0.1.11

• filter boundaries should be inclusive #276
• keep track of strategy index #275

0.1.10

• no signatures for NULL values #259

0.1.9

• Bump supported Python versions #245

0.1.8

• use logging instead of the print function #138
• use poetry for dependecy management #139
• validate blocking schemas with pydantic #140
• move CI from Azure pipelines to GitHub actions #179
• rewrite of asses_blocks_2party for better scaling #189

0.1.7

• added Python 3.9 support to CI pipeline #116
• fixed division by zero bug in evaluation #107

Support column names in blocking schema

Before we only support column name indices in blocking schema, which is not very convenient to use especially when there are many column names. In this release, we improve the code to support column names in blocking schema.

Enhance blocking statistics and generalise blocking key

• Enable blocking statistics on individual strategy of P-Sig
• Generalise all blocking key to be string type

Enhancement on p-sig blocking

• Remove uncommon blocks due to collisions in bloom filter to reduce size of returned filtered reversed indices
• Add feedback on coverage information of P-Sig blocking method i.e. if final blocks cover 100% records or not

Trivial fix

• Update requirements.txt
• Enable Pypi to render markdown properly

First formal release of blocklib

• Support two blocking methods

• Probabilistic signature
• LSH based $\Lambda$-fold

• Complete documentation
• Tutorial notebook
• Tests