As stated in the tutorial , each vertex has a unique number as its id, starting from 0 continuously to N-1, where N is the number of vertices in the graph. However, the id for a vertex p may change if one vertex prior to it was deleted from the graph. So what is the most efficient way to refer to p …

The name attribute is treated in a special way by the internals of the Python interface; in particular, the name attribute is indexed internally in a Python dict so looking up the ID of a node given its name attribute is an O(1) operation (on average). The index is invalidated if you mutate the grap…

Thanks for your comprehensive reply. This solved my problem. By the way, I find that there is the relative explanation in the tutorial. Maybe I missed the part when I read the tutorial. Anyway, thanks for your time and patience.

A further question: can we get the handle of an edge in a graph by using its name attribute in O(1)? I failed to get the handle of an edge by the following code: edge = g.es.find(edge_name) However, the following line works, which I do not know whether it is executed in O(1). edge = g.es.select(…

The name attribute of edges is treated the same way as any other attribute, so no, there is no indexing behind that and g.es.select(name_eq=edge_name) would make a full scan. In theory it wouldn’t be too hard to add indexing for the edges as well but so far it has been a relatively uncommon scenario…

Thanks for the reply. I agree with you that “add indexing for the edges as well but so far it has been a relatively uncommon scenario”. Currently g.get_eid(v1, v2) is enough for me.

This story about edge names will become much more interesting if we start working with multigraphs and remove edges at the same time. Consider the following graph: [image] Distinguishing the two edges between 1 and 3 is important because removing one merges the green/orange, removing the other me…

Can the special name code paths handle arbitrary Python objects? I could not tell from the docs or this thread. I have some classes that define __eq__ and __hash__ (so they can be used as keys in a dict) that represent vertices in a graph. It seems like there is some support for arbitrary hashable…

[image] adriangb: Can the special name code paths handle arbitrary Python objects? No, they cannot. Well, in theory they could because we use an internal Python dictionary to map from vertex names to indices; the part where it breaks is when we handle method calls like g.degree(some_obj); her…

Interesting, that is useful. One unfortunate thing is that even with that, if I want to do an operation on a sequence of vertices I still have to iterate over them and get their indices in Python. It would be nice to push that iteration into C since I think that would be faster. Other than that, i…

most efficient way to refer to a vertex in a graph

Usage

Topic		Replies	Views
Making it possible to use arbitrary vertex names Development Python	23	1878	15 November 2021
Check if vertex exists, create it if it doesn't--igraph is Much Slower than Networkx Usage Python	3	2033	9 September 2021
How can I query a vertex/edge according to its attribute? Usage C	4	552	20 February 2023
Does igraph accept json format and why are nodes represented as numbers? Usage Python	5	489	13 September 2021
What is the node order of Graph.indegree()? Usage Python	3	310	19 December 2022

most efficient way to refer to a vertex in a graph

Related topics