One repo to rule them all and its all Python

New Features:

• Merged all the code repos for the whole stack into one repo
• Pythonization of the whole BlazingSQL stack. See our blog post for more information
• New API for being able to query performance and execution logs
• Ability to create BlazingSQL tables from Hive tables
• Partial support for Non-equality joins. For example SELECT * FROM tableA as A INNER JOIN tableB as B ON A.key = B.key AND A.this_date > B.that_date
• Added arrow-provider

Improvements:

• Optimized simple queries that only have COUNT(*)
• Removed limitation on number of operands for outer joins
• Improved error messaging
• Improvements to relational algebra optimization

Bug Fixes:

• Fixed bug where a python script running BlazingSQL would hang at the end of a script
• Fixed bug when using wildcards for file paths and using dask distribution
• Fixed bug with HDFS
• Fixed bug with projects with large amounts of transformations on large GPUs
• Fixed bug with multiple projections on the same column
• Fixed COUNT(*) to properly ignore nulls
• Fixed stability issues with certains queries running on 3 or more nodes
• Fixed bug with querying a GDF and no transformations are applied
• Fixed bug with empty result sets
• Fixed bug with empty column names

New string operators, performance improvements and many bug fixes

New Features

• Implemented string concat operator
• Implemented substring operator

Improvements:

• Improved management of services
• Changed Apache Calcite schema database to an in-memory database
• Improved performance of communication between nodes by enabling parallel messaging
• Improved performance of data loading by enabling parallel file reading
• Added new distributed join method for joining small tables

Bug Fixes:

• Fixed various issues with Timestamp data types
• Fixed issue when column names were too long
• Fixed bug in relational algebra generation
• Fixed various bugs in communication layer
• Fixed bug with order by with strings
• Fixed issue with parsing Apache Parquet file schemas
• Fixed memory leak in joins
• Fixed memory leak in communication layer
• Fixed bug in table concatenation in disitrubiton algorithms
• Fixed bug when trying to join on columns of integers of different byte widths, or floats of different byte widths
• Fixed bug when trying to do a union on columns of integers of different byte widths, or floats of different byte widths
• Fixed bug in passing error message to user

Revamped data transport layer, LIKE operator and much more

New Features

• Completely revamped data transport layer is much faster and robust
• Added support for LIKE operator
• Added ability to create tables from Dask dataframes.
• Improved how services are launched from BlazingContext. Including new ready() function which checks to see if all services are online and shutdown() function to shutdown all services.

Improvements

• Improved performance logging
• Now using in-memory H2 database for Apache Calcite table catalog
• Updated to cudf v0.10

Bug Fixes

• Fixed bug in expression parsing
• Fixed various bugs with date literals, date functions and GDF_TIMESTAMP data type
• Fixed bug with aliases
• Fixed bug in order by for distributed queries when there are empty partitions
• Fixed bug in creating tables from S3 directories
• Fixed bug where predicate pushdown was not happening in certain types of queries

CAST and minor bug fixes

New Features

• Added support for CAST
• Added file_format parameter to create_table. This parameter is used for when the file format is not determinable from the file extension.

Bug Fixes

• Fixed bug where aliases would sometimes not be set correctly

Fixed Conda Packaging Versioning

New Features

• Added file_format parameter to create_table to help create tables from files that don't have extensions

Bug Fixes

• Fixed how releases are versioned for Conda
• Fixed bug with joining against an empty table

Google Cloud Storage support and MORE!

New Features

• Added support for CASE
• Improved support for Boolean columns
• Creating tables using wildcards in file paths
• Added support for Google Cloud Storage

Bug Fixes

• Fixed bug in groups by's with strings in distributed cluster
• Fixed issues in how BlazingContext launches processes
• Fixed issue where releases were being done in Debug mode
• Fixed bug related to creating multiple tables with the same name

Conda Install and JSON and ORC file tables

New Features:

• Ability to compile and install using Conda
• Creating BlazingContext can now automatically launches processes
• Support for creating tables from JSON and ORC files
• Added more CSV parsing parameters for creating tables from CSV files
• Updated to use cudf v0.9 release
• Added support for LIMIT

Bug fixes

• Fixed bug with processing queries using date literals
• Fixed distribution issues with data with nulls

Distributed Query Execution

A great deal has happened since we last released.

• We now support distributed query execution!
• Distributed results output to dask-cudf
• Updated to cuDF 0.9
• Millions, literally millions, of bug fixes.
• No longer use main. before any table names. That was awful.
  bc.sql('select * from main.table_name') --> bc.sql('select * from table_name')

simple-distribution-tcp-cudf0.7

before cudf 0.8 and before table scan