xarray accessor and helpers

Importing atlas registers an accessor at xr.Dataset.atlas (the side effect happens on the very first import atlas, so there's no extra setup). The accessor is a thin convenience wrapper around the Atlas.add_xr_dataset method.

import atlas, xarray as xr        # ds.atlas is registered here

ds = xr.Dataset(...)
with atlas.Atlas.create("/tmp/store") as atlas:
    ds.atlas.write(atlas, "jan_2024")           # accessor form
    atlas.add_xr_dataset(ds, "feb_2024")        # equivalent Atlas method

See xarray integration for the full storage conventions and dtype mapping.

ds.atlas.write

Append this Dataset to the open atlas store under name.

Equivalent to atlas.add_xr_dataset(self_ds, name, chunks).

atlas.init_tracing

Top-level helper for enabling the Rust core's structured logging. Useful when debugging a slow read path or verifying which chunks a lazy compute actually touched.

atlas.init_tracing("debug")           # everything at debug+
atlas.init_tracing("atlas_python=info")    # just atlas crate at info+
atlas.init_tracing()                  # re-read ATLAS_LOG / RUST_LOG env vars

Python-callable variant: atlas.init_tracing("debug") forces a filter directive regardless of env vars. Passing None re-reads env vars.