0.8.6

• Fixed an issue in approx_percentile caused by precision loss in float number
• Support for EMR 7.0 (This is not currently compatible with Kinesis-based Stream Feature Views -- support will be added soon)
• For materialization of Batch Feature Views on EMR: when spark.sql.shuffle.partitions or spark.default.parallelism is explicitly set (via the Cluster Config), Tecton will no longer repartition or coalesce during offline writes