speed of all_neighbours

graph-tool
1

I am new to graphtool but super excited by what I have implemented so far. I
have a question however about the speed of the all_neighbors method.

From trying it on various machines, it seems that it takes around 0.01s for

v.all_neighbors() where v is an arbitrary vertex. I have tried this on a 4
node graph and a 2M node graph getting the same result. Has anyone else
found the same thing? It seems to me to be particularly slow. Storing a
graph as an adjacency list in a hashtable should have microsecond look up
time, I assumed graph-tool stored the graph in such a way but I haven't yet
come across the place where the architecture it is described.

At any rate, am I doing something wrong or is 0.01s the expected look up
time?

I am trying to build a list of "friends of friends" on a graph of around 2M
nodes 200M edges, so if all_neighbors takes 0.01s, then to get "friends of
friends" takes about 1s for each node....which prohibits traversing the
graph in this way in a reasonable time

Nk

2

In [2]: %timeit v.all_neighbors()
263 ns ± 1.4 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

So, not quite.

As usual, without a minimal self-contained example of what you are doing, it
is impossible to give any concrete help.

Best,
Tiago

3

yes, sorry for the lack of detail. Indeed, the issue was my method of timing

start=time.time()
v.all_neighbors()
print(time.time()-start)

there is overhead with the print statement, I suspect it is in converting
the decimal.

thanks for your help!