error when running compute_geodist_PtE

[LLJiang01]

Error in [<-(*tmp*, graph.temp$.enclos_env$private$data[["__dummy"]], :
subscript out of bounds

issue.zip

[AlexandreBSimas]

Hi Lulu,

Thanks for raising this issue, however you just pasted some error message and attached a file without giving any context. Unlike ChatGPT or Claude, where people only paste the error and do not say anything else, to better fix the issue, we rely on clear explanations to understand and troubleshoot problems effectively.

Also, if possible, can you create a small reproducible example with your issue? The graph in your attached file has nearly 30,000 vertices, which makes it quite difficult to test efficiently. A more manageable example would make debugging much easier.

Best,
Alexandre

[LLJiang01]

Ah I see. Sorry for the explanation, I explained it with professor David today, so I just giving the error context. Large number of vertices: some choice of node will not return error, and currently only when I choose node as I attached above it would return error. I will try to make it smaller now.

[LLJiang01]

I tried running Graph_WM$compute_geodist_PtE( node, normalized = TRUE, include_vertices = FALSE, verbose = 2 ).

I tested it with 10 test data points and 411 real data points, and both ran successfully, returning distance matrices of size 10×10 and 411×411, respectively.

Now, I want to compute the distance from my generated mesh nodes to the 411 real data points. So, I combined the real data and mesh nodes using rbind and ran the function, then attempted to extract the distances between them.

However, when I did this, the error occurred. This is why I didn’t choose to upload a smaller set of nodes.

[AlexandreBSimas]

From your message it seems you only tried with Graph_WM (which is very big), is that so?

The idea of a smaller example would be to also consider a much smaller graph to begin with, and try to replicate this problem there. Did you try with a smaller graph?

[LLJiang01]

Yes. I only tried Graph_WM.

Not yet. I'm now trying to cut the node into 3 part, and then run them one by one to see is it because of some special node that the error occurs.

[LLJiang01]

I can try generating a smaller graph, but since this graph runs fine on other smaller test nodes and only returns an error when I select this specific graph and node, I’m not sure if I can reproduce the error on a smaller graph.

[LLJiang01]

load("~/Desktop/Crime/issue/node.RData")
ee <- Graph_WM$compute_geodist_PtE(

• node,
• normalized = TRUE,
• include_vertices = FALSE,
• verbose = 0
• )
  Error in [<-(*tmp*, graph.temp$.enclos_env$private$data[["__dummy"]], :
  subscript out of bounds
  In addition: Warning message:
  In system.time({ :
  Duplicated locations were found when computing geodist. The returned values are given for unique locations.

load("~/Desktop/Crime/issue/node.RData")
node <- node[1:10000,]
ee1 <- Graph_WM$compute_geodist_PtE(

• node,
• normalized = TRUE,
• include_vertices = FALSE,
• verbose = 0
• )
  Warning message:
  In system.time({ :
  Duplicated locations were found when computing geodist. The returned values are given for unique locations.

load("~/Desktop/Crime/issue/node.RData")
node <- node[10001:20000,]
ee2 <- Graph_WM$compute_geodist_PtE(

• node,
• normalized = TRUE,
• include_vertices = FALSE,
• verbose = 0
• )

load("~/Desktop/Crime/issue/node.RData")
node <- node[20001:32559,]
ee3 <- Graph_WM$compute_geodist_PtE(

• node,
• normalized = TRUE,
• include_vertices = FALSE,
• verbose = 0
• )

I cut the node into 3 part, and they all ran successfully...

[AlexandreBSimas]

I found the issue. One thing is that you were using node with duplicated entries (currently we are just using the unique entries, so it means that even if it did not give an error, the output would have a different dimension than your input). Also, there were observations that were close (either very close, or mapped to the same point, for example, the point at position 1 of edge 1 might coincide with position 0 of edge 2).

We will make the function robust to these kinds of inputs, but for now, I guess the best way if for you to try to build node more carefully.

I will let you know here when we fix it.

[LLJiang01]

So quick) OK! Sorry to bother you. Thank you!

[AlexandreBSimas]

Don’t worry, it’s all good.

For next time, it would be really helpful if you could describe the problem in more detail and, if possible, provide a small reproducible example. That makes it much easier to understand and debug the issue.

Just to illustrate, here’s a simple and small reproducible example:

library(MetricGraph)

edge1 <- rbind(c(0,0),c(1,0))
edge2 <- rbind(c(0,0),c(0,1))
edge3 <- rbind(c(0,1),c(-1,1))
theta <- seq(from=pi,to=3*pi/2,length.out = 50)
edge4 <- cbind(sin(theta),1+ cos(theta))
edges = list(edge1, edge2, edge3, edge4)

graph <- metric_graph$new(edges = edges)

node <- cbind(c(1,2), c(0,0))

graph$compute_geodist_PtE(node)

(But I have to admit, for this specific issue, might not be easy to create a simple and small example without knowing the cause)

[LLJiang01]

OKOK. I see :)