# Problem with weight retrieval in Directed Graphs
**URL:** https://discourse.julialang.org/t/problem-with-weight-retrieval-in-directed-graphs/120784
**Category:** Graphs
**Tags:** question, graphs
**Created:** [October 1, 2024, 1:30pm UTC](https://discourse.julialang.org/t/problem-with-weight-retrieval-in-directed-graphs/120784 "2024-10-01T13:30:52Z")
**Posts on this page:** 4
**Page:** 1
### Author:  [@paul20](https://discourse.julialang.org/u/paul20)
#### Post date: [October 1, 2024, 1:30pm UTC](https://discourse.julialang.org/t/problem-with-weight-retrieval-in-directed-graphs/120784/1 "2024-10-01T13:30:52Z")
I’ve been using the `SimpleWeightedDiGraph` from the SimpleWeightedGraphs package for creating directed weighted graphs. Recently, I encountered a problem related to weight retrieval of edges.
When initializing a `SimpleWeightedDiGraph` with a predefined adjacency matrix, retrieving the weight of a specific edge using `weights` indexing returns an incorrect value, while using the `get_weight` function provides the correct weight.
```julia
test_matrix = [0.0 1.0 0.0;
0.0 0.0 1.0;
1.0 0.0 0.0]
test_graph = SimpleWeightedDiGraph(test_matrix)
println("Outneighbors of node 1: ", outneighbors(test_graph, 1))
test_graph.weights[1,2] # it really seems there is a bug there...
get_weight(test_graph, 1, 2)
```
I get inconsistent outputs:
```julia
julia> test_graph.weights[1,2]
0.0
julia> get_weight(test_graph, 1, 2)
1.0
```
If I initialise the directed graph using the transpose of the adjency matrix, then the test\_graph.weights matches my matrix, but then the get\_weight and the outneighbors function do not match anymore…so it seems the stored adjency matrix in the directed graph is the tranposed of what it should be.
I saw in another post ([Deleting edges from SimpleWeightedDiGraph based on edge weight - #3 by anon94023334](https://discourse.julialang.org/t/deleting-edges-from-simpleweighteddigraph-based-on-edge-weight/35835/3)) that someone already raised this question. I don’t know if I am missing something/doing something wrong or it is a bug.
Thank you so much for your help!
---
### Author:  [@gdalle](https://discourse.julialang.org/u/gdalle)
#### Post date: [October 1, 2024, 2:30pm UTC](https://discourse.julialang.org/t/problem-with-weight-retrieval-in-directed-graphs/120784/2 "2024-10-01T14:30:09Z")
Hi @paul20, welcome to the community 👋
> [@paul20](#):
>
> so it seems the stored adjency matrix in the directed graph is the tranposed of what it should be.
You figured it out, that’s exactly what happens under the hood!
This transposed storage was chosen for performance reasons. Since (sparse) matrices in Julia are column-major, it is very fast to query all the nonzero coefficients in the column `j`. Usually, for directed graphs, we want `outneighbors` to be fast, so we need the out-neighbors of vertex `j` to be stored in the same column… which means we should work with the transpose of the adjacency matrix.
I agree that it is not very well documented at the moment, but there is a clue in the docstring for `SimpleWeightedDiGraph`:
```julia
help?> SimpleWeightedDiGraph
search: SimpleWeightedDiGraph SimpleWeightedDiGraphEdge SimpleWeightedEdge
SimpleWeightedDiGraph{T,U}
A type representing a directed weighted graph with vertices of type T and
edge weights of type U.
Fields
≡≡≡≡≡≡
• weights::SparseMatrixCSC{U,T}: weighted adjacency matrix, indexed
by (dst, src)
```
Right at the end, it says "indexed by `(dst, src)`, i.e. `(destination, source)` of each edge. You’re welcome to submit a pull request improving this documentatin 🙂
Does that clear things up for you?
---
### Author:  [@paul20](https://discourse.julialang.org/u/paul20)
#### Post date: [October 1, 2024, 2:35pm UTC](https://discourse.julialang.org/t/problem-with-weight-retrieval-in-directed-graphs/120784/3 "2024-10-01T14:35:48Z")
Yes, I oversaw this comment in the end:) Thanks a lot for your answer!
---
### Author:  [@gdalle](https://discourse.julialang.org/u/gdalle)
#### Post date: [October 1, 2024, 2:41pm UTC](https://discourse.julialang.org/t/problem-with-weight-retrieval-in-directed-graphs/120784/4 "2024-10-01T14:41:44Z")
Feel free to mark my answer as the solution if you have no further questions, so that other people know this topic has been handled