You don't have to check after every removed starlane (removing a starlane is the same as adding a wall). You just do a graph traversal at the end, and undelete starlanes to reconnect isolated sub-webs to the main web. This is relatively simple, since all deleted starlanes were originally good and not conflicting with others, so you don't have to calculate whether they'd create secondary problems. (Your tree branching would need a similar system to add starlane in a later step, though it would be a bit easier to calculate since you'd already have connectivity, though you would probably need to add more connections.)drek wrote:Wall adder tend to be slower and more complex, since you need to check for connectivety between all nodes after placing your wall.
Deleting starlanes would happen one at at time, and would be either removal of an isolated starlane, or removal of a starlane that connect to a star where a starlane has already been removed. The ratio of how often these two are done would determine the ratio of dead ends to looped connections, I think.
There should be a minimum number of starlanes out of a system, at which more starlanes can't be removed (probably 1 starlane... or perhaps more).
Any particular reason why?I really want to avoid crisscrossing, aka weaving.
The basic geometry of growing tree is rather boring though... and is always basically the same. There's never any blockage of the direct path down the tree to the central / originating system. You can have a second step to add interconnections in the mid-branches (and leaves), but what makes things really interesting is when there's an interesting absence of starlanes, imho.With the Growing tree on that page, there's no need for dead end checking or an isolation removing step.