Repository files navigation

maize is a graph-based workflow manager for computational chemistry pipelines.

It is based on the principles offlow-based programming and thus allows arbitrary graph topologies, including cycles, to be executed. Each task in the workflow (referred to asnodes) is run as a separate process and interacts with other nodes in the graph by communicating through unidirectionalchannels, connected toports on each node. Every node can have an arbitrary number of input or output ports, and can read from them at any time, any number of times. This allows complex task dependencies and cycles to be modelled effectively.

This repository contains the core workflow execution functionality. For domain-specific steps and utilities, you should additionally installmaize-contrib, which will have additional dependencies.

You can find guides, examples, and the API in thedocumentation.

A taste for defining and running workflows withmaize.

"""A simple hello-world-ish example graph."""frommaize.core.interfaceimportParameter,Output,MultiInputfrommaize.core.nodeimportNodefrommaize.core.workflowimportWorkflow# Define the nodesclassExample(Node):data:Parameter[str]=Parameter(default="Hello")out:Output[str]=Output()defrun(self)->None:self.out.send(self.data.value)classConcatAndPrint(Node):inp:MultiInput[str]=MultiInput()defrun(self)->None:result=" ".join(inp.receive()forinpinself.inp)self.logger.info("Received: '%s'",result)# Build the graphflow=Workflow(name="hello")ex1=flow.add(Example,name="ex1")ex2=flow.add(Example,name="ex2",parameters=dict(data="maize"))concat=flow.add(ConcatAndPrint)flow.connect(ex1.out,concat.inp)flow.connect(ex2.out,concat.inp)# Check and run!flow.check()flow.execute()

If you plan on not modifying maize, and will be usingmaize-contrib, then you should just follow the installation instructions for the latter. Maize will be installed automatically as a dependency.

Note thatmaize-contrib requires several additional domain-specific packages, and you should use its own environment file instead if you plan on using these extensions.

To get started quickly with running maize, you can install from an environment file:

conda env create -f env-users.ymlconda activate maizepip install --no-deps ./

If you want to develop the code or run the tests, use the development environment and install the package in editable mode:

conda env create -f env-dev.ymlconda activate maize-devpip install --no-deps ./