exor_filter changelog

Changelog History

• v0.6.0 Changes

  November 08, 2020

  • ➕ Adding support for incremental initialization api.
    • A more space efficient way to use the library.
    • Warning : currently does not support deduplication.
  • 🛠 Various bug fixes.
  • 🛠 Fixing test macros.

• v0.5.2 Changes

  October 24, 2020

  Fix nif compile on Erlang 23 by removing erl_interface from c_src Makefile.

• v0.5.1 Changes

  October 24, 2020

  👉 Use of contains with a binary created from to_bin.

• v0.5.0 Changes

  February 24, 2020

  ➕ Addition of serialization / deserialization of the xor filters to get them in pure binary. Thanks to @Vagabond !

• v0.4.2 Changes

  February 22, 2020

  🚀 Included in this release:

  Increased memory management, courtesy of @Vagabond

• v0.4.1 Changes

  January 20, 2020

  🚀 Included in this release:

  • ⬆️ Upgraded xor_filter headerfile. Includes protections against dups.
  • Included fix in xor_filter header file found by cppcheck.
    • PR for upstream found here: FastFilter/xor_singleheader#14
  • 👉 Used cppcheck on nif code, looks clean.
  • 🐎 Performance increases
    • Using a custom tail-recursive over_10_thousand/1 instead of length/1 function for greater performance on large lists.
    • Removing dup key checking from erlang code, now found in underlying library.
  • ➕ Address typo for freeing the wrong pointer
    • #6
  • 0️⃣ Using lists:map instead of lists:foldr, as it isn't tail recursive, for the default hash.
    • #7
  • ✂ Removed raw interface, no longer has any speed benefits.

• v0.3.0 Changes

  January 06, 2020

  🚀 This release is the 'production readiness' release.

  Includes:

  1. Error when duplicate keys are present in the hash function and pre-hashed options, there is no longer the risk of an infinite loop during initialization!
     • The default_hash squashes dupes, so that there is no risk of this happening here either.
  2. A 'raw' interface to avoid this, use with caution.
  3. Convenience modules xor8 and xor16. 🛠 4. Dialyzer fixes ✅ 5. Moar tests 📚 6. Documentation updates.
  4. Simplified C and erlang code
  5. Removal of fast_hash option, as duplication checking would have voided the usefulness.

• v0.0.1 Changes

  December 26, 2019

  v0.0.1 CHANGELOG

  • Two hashing methods are provided
    • default_hash, which uses the erlang:phash2/1 function
    • fast_hash, which is faster but inconsistent across nodes.
  • A custom hash function may be provided
  • Pre-hashed data may be passed as well.
  • 🎉 Initialization using a dirty nif is used if the passed list is greater than 10K elements. So it is 'safe' to load large amounts of data.
  • It takes 2 seconds to initialize a filter of 10M elements using the exor_filter:xor8_buffered/2 function with pre-hashed data.
    • Our example used keys that were generated sequentially 😛