missing.

In summary, I'd expect DataFile.__enter__ to return self so that f in the with-statement is a DataFile object not one of the implementations. Is this a bug?

[ZedThree]

I think with our current implementation, DataFile.__enter__ should call self.impl.__enter__() but return self. Any public methods that the implementations define should also be defined in DataFile, which can just call self.impl.foo()

[dschwoerer]

Fixed by #1243