quantlab.factor.kunquant.run_batch_graph#

quantlab.factor.kunquant.run_batch_graph(name, function, arrays, timestamps, symbols, njobs)[source]#

Compile function for a whole range (through compile_graphs) and run it over arrays.

For a computation whose inputs do not come from a dataset panel (a factor reading another factor’s store, say). The graph is compiled in the TS layout with BATCH_OPTIONS, the symbols padded and cut as FactorKunQuant does, and run on the shared executor.

Parameters:

namestr: The compiled module’s name.
functionKunQuant.Stage.Function: The graph; its Input names are the keys of arrays.
arraysdict[str, np.ndarray]: Contiguous float32 [time, symbol] inputs.
timestamps, symbolsnp.ndarray: The axes of the inputs and of the returned panel.
njobsint: Executor threads.

Returns:

xr.Dataset: Every output of the graph on (timestamp, symbol).

Examples

>>> b = Builder()
>>> with b:
...     Output(Input("x") * 2.0, "y")
>>> run_batch_graph("Double", Function(b.ops), {"x": x}, times, symbols, 2)["y"].shape
(3, 8)