Our Journey with Apache Arrow (Part 2): Adaptive Schemas and Sorting
Hacker News
JULY 4, 2023
Likewise, a column with dictionary encoding that indexes a uint64 will occupy four times more memory than the same column with a dictionary encoding based on a uint8. To optimize such scenarios, we have adopted an intermediary approach that we have named dynamic Arrow schema, aiming to gradually adapt the schema based on the observed data.
Let's personalize your content