recordlinkage

• Remove deprecated recordlinkage classes (#173)
• Bump min Python version to 3.6, ideally 3.8+ (#171)
• Bump min pandas version to >=1
• Resolve deprecation warnings for numpy and pandas
• Happy lint, sort imports, format code with yapf
• Remove unnecessary np.sort in SNI algorithm (#141)
• Fix bug for cosine and qgram string comparisons with threshold (#135)
• Fix several typos in docs (#151)(#152)(#153)(#154)(#163)(#164)
• Fix random indexer (#158)
• Fix various deprecation warnings and broken docs build (#170)
• Fix broken docs build due to pandas depr warnings (#169)
• Fix broken build and removed warning messages (#168)
• Update narrative
• Replace Travis by Github Actions (#132)
• Fix broken test NotFittedError
• Fix bug in low memory random sampling and add more tests (#130)
• Add extras_require to setup.py for deps management
• Add banner to README and update title
• Add Binder and Colab buttons at tutorials (#174)

Special thanks to Tomasz Waleń @twalen and other contributors for their work on this release.

• Drop Python 2.7 and Python 3.4 support. (#91)
• Upgrade minimal pandas version to 0.23.
• Simplify the use of all cpus in parallel mode. (#102)
• Store large example datasets in user home folder or use environment variable. Before, example datasets were stored in the package. (see issue #42) (#92)
• Add support to write and read annotation files for recordlinkage ANNOTATOR. See the docs and https://github.com/J535D165/recordlinkage-annotator for more information.
• Replace .labels by .codes for pandas.MultiIndex objects for newer versions of pandas (>0.24). (#103)
• Fix totals for pandas.MultiIndex input on confusion matrix and accuracy metrics. (see issue #84) (#109)
• Initialize Compare with (a list of) features (Bug). (#124)
• Various updates in relation to deprecation warnings in third-party libraries such as sklearn, pandas and networkx.

Fix distribution problem.

resolve conflict with threshold and missing value (#85)

Closes #70

• Minor installation improvement. Exclude unwanted files

• Fix installation issue. Submodule 'preprocessing' was not added to the
  source distribution.

• The submodule 'standardise' is renamed. The new name is 'preprocessing'.
  The submodule 'standardise' will get deprecated in a next version.
• Deprecation errors were not visible for many users. In this version, the
  errors are better visible.
• Improved and new logs for indexing, comparing and classification.
• Faster comparing of string variables. Thanks Joel Becker.
• Changes make it possible to pickle Compare and Index objects. This makes it
  easier to run code in parallel. Tests were added to ensure that pickling
  remains possible.
• Important change. MultiIndex objects with many record pairs were split into
  pieces to lower memory usage. In this version, this automatic splitting is
  removed. Please split the data yourself.
• Integer indexing. Blog post will follow on this.
• The metrics submodule has changed heavily. This will break with the previous
  version.
• repr() and str() will return informative information for index and compare
  objects.
• It is possible to use abbreviations for string similarity methods. For example
  'jw' for the Jaro-Winkler method.
• The FEBRL dataset loaders can now return the true links as a
  pandas.MultIndex for each FEBRL dataset. This option is disabled by default.
  See the FEBRL datasets for details.
• Fix issue with automatic recognision of license on Github.
• Various small improvements.

Note: In the next release, the Pairs class will get removed. Migrate now.

• print statement in the geo compare algorithm removed.
• String, numeric and geo compare functions now raise directly when an incorrect algorithm name is passed.
• Fix unit test that failed on Python 2.7.

• A new compare API. The new Compare class no longer takes the datasets and pairs as arguments. The actual computation is now performed when calling .compute(PAIRS, DF1, DF2). The documentation is updated as well, but still needs improvement.
• Two new string similarity measures are added: Smith Waterman (smith_waterman) and Longest Common Substring (lcs). Thanks to Joel Becker and Jillian Anderson from the Networks Lab of the University of Waterloo.
• Added and/or updated a large amount of unit tests.
• Various small improvements.

• A new index API. The new index API is no longer a single class (recordlinkage.Pairs(...)) with all the functionality in it. The new API is based on Tensorflow and FEBRL. With the new structure, it easier to parallise the record linkage process. In future releases, this will be implemented natively. See the reference page for more information and migrating. <http://recordlinkage.readthedocs.io/en/latest/ref-index.html>_
• Significant speed improvement of the Sorted Neighbourhood Indexing algorithm. Thanks to @perryvais (PR #32).
• The function binary_comparisons is renamed. The new name of the function is binary_vectors. Documentation added to RTD.
• Added unit tests to test the generation of random comparison vectors.
• Logging module added to separate module logs from user logs. The implementation is based on Tensorflow.

