Violation generation

Experimental: this can change at any time

Violation generation is experimental. The shifty.experimental.vio Python module, its command line, its output files, and shifty_experimental::vio in Rust can change in any release, without a deprecation period. Pin the Shifty version if you depend on them.

A repair tool takes a graph that fails validation and edits it until it passes. Passing is easy to check, and it says little: deleting the failing node passes too. To score a repair, you need to know what the graph was meant to be.

shifty.experimental.vio gets that answer by working backwards. Start from a graph that conforms, break it with a few recorded edits, and keep the edits. Undoing them is a repair known to be correct.

from shifty.experimental import vio

universe = vio.Universe("model.ttl", "shapes.ttl")
case = universe.draw(3, seed=0)

case.graph()   # the broken graph: give this to the repair tool
case.repair    # the edits that undo the breakage: the answer

A case can hold several violations at once, and they can interact: fixing one can expose or hide another. Every failure a case lists was confirmed by validating the broken graph. The same seed always gives the same case.

It works from Python, from the command line (python -m shifty.experimental.vio), and from Rust (shifty_experimental::vio).

Break a graph and repair it	Break a small graph, catch it with the validator, and repair it.
Build a repair benchmark	Build a benchmark suite from your own model, and score a tool on it.
How violation generation works	How the edits are found, picked, and checked.
Violation generation reference	Classes, options, output files, and limitations.