\(\renewcommand\AA{\unicode{x212B}}\)

WorkflowAlgorithmRunner v1

Enable screenshots using DOCS_SCREENSHOTS in CMake

Summary

Manages dependency resolution, input/output mapping and running of algorithms.

Properties

Name	Direction	Type	Default	Description
Algorithm	Input	string	Mandatory	Name of the algorithm to run
SetupTable	Input	TableWorkspace	Mandatory	Table workspace containing the setup of the runs.
InputOutputMap	Input	TableWorkspace	Mandatory	Table workspace mapping algorithm outputs to inputs.

Description

This utility algorithm is used to control consecutive executions of another algorithm. After defining a table of desired input and output properties, WorkflowAlgorithmRunner connects the outputs to input properties, declares the order of execution, and runs the algorithm. It is mainly meant for managing the execution of workflow algorithms for data reduction purposes although it can run all types of algorithms.

The runs are specified in a TableWorkspace and given as the SetupTable input property to the algorithm. The first column has to contain unique string identifiers for each row. The rest of the columns have to be named same as the input/output properties of the managed algorithm.

Each row in the SetupTable corresponds to a single run of the managed algorithm. Its properties are set without modifications from the corresponding columns of the row except for special input/output columns declare in the InputOutputMap. For example, two independent runs of the Scale v1 would be specified as

Id	InputWorkspace	OutputWorkspace	Factor	Operation
1st run	ws1	ws1	0.42	Multiply
2nd run	ws2	scaled_ws2	100	Multiply

The Operation column could have been omitted in the above to use the default value of the Operation property of Scale.

InputOutputMap connects the outputs of a run to the inputs of another run. It is a TableWorkspace where the column names correspond to some or all of the managed algorithm’s input properties. The table has only a single row which lists the names of the output properties whose final values will be forwarded to the corresponding input property. To continue the Scale v1 example above, the InputWorkspace and OutputWorkspace properties would be connected by the following single column table:

InputWorkspace
OutputWorkspace

Note

Only workspace properties can be specified in the InputOutputMap.

Note

A single output property can be wired to several input properties. However, an input property cannot have multiple output properties.

If the one wanted to give the workspace scaled by ‘2nd run’ as an input for ‘1st run’, the InputOutputMap would look like the following:

Id	InputWorkspace	OutputWorkspace	Factor	Operation
1st run	2nd run	‘ws1’	0.42	Multiply
2nd run	‘ws2’	scaled_ws2	100	Multiply

This would result in ‘2nd run’ to be executed first to produce the ‘scaled_ws2’ workspace which in turn would be fed to ‘1st run’ as the InputWorkspace property. Behind the scenes, WorkflowAlgorithmRunner will replace the ‘2nd run’ in the InputWorkspace column of ‘1st run’ by scaled_ws2.

In the above, note the quotation marks around the ‘ws2’ and ‘ws1’ property names. These are hard coded input and forced output values, respectively. This tells WorkspaceAlgorithmRunner to omit input/output mapping and to keep these values as they are.

Note

A non-forced output property which is not needed by the other runs will be cleared by WorkflowAlgorithmRunner.

An equivalent python commands to running WorkflowAlgorithmRunner with the above InputOutputMap and SetupTable would be

Scale(InputWorkspace='ws2', OutputWorkspace='scaled_ws2', Factor=100, Operation='Multiply')
Scale(InputWorkspace='scaled_ws2', OutputWorkspace='ws1', Factor=0.42, Operation='Multiply')

Usage

Example: Let’s recreate the example in the Description section

# This is our initial workspace. We want to scale it first by 100
# and then by 0.42
CreateSingleValuedWorkspace(OutputWorkspace='ws2', DataValue=1.0)

# Setup the runs for the Scale algorithm
setupTable = WorkspaceFactoryImpl.Instance().createTable()
setupTable.addColumn('str', 'Run name') # First column can have arbitrary name.
# The rest of the columns can be in arbitrary order
setupTable.addColumn('str', 'InputWorkspace')
setupTable.addColumn('double', 'Factor') # Scale expects to get a number here.
setupTable.addColumn('str', 'OutputWorkspace')
row = {
    'Run name': '1st run',
    'InputWorkspace': '2nd run',
    'Factor': 0.42,
    'OutputWorkspace': '"ws1"' # Forced output either by '' or "".
}
setupTable.addRow(row)
row = {
    'Run name': '2nd run',
    'InputWorkspace': "'ws2'",
    'Factor': 100,
    'OutputWorkspace': 'scaled_ws2'
}
setupTable.addRow(row)
AnalysisDataServiceImpl.Instance().addOrReplace('setupTable', setupTable)

# Map OutputWorkspace to InputWorkspace
ioMap = WorkspaceFactoryImpl.Instance().createTable()
ioMap.addColumn('str', 'InputWorkspace')
ioMap.addRow({'InputWorkspace': 'OutputWorkspace'})
AnalysisDataServiceImpl.Instance().addOrReplace('ioMapTable', ioMap)

# Execute the algorithm
WorkflowAlgorithmRunner('Scale', SetupTable=setupTable, InputOutputMap=ioMap)

# Print some results
print('Original input value: {0}'.format(mtd['ws2'].dataY(0)[0]))
print('After scaling by 100: {0}'.format(mtd['scaled_ws2'].dataY(0)[0]))
print('After further scaling by 0.42: {0}'.format(mtd['ws1'].dataY(0)[0]))

Original input value: 1.0
After scaling by 100: 100.0
After further scaling by 0.42: 42.0

Categories: AlgorithmIndex | Utility

Source

C++ header: WorkflowAlgorithmRunner.h

C++ source: WorkflowAlgorithmRunner.cpp