cellmap_flow.config.yaml
========================

.. py:module:: cellmap_flow.config.yaml

.. autoapi-nested-parse::

   Reading a ``cellmap_flow yaml`` or blockwise YAML file.

   ``load_config`` reads and checks the file's top level, filling
   ``charge_group`` and ``queue`` from the dashboard's saved settings (then the
   site's default queue) when the file leaves them out; the model entries under
   ``models`` are built by ``models.registry.build_models``. Every problem is a ``ConfigError``.
   ``resolve_data_path`` is the one rule for a model's ``data_path`` and
   ``scale``, used by every launcher. Every model's input is resampled to its
   voxel size when the data has no level at it (``cellmap_flow yaml`` passes
   ``--resample`` to each server; blockwise reads with
   ``on_voxel_size_mismatch="resample"``) unless the YAML says
   ``resample: false``.



Attributes
----------

.. autoapisummary::

   cellmap_flow.config.yaml.logger


Exceptions
----------

.. autoapisummary::

   cellmap_flow.config.yaml.ConfigError


Functions
---------

.. autoapisummary::

   cellmap_flow.config.yaml.resolve_data_path
   cellmap_flow.config.yaml.load_config


Module Contents
---------------

.. py:data:: logger

.. py:exception:: ConfigError



   The YAML configuration is not usable; the message says why.

   Raised rather than calling sys.exit: this code also runs inside the
   dashboard (the blockwise precheck, a finetuned model's base config),
   where SystemExit slips past ``except Exception`` and the request thread
   dies without sending a response. The CLIs catch it and exit non-zero.


.. py:function:: resolve_data_path(data_path: str, scale: Optional[str]) -> str

   The dataset a model reads: ``data_path``, with ``scale`` applied.

   The one rule every launcher uses (``cellmap_flow infer``, ``yaml`` and
   ``blockwise``), so the same YAML reads the same data in each:

   - ``data_path`` is an array: it is used as it is. A ``scale`` naming a
     different level is ignored, with a warning.
   - otherwise (a multiscale group, or a path that cannot be inspected
     here, such as a URL): ``scale`` selects the level under it.


.. py:function:: load_config(path: str) -> Dict[str, Any]

   Load and validate the YAML configuration.

   :param path: Path to YAML configuration file: a str, bytes or os.PathLike.
                Anything else is refused, because open() takes an int as a file
                descriptor: it would read through whatever the process has open
                under that number, and then close it.

   :returns: Validated configuration dictionary

   :raises ConfigError: ``path`` is not a path, or a required field is missing
       or malformed