[xquery-talk] [ANN] Rumble 1.9.0
gfourny at inf.ethz.ch
Wed Oct 28 09:10:26 PDT 2020
We are happy to announce the release of Rumble 1.9.0 "Ficus Bonsai".
It is a free, open source JSONiq engine (JSON-tuned cousin of XQuery), and can be downloaded as a jar from
or you can directly try out a JSONiq tutorial running on our public backend in your browser (it is a Jupyter notebook running on Google's Colab) from
(this public backend is just one small machine for small queries -- if you need a larger Spark cluster, you can create one on your favorite cloud provider, e.g., Amazon EMR, in a few minutes, connect to it, download the Rumble jar and start querying)
New features (since the previous announcement of version 1.7.0 in this list) include:
- Detection of joins in many forms: two for clauses binding large sequences with a predicate in the second one (with support for allowing empty for left outer joins), two for clauses binding large sequences followed by a where clause, or a let clause binding a large sequence with a predicate, for "group joins" (to denormalize data easily).
- Efficient evaluation of equi-joins (pushed down to Spark) if the predicate involves equality comparison of (arbitrary) expressions from each side (or conjunctions thereof).
- Support for positional variables.
- Outputing large quantities of structured data to other formats: avro, parquet, csv, etc. This allows using Rumble to easily clean up a messy dataset with JSONiq, structure it, annotate it, and output it in a structured format, or to easily convert from one format to another, and use it in turn as input to another JSONiq query, or as input to a Python script for visualization with your favorite libraries (pie charts, etc).
- Various performance improvements and bugfixes.
More information about the talk