net.declared_track_width

What it is ¶

net.declared_track_width(net, mm) yields the track width a net SHOULD be routed at, according to the project’s own net-class declarations. The value is held in metres and shown in mm. There is one row for each net the project actually constrained. It is the resolved answer, computed after the cascade across the net’s classes.

Pair it with board.track_width(net, mm), the ACTUAL routed value in the same units. Declared versus actual, with no number the engine invented.

For hardware engineers ¶

A net can sit in several classes at once, and the classes can disagree. KiCad does not pick one class and use it wholesale. It fills each constraint from the highest-priority class that states that particular constraint, and the Default class fills whatever is left over. So a net can take its clearance from one class and its track width from another, exactly as the tool does when it routes.

This relation runs that same resolution, so your layout tool would enforce the number here. Ask it when you want to know what a net was supposed to be, not what it is.

For software engineers ¶

Join THIS, not netclass.track_width. Membership is 1:many (WS1-050). Joining net.netclass(?net, ?class) to the per-class relation fans out over every class the net belongs to, and a comparison then fires against classes that lost the cascade. Imagine a net routed correctly to its winning class’s value. That join reports it as a confident, wrong finding.

Rows are 1:1 with constrained nets. If none of a net’s classes states this quantity, the net yields no row. A rule joining it then selects only nets the project genuinely constrained, and not net.declared_track_width(?n, ?_) reads as unconstrained.

Go projector ¶

netDeclaredFacts in stdlib/relations/facts.go projects it. For each net it orders the net’s classes by the project’s priority, ascending, with unstated and unknown classes last. It then takes the first class that states the field, independently per field. The citation records WHICH class supplied the value (net_settings:<class>), so a finding can say where the limit came from.

Absence is not a pass ¶

Only a KiCad project read populates this, and only when the project declares class definitions. Take a rule comparing declared against actual. On every other source it finds nothing to compare and reports clean, and a review cannot tell that from a genuine pass. Such a rule gates on design.has_netclass_defs, and on the board tier for the actual side, so an absent declaration reads not-applicable.

Datalog ¶

What every constrained net should route at:

net.declared_track_width(?net, ?mm) => ?net, ?mm

Nets routed narrower than their own project demands:

net.declared_track_width(?net, ?declared), board.track_width(?net, ?actual), ?actual < ?declared => ?net, ?actual, ?declared