Roadmap#

Goals#

  1. Enable easy use of additional CF attributes that are not decoded by xarray.

  2. Provide a consolidated set of public helper functions that other packages can use to avoid duplicated efforts in parsing CF attributes.

Scope#

  1. This package will not provide a full implementation of the CF data model using xarray objects. This use case should be served by Iris.

  2. Unit support is left to pint-xarray and future xarray support for pint until it is clear that there is a need for some functionality.

  3. Projections and CRS stuff is left to rioxarray and other geo-xarray packages. Some helper functions could be folded in to cf-xarray to encourage consolidation in that sub-ecosystem.

Design principles#

  1. Be uncomplicated.

    Avoid anything that requires saving state in accessor objects (for now).

  2. Be friendly.

    Users should be allowed to mix CF names and variables names from the parent xarray object e.g. ds.cf.plot(x="X", y="model_depth"). This allows use with “imperfectly tagged” datasets.

  3. Be loud when wrapping to avoid confusion.

    For e.g. the repr for cf.groupby("X") should make it clear that this is a CF-wrapped Resample instance i.e. cf.groupby("X").mean("T") is allowed. Docstrings should also clearly indicate wrapping by cf-xarray; for e.g. ds.cf.isel.

  4. Allow easy debugging and help build understanding.

    Since cf_xarray depends on attrs being present and since attrs are easily lost in xarray operations, we should allow easy diagnosis of what cf_xarray can decode for a particular object.