Generate only directed acyclic graphs about hypothesis-networkx HOT 7 CLOSED

Unless there's anything further on generating DAGs I'll close the issue. Feel free to reopen it if it's not resolved.

from hypothesis-networkx.

Interesting question :)
The directed part is easy (graph_type=nx.DiGraph).
For the acyclic part: Could you try with max_edges=0, connected=True? The algorithm won't make (undirected) cycles in the first phase, and uses any "leftover" edges to create possibly cyclic edges later. However, this will greatly limit the type of graphs you'll get. I would try filtering/assuming first, since I don't really expect too many cyclic directed graphs.

from hypothesis-networkx.

But won't max_edges=0 mean there won't be any edges to any node?

from hypothesis-networkx.

No, that constraint will be violated to ensure the graph is (weakly) connected.

from hypothesis-networkx.

Well it makes my test run forever since sometimes there are too many nodes.

from hypothesis-networkx.

It scales poorly with the number of nodes (at least N^2)
From the readme:

Graph generation may be slow for large graphs.

from hypothesis-networkx.

Nah it's my test that is slow, not the graph generation itself.

from hypothesis-networkx.