test_lazy_exports
=================

.. py:module:: test_lazy_exports

.. autoapi-nested-parse::

   Importing this package must stay cheap, and its exports must stay honest.

   Every model in this stack is a ccflow model, and importing ccflow imports
   csp and pandas: about 0.9 s. A package that imports its model modules
   at import time puts all of that in front of anything that merely imports
   it -- including hydra's searchpath plugin, which imports the config package
   to discover what is installed and reads none of these names.



Attributes
----------

.. autoapisummary::

   test_lazy_exports.INIT


Functions
---------

.. autoapisummary::

   test_lazy_exports._type_checking_imports
   test_lazy_exports.test_the_declaration_and_the_type_checking_block_agree
   test_lazy_exports.test_every_exported_name_resolves
   test_lazy_exports.test_an_unknown_name_still_raises_attribute_error
   test_lazy_exports.test_importing_the_package_stays_cheap


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

.. py:data:: INIT

.. py:function:: _type_checking_imports() -> dict[str, tuple[str, ...]]

   What the ``TYPE_CHECKING`` block declares, read from the source.


.. py:function:: test_the_declaration_and_the_type_checking_block_agree() -> None

   The two exist for different readers -- the resolver and a type
   checker -- and a name in one and not the other is a name that imports
   for one of them and fails for the other.


.. py:function:: test_every_exported_name_resolves() -> None

   A name declared here and missing from its submodule would raise on
   first access, which is a broken import nobody sees until a caller
   reaches for it. Resolve all of them now.


.. py:function:: test_an_unknown_name_still_raises_attribute_error() -> None

.. py:function:: test_importing_the_package_stays_cheap() -> None

   Run in a fresh interpreter, because this one has already imported
   everything.