`{}` Real-time UDFs in q¶

take either no arguments, or a 2-list: (tableName;tableData)
have access to the get* APIs, that operate on a local set of data
should return a table; if the result is not a table it will be wrapped as a 1×1 table with column name result

Examples¶

With arguments:

{[tabName;data] select avg price from data}

With no arguments, using getTicks API:

{[] select avg price from getTicks`symList`assetClass`dataType!`VOD.L`equity`trade}

Trigger functions¶

Trigger functions

operate on a table to determine whether a UDF should be run
take a single argument, the table listed in the trigTab field
return a boolean atom: whether to run the UDF or not

Example:

{[data] `mySym in data`sym}

In this example, we would execute our UDF if the trigTab contained the symbol `mySym.

Initialization functions¶

Initialization functions are run on boot of the UDF worker nodes to prepare the process for UDF execution. They are useful for one-time loads of configuration, libraries, or reference data. Initialization functions take no arguments.

Example:

{[]
    system"l ",getenv[`UDF_LIBS_DIR],"/myDependency.q";
    .myUDF.referenceTable:([]sym:`a`b;num:1 2); }

In this example, we load a q script and set a table that we can later reference in a namespace. Dependencies (such as the q script in the example above) can be placed in the UDF_LIBS_DIR which is located in the udfdir. Dependencies will then be included when the UDF package is exported via the command line interface.

Recommended dependencies¶

ml.q (KX Machine Learning library)
Pandas (Python library)

Recommended dependencies enable smoother conversion between q tables and Python, as in the pre-execution function example above. New UDFs can be added to the system and managed with

Querying results¶

Outputs of real-time UDFs can be accessed through the getUDF API call.

{} Real-time UDFs in q¶