kappa-dev / kami-js Goto Github PK

Using link-left and link-right to connect action boxes to agents makes it awkward to re-arrange agents. When importing a large amount of nuggets, re-arranging is crucial to visualizing the graph. Unless the modeler has the foresight to link-left & link-right with a global perspective, the graphs become "tangled".

Unless link-left & link-right will be deprecated, I would suggest adding a "flip link-right & link-left" to an action's context menu, so that the modeler can alter this visual/logic element.

Attributes on nodes that have the same name will get their parents re-assigned when exporting to JSON. For example, the following graph:

Becomes the following JSON, where the rate attributes are now pointing to the same entity on the tree:

{"version":2.0,
 "agents":[
{"classes":["agent"],"father_classes":[],"path":[],"labels":["C"],"values":[],"x":474,"y":252},
{"classes":["agent"],"father_classes":[],"path":[],"labels":["D"],"values":[],"x":802,"y":248},
{"classes":["agent"],"father_classes":[],"path":[],"labels":["A"],"values":[],"x":474,"y":139},
{"classes":["agent"],"father_classes":[],"path":[],"labels":["B"],"values":[],"x":809,"y":127}
],
 "regions":[
],
 "key_rs":[
],
 "attributes":[
{"classes":["attribute","list"],"father_classes":["action","bnd","revers"],"path":["rev-bind"],"labels":["rate"],"values":["2.0"],"x":620.0760961020505,"y":323.9735501703761},
{"classes":["attribute","list"],"father_classes":["action","bnd","revers"],"path":["rev-bind"],"labels":["rate"],"values":["1.0"],"x":620.093746391759,"y":161.53215372790945}
],
 "flags":[
],
 "actions":[
{"classes":["action","bnd","revers"],"labels":["rev-bind"],
"left":[{"ref":["agents",0]}],
"right":[{"ref":["agents",1]}],
"context":[],
"x":596,"y":275.51666259765625},
{"classes":["action","bnd","revers"],"labels":["rev-bind"],
"left":[{"ref":["agents",2]}],
"right":[{"ref":["agents",3]}],
"context":[],
"x":585,"y":123}
]
}

When read back, Kami displays the following graph, where the rate attribute got its parent re-assigned:

This means the modeler must rename the actions with globally unique labels when writing the "local" nuggets, which is not feasible outside toy models.

With all nodes unlocked, attributes of actions are attracted to the canvas' top-left corner, and pull in that direction the rest of the graph & connected components.

Since one can not lock these attribute nodes into position, they often end up divorced from their parent node, sometimes gravitating to positions behind other nodes, making large graphs impossible to organize.