Workspace¶

What are Workspaces?¶

Workspaces are the nouns of Mantid (while algorithms are the verbs). Workspaces hold the data in Mantid.

They come in several forms, but the most common by far is the MatrixWorkspace which contains measured or derived data with associated errors. Matrix Workspaces are typically created initially by executing one of Mantid’s ‘Load’ algorithms, for example LoadRaw or LoadNexus, or they are the output of algorithms which took a matrix workspace as input. In MantidPlot the data from the workspace can viewed as a table, and graphed in many ways.

Another form of workspace is the TableWorkspace. This stores data of (somewhat) arbitrary type in rows and columns, much like a spreadsheet. These typically are created as the output of certain specialized algorithms (e.g. curve fitting).

In addition to data, workspaces hold a workspace history, which details the algorithms which have been run on this workspace.

In software engineering terms, the ‘abstract’ concept of a workspace is an ‘interface’, in that it defines common properties that are implemented by various ‘concrete’ workspaces. Interaction with workspaces is typically through an interface. The concrete workspaces themselves are loaded in via Mantid’s plugin mechanism and are created using the Workspace Factory.

Example Workspaces¶

• MatrixWorkspace - A base class that contains among others:
  • Workspace2D - A workspace for holding two dimensional data in memory, this is the most commonly used workspace.
  • EventWorkspace - A workspace that retains the individual neutron event data.
• TableWorkspace - A workspace holding data in rows of columns having a particular type (e.g. text, integer, ...).
• WorkspaceGroup - A container for a collection of workspaces. Algorithms given a group as input run sequentially on each member of the group.

Working with Workspaces in Python¶

Workspace is an abstract description of an specific workspace implementation. It provides access to a few common properties without any knowledge of what the type of the workspace.

# This creates a workspace without explicitly capturing the output
CreateSampleWorkspace(OutputWorkspace="MyNewWorkspace")

# You can get a python variable pointing to the workspace with the command
myWS = mtd["MyNewWorkspace"]
print "The variable myWS now points to the workspace called ", myWS

# You can also ask Mantid to create matching python variables for all of it's workspaces
mtd.importAll()
print "MyNewWorkspace has been created that also points to the workspace called ", MyNewWorkspace

# You can assign a python variable when calling an algorithm and the workspace will match the variable name
myOtherWS = CreateSampleWorkspace()
print "myOtherWS now points to the workspace called ", myOtherWS

The variable myWS now points to the workspace called MyNewWorkspace
MyNewWorkspace has been created that also points to the workspace called MyNewWorkspace
myOtherWS now points to the workspace called myOtherWS

myWS = CreateSampleWorkspace()
print "name = " + myWS.getName()

myWS.setTitle("This is my Title")
print "getTitle = " + myWS.getTitle()

myWS.setComment("This is my comment")
print "comment = " + myWS.getComment()

print "id = " + myWS.id()

print "getMemorySize = " + str(myWS.getMemorySize())

name = myWS
getTitle = This is my Title
comment = This is my comment
id = Workspace2D
getMemorySize = ...

# Run a few algorithms
myWS = CreateSampleWorkspace()
myWS = ConvertUnits(myWS,Target="Wavelength")
myWS = Rebin(myWS,Params=200)

# You can access the history using getHistory()
history = myWS.getHistory()
for algHistory in history.getAlgorithmHistories():
    print algHistory.name()
    for property in algHistory.getProperties():
        if not property.isDefault():
            print "\t" + property.name() + " = " + property.value()

CreateSampleWorkspace
    OutputWorkspace = myWS
ConvertUnits
    InputWorkspace = myWS
    OutputWorkspace = myWS
    Target = Wavelength
Rebin
    InputWorkspace = myWS
    OutputWorkspace = myWS
    Params = 200

Writing you own workspace¶

This is perfectly possible, but not as easy as creating your own algorithm. Please talk to a member of the development team if you wish to implement you own workspace.

Category: Concepts