how to access the layout of a dataset?

[Remi-Gau]

I have a bids apps that requires to get some "stats" about the dataset as a whole (on top of accessing some specific files).

Is there a way that snakebids "exposes" the layout so I can access it in script or should I just create a script that will reindex the dataset with pybids?

[pvandyken]

Not yet, but this has come up before and should be low-hanging fruit. We can definitely prioritize this.

I think a reasonable API would be to just have a BidsDataset.layout property that points to the underlying layout object (trying to think of this will have any incompatibilities with future filtering apis, but I can't come up with anything).

[Remi-Gau]

Thanks. For now I can still reindex the dataset but it will become a pain for large datasets.

[akhanf]

Note that another workaround right now would be to use the --pybidsdb-dir option in snakebids to avoid re-indexing, and always use the same path to the pybidsdb directory (e.g. <bids_dir>/.pybids), so that your script can also be hardcoded to look in the same location (e.g. <bids_dir>/.pybids).

We also have this useful tool for generating the pybids db dir on a network file system (https://github.com/pvandyken/pybidsdb)

[pvandyken]

I'm actually facing a bit of an implementation detail on this, having to do with the types and the fact that, in testing, the fake Datasets I'm making will not have an associated layout.

There's two approaches to this: one is an Optional type annotation:

class BidsDataset:
    layout: Optional[BIDSLayout] = None

I don't care for this because in practice, any BidsDataset generated by generate_inputs() will indeed have a layout, so from the user perspective, layout is not really optional.

The other is to have two classes

class AbstractBidsDataset:
   ...

class BidsDataset(AbstractBidsDataset):
    layout: BIDSLayout

I'm planning on going with this approach: the implementation is invisible to the user and it solves the problem in the first approach. AbstractBidsDataset would be used in tests.