Blueetl

--------------

- Revert raising an error if the env variable SHMDIR isn't set. Log a more detailed warning instead.

--------------

Improvements
~~~~~~~~~~~~

- Allow to access the cache config and the output path of individual analyses with ``analysis_config.cache`` and ``analysis_config.output``, as a shortcut to ``analysis_config.cache.path``.
- Raise an error if the env variable ``SHMDIR`` isn't set, instead of logging a warning.

--------------

Bug Fixes
~~~~~~~~~

- Fix link to the example configuration in the documentation.

--------------

New Features
~~~~~~~~~~~~

- Add ``multi_index`` option to the features configuration, to decide whether ``reset_index()`` should be applied to the features DataFrames.
Note: the features cache will be rebuilt, although the resulting DataFrames are unchanged (because the default value of the new option is ``True``).

-------------

Improvements
~~~~~~~~~~~~

- Improve performance of report extraction and features calculation by writing partial DataFrames to the shared memory (or temp directory, if shared memory is not available).
Both the used memory and the execution time should be lower than before, when processing large DataFrames.
- Use zstd compression instead of snappy when writing parquet files.
- When ``repo`` is pickled, extract the DataFrames only if they aren't already stored in the cache.
- Remove fastparquet extra dependency.

-------------

New Features
~~~~~~~~~~~~

- Add ``cache.readonly``, to be able to use an existing cache without exclusive locking [NSETM-2310].
- Add ``cache.store_type``, to change the file format (experimental).
- Add ``cache.skip_features``, to skip writing the features DataFrames [NSETM-2312].

Deprecations
~~~~~~~~~~~~

- Deprecate ``output``, use ``cache.path`` instead.
- Deprecate ``clear_cache``, use ``cache.clear`` instead.