• Issues solved with rendering docs on ReadTheDocs. Still not clear what is going on with the autodoc_mock_imports in the sphinx conf.py file. Maybe a bug in sphinx.
• Move six to dependencies.
• The reference part of the docs is split into separate subsections. This makes the reference better readable.
• The landing page of the docs is slightly changed.

• Add additional arguments to the function that downloads and loads the krebsregister data. The argument missing_values is used to fill missing values. Default: nothing is done. The argument shuffle is used to shuffle the records. Default is True.
• Remove the lastest traces of the old package name. The new package name is 'Python Record Linkage Toolkit'
• Better error messages when there are only matches or non-matches are passed to train the classifier.
• Add AirSpeedVelocity tests to test the performance.
• Compare for deduplication fixed. It was broken.
• Parameterized tests for the Compare class and its algorithms. Making use of nose-parameterized module.
• Update documentation about contributing.
• Bugfix/improvement when blocking on multiple columns with missing values.
• Fix bug #29. Package not working with pandas 0.18 and 0.17. Dropped support pandas 0.17 and fixed support for 0.18. Also added multi-dendency tests for TravisCI.
• Support for dedicated deduplication algorithms
• Special algorithm for full index in case of finding duplicates. Performce is 100x better.
• Function max_number_of_pairs to get the maximum number of pairs.
• low_memory for compare class.
• Improved performance in case of comparing a large number of record pairs.
• New documentation about custom algorithms
• New documentation about the use of classifiers.
• Possible to compare arrays and series directly without using labels.
• Make a dataframe with random comparison vectors with the binary_comparisons in the recordlinkage.datasets.random module.
• Set KMeans cluster centers by hand.
• Various documentation updates and improvements.
• Jellyfish is now a required dependency. Fixes bug #30.
• Added tox.ini to test packaging and installation of package.
• Drop requirements.txt file.
• Many small fixes and changes. Most of the changes cover the Compare module. Especially label handling is improved.

This version includes the following updates:

• Reformatting the code such that it follows PEP8.
• Add Travis-CI and codecov support.
• Switch to distributing wheels.
• Fix bugs with depreciated pandas functions. __sub__ is no longer used for computing the difference of Index objects. It is now replaced by ``INDEX.difference(OTHER_INDEX).
• Exclude pairs with NaN's on the index-key in Q-gram indexing.
• Add tests for krebsregister dataset.
• Fix Python3 bug on krebsregister dataset.
• Improve unicode handling in phonetic encoding functions.
• Strip accents with the clean function.
• Add documentation
• Bug for random indexing with incorrect arguments fixed and tests added.
• Improved deployment workflow
• And much more

• Fixes a serious bug with deduplication (thanks to https://github.com/dserban).
• Fixes undesired behaviour for sorted neighbourhood indexing with missing values.
• Add new datasets to the package like Febrl datasets
• Move Krebsregister dataset to this package.
• Improve and add some tests
• Various documentation updates

This version contains a lot of changes to the API. Hopefully, there are no large API changes needed for now.

• Total restructure of compare functions (The end of changing the API is close to now.)
• Compare method numerical is now named numeric and fuzzy is now named string.
• Add haversine formula to compare geographical records.
• Use numexpr for computing numeric comparisons.
• Add step, linear and squared comparing.
• Add eye index method.
• Improve, update and add new tests.
• Remove iterative indexing functions.
• New add chunks for indexing functions. These chunks are defined in the class Pairs. If chunks are defined, then the indexing functions returns a generator with an Index for each element.
• Update documentation.
• Various bug fixes.

• Full Python3 support
• Update the parameters of the Logistic Regression Classifier manually. In literature, this is often denoted as the deterministic record linkage.
• Expectation/Conditional Maxisation algorithm completely rewritten. The performance of the algorithm is much better now. The algorithm is still experimental.
• New string comparison metrics: Q-gram string comparing and Cosine string comparing.
• New indexing algorithm: Q-gram indexing.
• Several internal tests.
• Updated documenation.
• BernoulliNBClassifier is now named NaiveBayesClassifier. No changes to the algorithm.
• Arguments order in compare functions corrected.
• Function to clean phone numbers
• Return the result of the classifier as index, numpy array or pandas series.
• Many bug fixes