Basic concepts

In SciPipe, we are discussing a few concepts all the time, so to make sure we are on the same page, we will below go through the basic ones briefly.


The probably most basic concept in SciPipe is the process. A process is an asynchronously running component that is typically defined as a static, "long-running" part of the workflow, and the number of processes thus is typically fixed for a workflow during its execution.

One can create customized types of processes, but for most basic workflows, the scipipe.SciProcess will be used, which is specialized for executing commandline applications. New SciProcess-es are typically created using the scipipe.NewProc(procName, shellPattern) command.


The "long-running" processes mentioned above, will receive input files on its in-ports, and for each complete set of input files it receives, it will create a new task. Specifically, scipipe.SciProcess will create scipipe.SciTask objects, and populate it with all data needed for one particular shell command execution. SciTask objects are executed via their Execute() method, or CustomExecute(), if custom Go code is supposed to be executed instead of a shell command.

The distinction between processes and tasks is important to understand, for example when doing more advanced configuration of file naming strategies, since the custom anonymous functions used to format paths are taking a SciTask as input, even though these functions are saved on the process object.

To understand the difference between processes and tasks, it is helpful to remember that processes are long-running, and typically fixed during the course of a workflow, while tasks are transient objects, created temporarily as a container for all data and code needed for each execution of a concrete shell command.


Central to the way data dependencies are defined in SciPipe, is ports. Ports are fields on processes, which are connected to other ports via channels (see separate section on this page).

In SciPipe, each port must have a unique name within its process (there can't be an in-port and out-port named the same), and this name will be used in shell command patterns, when connecting dependencies / dataflow networks, and when configuring file naming strategies.

In SciProcess objects, in-ports are are accessed with myProcess.In("my_port"), and out-ports are similarly accessed with myProcess.Out("my_other_port"). Both are of type FilePort.

Some pre-made components might have ports bound to custom field names though, such as myFastaReader.InFastaFile, or myZipComponent.OutZipFile.

Port objects have some methods bound to them, most importantly the Connect() method, which takes another port, and connects to it, by stitching a channel between the ports.

On SciProcess objects, there is also a third port type, ParamPorts, which is used when it is needed to send a stream of parameter values (in string format) to be supplied to as arguments to shell commands.


Ports in SciPipe are connected via channels. Channels are plain Go channels and nothing more. Most of the time, one will not need to deal with the channels directly though, since the port objects (see separate section for ports) have all the logic to connect to other ports via channels, but it can be good to know that they are there, in case you need to do something more advanced.


The Workflow is a special object in SciPipe, that just takes care of running a set of components making up a workflow.

There is not much to say about the workflow component, other than that it is created with scipipe.NewWorkflow(workflowName), that all processes need to be added to it with wf.Add(proc) while the "last", or "driving" process needs to be specified with wf.SetDriver(driverProcess), and that it should be run with wf.Run(). But this is already covered in the other examples and tutorials.

Shell command pattern

The SciProcess has the speciality that it can be configured using a special shell command pattern, supplied to the NewProc() factory function. It is already explained in the section "writing workflows", but in brief, it is a normal shell command, with placeholders for in-ports, out-ports and parameter ports, on the form {i:inportname}, {o:outportname}, and {p:paramportname}, respectively.