Architecture & Design
How g.Pype is built, for the reader who writes against it: what a node, a pipeline and a run are made of, what a call accepts and refuses, and why it is shaped that way. What a running system does – install, operate, observe, deploy – is Advanced Topics.
The three parts of every node, why they run at three different times, and how to write one of your own.
Realtime and batch: how a pipeline runs, which call drives it, and the
pipelines start() and run() refuse.
What holds only offline: a Result and what it carries, zero-phase
filters, a plain function as a node, exceptions through your own line,
and the bridges to MNE and scikit-learn.
One master timeline per run, the arrival stamp every live source adds, and how an event lands on the sample it happened in.
Labels, roles and units: what travels with every channel, what reads it, and where it gets lost.
The three moments a pipeline is checked before data flows, and the message each refusal gives.
Three ways to the samples as an array: live in memory, live frame by frame, and from a recording.
Shape, dtype, units, channel order, what time means in each output, frames and buffering, in one place.
Which thread runs what, how a frame travels through the nodes, the GUI thread, and what happens when the pipeline cannot keep up.