Open deitch opened 1 year ago
I delved into it quite a bit. It looks like it is because many of the elements (clusters in their parents, elements in their clusters, etc.) are all kept in maps, and then it ranges over the maps. It should be possible to range over the maps to get keys, then sort, then range over that. I tried, but I could not quite figure out where the right place to do it, as I don't understand the templating structure.
Is there a way to get consistent (or nearly consistent) comparable output?
My specific use case is that we have protobufs, and expect to have the graph files checked into the repo as part of the commit. The simplest way to do this (we do it for generated language bindings) is to run
protodot
, and check if the file is different from what is checked in.Except the output isn't consistent.
From my checks, I found several things:
generated by
comment at the end; we can ignore thatsubgraph cluster_<id>
: the ID appears to be generated using some data about the content. As far as I can tell, this is consistentNode_Ja_<number>
: the numbers appear to be consistent with the content in themI don't know the internals of
protodot
well enough, but it looks like consistent sorting of the subgraphs and nodes within subgraph might go a long way?