Citrine

This release brings several formulations-related predictors to Beta. This means that they are well-documented, stable, and expected to work for a variety of problems. We are excited to have them used widely and welcome your feedback.

This release also contains several changes in service of cutting the eventual v1.0.0 release of citrine-python. Releasing v1.0.0 indicates that we expect the software to be polished, robust, and resistant to change. We will be making many upgrades in the next few weeks, including some breaking changes, while working towards this goal.

What's New
* The `MeanPropertyPredictor` now implements stricter behavior when `impute_properties` is `False`. If a formulation does not have a value for a given property, then it will not be featurized. (583).
* The `get` methods for `GemTable` and `TableConfig` no longer require that a version number be specified. If not specified, the most recent version is returned (594).

Improvements
* More flexibility when passing `set`, `list`, and `tuple` arguments. A method that requires a collection will now generally accept all three types. (585)
* Several not-yet-implemented and not-sensical REST methods have been given clear error messages (590).
* More thorough, correct, and uniform documentation (588, 595)
* Enumerated the allowed `resource_type` values, for use in access control calls (599).

Deprecated
* `get_with_version` is deprecated for `TableConfig`. Instead, use `get` and specify the version number (594).

Coming Soon

This release adds the ability to limit the number of design candidates that are evaluated when using the `MonteCarloProcessor`, which should make it easier to develop design workflows before scaling them up.

What's New
* The `MonteCarloProcessor` has a new `max_candidates` parameter that limits the number of design candidates that are evaluated (https://github.com/CitrineInformatics/citrine-python/pull/584)

This release includes new functionality around pulling ingredient labels into models and generating features from chemical formulas.

What's New
* Including ingredient labels in a table is now easier. With IngredientLabelsSetByProcessAndName and IngredientLabelsSetInOutput, users can now include all labels from an ingredient (using the ConcatColumn) in one cell rather than a series of Booleans to describe their presence of absence. (576, 580)

Improvements
* `build_mean_feature_property_predictors()`, a builder method that makes it easier for a graphical model calculate the mean of featurized properties, now accepts a `ChemicalFormulaFeaturizer` in addition to a `MolecularStructureFeaturizer` (578).

Deprecated
* We've removed support for using pipenv for running tests. Users should be conscientious about additionally installing the test_requirements.txt file when trying to test code. (579)

Coming Soon
* Even more ways of mapping your GEMD data into tables, feeding your materials design processes.
* Most formulations-related predictors promoted to Beta.

This is a minor bugfix release to correct an oversight in [v0.117.0](https://github.com/CitrineInformatics/citrine-python/releases/tag/v0.117.0).

Bugfixes
* Fixed support for deserialization of `ChemicalFormulaFeaturizer` (https://github.com/CitrineInformatics/citrine-python/pull/577)

This release includes significant new functionality around featurization of molecular structures, chemical formulas, and formulation ingredients.

What's new
* The `MolecularStructureFeaturizer` is out of alpha! (https://github.com/CitrineInformatics/citrine-python/pull/573)
* A new `ChemicalFormulaFeaturizer` that can be used in `GraphPredictors` (https://github.com/CitrineInformatics/citrine-python/pull/575)
* A new builder method to make it easier to use molecular structure and chemical formula properties as ingredient properties in formulations problems (https://github.com/CitrineInformatics/citrine-python/pull/575)

Improvements
* The `training_data_count` field is now available in the predictor report's `ModelSummary` (https://github.com/CitrineInformatics/citrine-python/pull/574)

This release includes significant new functionality around deleting GEMD objects, particularly Attribute Templates. It also contains considerable quality of life improvements, both for developers and for users.

What's new
* Support for deleting Attribute Templates (https://github.com/CitrineInformatics/citrine-python/pull/561)
* Support for deleting the entire contents of a dataset (https://github.com/CitrineInformatics/citrine-python/pull/567)

Improvements
* Default resolution for formulation design space reduced to `0.0001` (https://github.com/CitrineInformatics/citrine-python/pull/564)
* Made the GEMD batch delete async, increasing its flexibility (https://github.com/CitrineInformatics/citrine-python/pull/561)
* Predictor reports are now fetched only when needed, avoiding a rare race condition (https://github.com/CitrineInformatics/citrine-python/pull/571)
* Many small quality of life improvements for developers and users (https://github.com/CitrineInformatics/citrine-python/pull/568)

Fixes
* Response parsing in error cases is more flexible (https://github.com/CitrineInformatics/citrine-python/pull/565/files)