googleapis / googleapis/google-cloud-java
[java-bigtable] Add the ability to get Row cells in serverside order
- Dominant language
- Java
- Stars
- 2.1k
- Forks
- 1.2k
- Avg merge
- 1d 23h
- Merged PRs (30d)
- 154
Description
Bigtable stores data in the following order:
1. lexicographically by row key
2. family creation order
3. lexicographically by qualifier
4. reverse chronological by timestamp
The family ordering is not guaranteed so the data client currently pre-processes each row to change the ordering of families to be lexicographic order. This makes it convenient to do binary searches for specific values.
Unfortunately, this has negative repercussions when combined with `cells_per_row_limit_filter` and `cells_per_row_offset_filter`. Those filters are executed on the server side using the family creation order.
At the very least we need to update our docs to caution users against using those filters outside of a chain that narrows it to a single family.
In addition we should expose the server ordering of the cells on each row as `Row#getRawCells()` (maybe with a better name).
Contributor guide
Assessment
This issue has not been assessed yet.