Abstract: A query directed at a source table organized into a set of batch units is received. The query comprises a regular expression search pattern. The regular expression search pattern is converted to a pruning index predicate comprising a set of substring literals extracted from the regular expression search pattern. A set of N-grams is generated based on the set of substring literals extracted from the regular expression search pattern. A pruning index associated with the source table is accessed. The pruning index indexes distinct N-grams in each column of the source table. A subset of batch units to scan for data matching the query are identified based on the pruning index and the set of N-grams. The query is processed by scanning the subset of batch units.

Abstract: Techniques for implementing cross in-database machine learning are disclosed. In some example embodiments, a computer-implemented method comprises training a machine learning model in a first database instance using a machine learning algorithm and a training dataset in response to receiving a request to train, serializing the trained machine learning model into a binary file in response to the training of the machine learning model, recreating the trained machine learning model in a second database instance using the binary file in response to receiving a request to apply the machine learning model, and generating an inference result by applying the recreated trained machine learning model on the inference dataset in the second database instance.

Abstract: In some example embodiments, a computer-implemented method may include training a machine learning model in a first database instance using a machine learning algorithm and a training dataset in response to receiving a request to train, serializing the trained machine learning model into a binary file in response to the training of the machine learning model, recreating the trained machine learning model in a second database instance using the binary file in response to receiving a request to apply the machine learning model, and generating an inference result by applying the recreated trained machine learning model on the inference dataset in the second database instance.

Abstract: A query directed at a source table organized into a set of batch units is received. The query comprises a regular expression search pattern. The regular expression search pattern is converted to a pruning index predicate comprising a set of substring literals extracted from the regular expression search pattern. A set of N-grams is generated based on the set of substring literals extracted from the regular expression search pattern. A pruning index associated with the source table is accessed. The pruning index indexes distinct N-grams in each column of the source table. A subset of batch units to scan for data matching the query are identified based on the pruning index and the set of N-grams. The query is processed by scanning the subset of batch units.

Abstract: A query directed at a source table organized into a set of batch units is received. The query comprises a regular expression search pattern. The regular expression search pattern is converted to a pruning index predicate comprising a set of substring literals extracted from the regular expression search pattern. A set of N-grams is generated based on the set of substring literals extracted from the regular expression search pattern. A pruning index associated with the source table is accessed. The pruning index indexes distinct N-grams in each column of the source table. A subset of batch units to scan for data matching the query are identified based on the pruning index and the set of N-grams. The query is processed by scanning the subset of batch units.

Abstract: Techniques for implementing cross in-database machine learning are disclosed. In some example embodiments, a computer-implemented method comprises training a machine learning model in a first database instance using a machine learning algorithm and a training dataset in response to receiving a request to train, serializing the trained machine learning model into a binary file in response to the training of the machine learning model, recreating the trained machine learning model in a second database instance using the binary file in response to receiving a request to apply the machine learning model, and generating an inference result by applying the recreated trained machine learning model on the inference dataset in the second database instance.

Abstract: In some example embodiments, a computer-implemented method may include training a machine learning model in a first database instance using a machine learning algorithm and a training dataset in response to receiving a request to train, serializing the trained machine learning model into a binary file in response to the training of the machine learning model, recreating the trained machine learning model in a second database instance using the binary file in response to receiving a request to apply the machine learning model, and generating an inference result by applying the recreated trained machine learning model on the inference dataset in the second database instance.

Abstract: Provided herein are systems and methods for indexed geospatial predicate search. An example method performed by at least one hardware processor includes decoding a query with a geospatial predicate. The geospatial predicate is configured between a geography data column and a constant geography object. The method further includes computing a first covering for a data value of a plurality of data values in the geography data column. The first covering includes a first set of cells in a hierarchical grid representation of a geography. The first set of cells represents a surface of the geography associated with the data value. A second covering is computed for the constant geography object. A determination is made on whether to prune at least one partition of a database organized into a set of partitions and including the geography data column based on a comparison between the first covering and the second covering.

Abstract: Techniques for implementing cross in-database machine learning are disclosed. In some example embodiments, a computer-implemented method comprises training a machine learning model in a first database instance using a machine learning algorithm and a training dataset in response to receiving a request to train, serializing the trained machine learning model into a binary file in response to the training of the machine learning model, recreating the trained machine learning model in a second database instance using the binary file in response to receiving a request to apply the machine learning model, and generating an inference result by applying the recreated trained machine learning model on the inference dataset in the second database instance.

Abstract: A bit vector having a bit vector length is accessed. A select operator directory tree can be generated using the bit vector. The select operator directory tree includes a first level of superblocks including large superblocks and small superblocks, a second level of blocks including large blocks and small blocks, each block associated with one of the superblocks, and a third level of sub-blocks, each sub-block associated with a block. The large superblocks each have, a length greater than a first constant that is independent of the bit vector length and the large blocks each have a length greater than a second constant that is independent of the bit vector length. The select operator directory tree can be stored. Related apparatus, systems, techniques and articles are also described.

Abstract: A plus-minus-one array in which adjacent entries vary by no more than positive one and no less than negative one is accessed. A range minimum query directory tree including blocks and subblocks of the plus-minus-one array is determined. Blocks are contained in the plus-minus-one array and subblocks are contained in the blocks. A data structure characterizing positions of minimum elements within the range minimum query directory tree is generated. The characterization includes positions of minimums within each subblock, between subblocks in a respective block, within each block, and between blocks. The data structure is stored. Related apparatus, systems, techniques and articles are also described.

Abstract: A plus-minus-one array in which adjacent entries vary by no more than positive one and no less than negative one is accessed. A range minimum query directory tree including blocks and subblocks of the plus-minus-one array is determined. Blocks are contained in the plus-minus-one array and subblocks are contained in the blocks. A data structure characterizing positions of minimum elements within the range minimum query-directory tree is generated. The characterization includes positions of minimums within each subblock, between subblocks in a respective block, within each block, and between blocks. The data structure is stored. Related apparatus, systems, techniques and articles are also described.

Abstract: A bit vector having a bit vector length is accessed. A select operator directory tree can be generated using the bit vector. The select operator directory tree includes a first level of superblocks including large superblocks and small superblocks, a second level of blocks including large blocks and small blocks, each block associated with one of the superblocks, and a third level of sub-blocks, each sub-block associated with a block. The large superblocks each have, a length greater than a first constant that is independent of the bit vector length and the large blocks each have a length greater than a second constant that is independent of the bit vector length. The select operator directory tree can be stored. Related apparatus, systems, techniques and articles are also described.