alamb opened a new issue, #9671:
URL: https://github.com/apache/arrow-rs/issues/9671

   **Is your feature request related to a problem or challenge? Please describe 
what you are trying to do.**
   @adriangb  implemented a really neat bloom filter sizing optimization in 
   - https://github.com/apache/arrow-rs/pull/9628
   
   A problem with most Parquet bloom filter implementations is that to avoid 
wasting a lot of space, you need to know beforehand the Number of Distinct 
Values (NDV), which is challenging for most systems to configure correctly. 
This PR very nicely handles the case where the number of distinct values is not 
known beforehand -- the bloom filter undergoes a folding process.
   
   So if point lookups are important, systems can now enable bloom filters 
while writing Parquet and the writer will automatically resize the filters to 
the minimal size needed for a false positive percentage. This technique 
requires more CPU to compute and fold the bloom filters, but is very effective 
and results in byte-for-byte identical bloom filters. 
   
   The high level idea is mentioned in other sources
   * Sailhan & Stehr, ["Folding and Unfolding Bloom 
Filters"](https://hal.science/hal-01126174v1/document), IEEE iThings 2012.
   * Liang, ["Blocked Bloom Filters: Speeding Up Point Lookups in Tiger 
Postgres' Native 
Columnstore"](https://www.tigerdata.com/blog/blocked-bloom-filters-speeding-up-point-lookups-in-tiger-postgres-native-columnstore)
   
   However, it was not immediately apparently that those techniques do apply to 
parquet
   
   **Describe the solution you'd like**
   If would be great to write a blog post, ideally on one of
   * arrow.apache.org (https://github.com/apache/arrow-site) or  
   * parquet.apache.org (https://github.com/apache/parquet-site) 
   
   Explaining how the technique works
   
   The idea would be:
   1. Advertise that Parquet bloom filters are more useful now in arrow-rs
   2. Explain enough of the implementation that other Parquet implementations 
could implement it
   
   **Describe alternatives you've considered**
   <!--
   A clear and concise description of any alternative solutions or features 
you've considered.
   -->
   
   **Additional context**
   <!--
   Add any other context or screenshots about the feature request here.
   -->
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to