DolphinDB Release Notes

Release Date : 2019-04-10

Version : 0.95.0

DolphinDB Server

New Features

First release of 32 and 64 bit DolphinDB server for ARM on Linux.
First release of 32 bit DolphinDB server for x86 CPU on Windows and Linux.
Added the feature of filtering for streaming data subscription. For example, we can specify a list of stocks when subscribing to a table of stock trades. Only data about these stocks will be published to the subscriber.
Added cross-sectional aggregator for streaming clients. With this aggregator, we can calculate cross-sectional metrics with built-in or user-defined aggregated functions (e.g. min, max, avg, percentile, median, etc).
Added the feature of historical data replay. With this new feature, we can use the same script to process both historical data and real-time data. The related functions include replay and replayDS.
Added 3 machine learning methods: generalized linear regression, logistic regression and principal component analysis (PCA).
Added a synchronized dictionary object that supports concurrent data access and manipulation. Use function syncDict to create a synchronized dictionary.
Added a new type of in-memory table (mvcc table) that can be persisted to disk by write-ahead logging (WAL). When appending, updating or deleting rows of an mvcc table, another version of the table is created so that concurrent reading will not be blocked. The mvcc table is optimal for the use case with frequent read and append, but few update and delete operations.

Improvements

Integrated OpenBLAS for matrix multiplication optimization. The performance of large matrix multiplication speeds up 10~20 times.

Bug fixes

Fixed a bug in saving a symbol-type column in a partitioned local table on disk. When appending a string vector to a symbol-type column of an empty partitioned local table on disk, the underlying symbol base wouldn't be saved. (2019-04-16 0.95.3 Linux64 binary | Linux32 binary | Windows64 binary | Windows32 binary | ARM64 binary | ARM32 binary )
Fixed a bug of mvcc table with persistence. The system crashes when erasing all rows of a mvcc table that enables persistence. (2019-04-16 0.95.3 )
Fixed a performance issue of conditional statements that executes the "if" statement twice. (2019-04-15 0.95.2 Linux64 binary | Linux32 binary | Windows64 binary | Windows32 binary | ARM64 binary | ARM32 binary )
Fixed a potential deadlock issue of function getJobReturn. (2019-04-15 0.95.2 )
Fixed a bug of function replay with multiple tables. With this fix, all tables are replayed at the same speed. (2019-04-11 0.95.1 Linux64 binary | Linux32 binary | Windows64 binary | Windows32 binary | ARM64 binary | ARM32 binary )

Java API

Support data filtering and auto-reconnect when subscribing to a streaming table.

Python API

The Python API is rewritten in C++. The read/write performance is greatly improved. For batch data upload, the performance is improved 10-30 times; for batch data download, the performance is improved 5-10 times.
Python API now supports subscriptions to streaming tables. It also supports filtering for streaming data subscription and automatic reconnection after disconnection. It works with Python 3.4 to 3.7. The current version does not support jupyter notebook under Linux.

C++ API

Support data filtering and auto-reconnect when subscribing to a streaming table.

C# API

Support data filtering and auto-reconnect when subscribing to a streaming table.