JlTbl is a column-store table data structure and accessories written in Julia. It is designed and optimized for data analysis but can also be used for data processing and transformation. It is intended for single-machine in-memory use but does not preclude permanent-storage realizations nor use as a building block in distributed storage and processing.
- designed and optimized for reading, querying, and bulk loading
- supports all relational table operations and queries, including multi-table queries
- table size is known at all times and so is constant time to query
- lightweight dynamic views that can be materialized as new tables
- dynamic views can be used to modify the underlying table and reflect underlying modifications
- read-only views
- tables and views share the same core API
- convertable to a matrix (or a matrix view)
- in-memory implementation but API design does not preclude a permanent storage implementation
- supports random access of rows, columns, and cells (values)
- columns can store:
- any of the core data types, or arrays, or objects
- homogeneous values or a mix of sub-types
- sets or enumerations of values
- null
- columns can be indexed (and compressed?)
- both lightweight rows for reading and instantiated rows for streaming data processing
- table-level concurrency control that permits many concurrent readers but writers are serialized; appends can coexist with reads, but other writes are exclusive
- extent-based storage and allocation so that resizing does not need copying
- reading and writing a number of external file formats (CSV, YAML, ARFF, Prolog)
- tables can be grown by
- appending columns
- appending rows
JlTbl is free, open source software. It is released under the MIT
License. See the file LICENSE for details.
Contact for this software is handled through GitHub. Start by checking to see if there are any relevant issues. Then open a new issue to report a bug or ask a new question.
Copyright (c) 2014 Aubrey Barnard. This is free software. See LICENSE for details.