checking inputs to step

1. check that output from transformer is dict
2. check that input to step.fit_transform() and step.transform() is dict (None is Ok though).

chained operations on Step

Dev 015 (#116)

* allow to chain operations i.e. preprocessing_block = preprocessing_steps(cfg=cfg).set_mode_inference().set_parameters_upstream(steps_config)

* allow to chained operations

bug fix

bug was reported here: minerva-ml/steppy-examples#9

bug fixing

bug fix (#113)

* bug fix

* prepare for new version

removed ALL_STEPS_NAMES global, simplified logs

Dev s12 (#112)

* removed suffixes

* removed global _ALL_STEPS_NAMES, check upstream names by default - if two names are the same raise an error, simplified error logs

bug fix with transformers path

Revert "removed suffixes (#110)" (#111)

This reverts commit 520a17ad9abb49506774dcd29435d2cc14931847.

v0.1.10: S10 (#107)

* persist is with joblib by default

* Update setup.py

* Update conf.py

bug fixing release

Dev 0.1.9 (#106)

* corrected display bug

* made all steps list private

* prepare for 0.1.9

new defaults and simpler logic

v0.1.8

prepare for v0.1.8

new interface

removed cache dir, cache is now stored in memory as self.output
name is not obligatory. By default use transformer.__class__.__name__
name by default is added suffix '_0' to the name (or 1,2,3,...)
now you can clean_cache or clean_cache_upstream_steps
added name validation
new defaults:
    is_fittable=True
    force_fitting=True
    cache_output=False
    persist_output=True