Adding a non-protobuf allocate_ids method on Dataset.

[dhermes]

Partially address #336.

[dhermes]

I'm unclear if allocate_ids should return just the IDs or the keys (which is what it currently does).

[coveralls]

Coverage remained the same when pulling 0194f30 on dhermes:improve-allocate-ids into c2ba130 on GoogleCloudPlatform:master.

[tseaver]

I think keys makes more sense than IDs: the keys should match the incomplete one, for instance (otherwise, why are we passing it?)

[silvolu]

Yup, keys.

[coveralls]

Coverage remained the same when pulling aa63666 on dhermes:improve-allocate-ids into 1f2ed88 on GoogleCloudPlatform:master.

[dhermes]

Cool, so I can leave as-is with keys.

[tseaver]

@dhermes doesn't the test still assert that the result is the integer IDs, rather than keys? This bit:

self.assertEqual(result, range(NUM_IDS))

should be something like:

self.assertEqual([x.id for x in result], range(NUM_IDS))

[dhermes]

The place it comes in to play is the last line of Dataset.allocate_ids:

        return [incomplete_key.id(allocated_id)
                for allocated_id in allocated_ids]

so since my stubbed _Key.id returns the ID instead of self, the result of Dataset.allocate_ids is a list of IDs instead of a list of _Keys. I can change test_entity._Key.id to

    def id(self, id_to_set):
        self._called_id = id_to_set
        return self

to resolve if you think that's better.

[tseaver]

I think it would be better to emphasize, even in tests, that the result is a sequence of key instances, especially since the name, allocate_ids, might indicate otherwise.

[dhermes]

@tseaver Rebased and returned _Keys. PTAL.

[coveralls]

Changes Unknown when pulling 2ac3226 on dhermes:improve-allocate-ids into * on GoogleCloudPlatform:master*.

[coveralls]

Coverage remained the same when pulling ee29cbe on dhermes:improve-allocate-ids into 8a6feff on GoogleCloudPlatform:master.

[tseaver]

LGTM