You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At present there are a number of operations (such as GraphvizRemoveNode) that do nothing if their arguments are valid.
For example, doing GraphvizRemoveNode(gv, "non-existent-node"); just returns gv unaltered. This might be ok, but we should try to be consistent everywhere and do one of the following:
Nothing (currently what happens)
Issue a warning that "non-existent-node" doesn't exist and that the first argument will be unaltered;
Issue an error
There might be some merit in having code that just doesn't error out, but just carries on. On the other hand, I think this will violate the principal of least astonishment, i.e. I have a graph, I remove a node (mistyping its name), and no nodes are removed (surprised face!). I think issuing a warning is the minimum we should do, but think errors would be better (there's no way to inadvertently ignore an error whereas a warning can be disabled, swallowed by something else, ignored or just overlooked).
What say @mpan322 and @mtorpey ? I'm raising this issue for discussion about what's desirable, rather than as a TODO for anyone.
The text was updated successfully, but these errors were encountered:
At present there are a number of operations (such as
GraphvizRemoveNode
) that do nothing if their arguments are valid.For example, doing
GraphvizRemoveNode(gv, "non-existent-node");
just returnsgv
unaltered. This might be ok, but we should try to be consistent everywhere and do one of the following:"non-existent-node"
doesn't exist and that the first argument will be unaltered;There might be some merit in having code that just doesn't error out, but just carries on. On the other hand, I think this will violate the principal of least astonishment, i.e. I have a graph, I remove a node (mistyping its name), and no nodes are removed (surprised face!). I think issuing a warning is the minimum we should do, but think errors would be better (there's no way to inadvertently ignore an error whereas a warning can be disabled, swallowed by something else, ignored or just overlooked).
What say @mpan322 and @mtorpey ? I'm raising this issue for discussion about what's desirable, rather than as a TODO for anyone.
The text was updated successfully, but these errors were encountered: