Skip to main content

Node

Simple Story

Imagine the pupil's compiled answer drawn as a map of dots joined by arrows. Compiled, not written: the map holds things the compiler produced that the pupil never typed.

A node is one dot. Depending on how closely you look, a dot might be a whole class or a single method. Everything the reach check does is answering questions about which dots exist and which arrows lead where.

What it is​

A node is one element of the analysed program. Which elements count as nodes depends on the engine: ArchUnit reasons mostly about classes and their members, while T. J. Watson Libraries for Analysis (WALA) reasons about individual methods, because a call graph joins methods rather than classes.

Nodes come from compiled bytecode, not from source. That matters twice over: the analysis sees what the compiler produced, including synthetic and bridge methods the author never wrote, and it can analyse a dependency for which no source is available.

In Ares 2​

Nodes are classified before they are judged. A node belonging to student code is subject to the policy; a node belonging to Ares itself, to the test harness, or to a class named in theFollowingClassesAreTestClasses is trusted. That classification is the whole reason the trust boundary can be drawn at all, and it is why student code must never land in a reserved package: a node in a reserved package would be classified as trusted.

Further reading​