if the dataset is a temporary dataset.
Here is a summary of temporary datasets:
1. Different from a persistent dataset, reads and writes over a temporary dataset do not require any lock.
Writes over a temporary dataset do not generate any write-ahead update and commit log but generate
flush log and job commit log.
2. A temporary dataset can only be an internal dataset, stored in partitioned LSM-Btrees.
3. All secondary indexes for persistent datasets are supported for temporary datasets.
4. A temporary dataset will be automatically garbage collected if it is not active in the past 30 days.
A temporary dataset could be used for the following scenarios:
1. A data scientist wants to run some one-time data analysis queries over a dataset that s/he pre-processed
and the dataset is only used by her/himself in an one-query-at-a-time manner.
2. Articulate AQL with external systems such as Pregelix/IMRU/Spark. A user can first run an AQL
query to populate a temporary dataset, then kick off an external runtime to read this dataset,
dump the results of the external runtime to yet-another-temporary dataset, and finally run yet-another AQL
over the second temporary dataset.