mpadge
commented
2 years ago Thanks @demcortillas. If i understand correctly, your question is just how to extract the isodistance profile as a polygon? If you read the docs for ?dodgr_isodists, you'll see that the return value is described as:
https://github.com/ATFutures/dodgr/blob/fa79c6bb559f7811347dce3d94f207d6ef5943f9/R/iso.R#L16-L20
I think that means that return value is actually what you want, and you don't need to do any of the sf stuff in your example above at all. Here's an example of how to directly plot an isodistance contour, starting with your initial code.
library (dodgr)
osm <- readRDS ("<local>/<osm>/<network>")
graph1 <- weight_streetnet(osm, wt_profile = 1)
#> Loading required namespace: geodist
#> Loading required namespace: dplyr
graph1 <- graph1[graph1$component==1,]
graph1 <- dodgr_contract_graph(graph1)
vertices <- dodgr_vertices(graph1)You then use the following code to select a single point and map it onto the network:
# vertices <-st_as_sf(vertices, coords = c("x", "y"), crs = 4326, agr = "constant")
# p_real <- st_sfc(st_point(c(-73.0511859291954,-36.82369513651161)),crs = 4326)
# I <- st_nearest_feature(p_real,vertices)
# I_net <- vertices[I,]$idBut that is not needed as dodgr maps coordinates to nearest network point anyway, so instead you can just submit p_real as your single from value. For the above network I'll fake a value as:
set.seed (1)
p_real <- c (x = min (vertices$x) + runif (1) * diff (range (vertices$x)),
y = min (vertices$y) + runif (1) * diff (range (vertices$y)))The call to isodists is then:
# iso_dist <- dodgr_isodists(graph1, from = I_net , dlim = 10000, heap = "BHeap")
iso_dist <- dodgr_isodists (graph1, from = p, dlim = 10000)That result can be used to plot a contour directly, for example using these lines:
plot (iso_dist$x, iso_dist$y, type = "l")
points (p [1], p [2], pch = 19, cex = 2)
But it seems from the image you included above that what you're actually trying to do with the following lines is work out every possible path within that isodistance contour, right?
fmat <- matrix (1, nrow = 1, ncol = dim(iso_dist)[1])
f <- dodgr_flows_aggregate (graph1,
from = iso_dist$from,
to = iso_dist$id,
flows = fmat) %>%
merge_directed_graph() %>%
dodgr_to_sf()
plot(f$geometry)The problem is that isodistance contours in dodgr, in indeed as a general principle of what isodistances actually are, are the collection of points which are maximally far from the defined from point yet just within the specified isodistance. So each point is a single network point which is either (1) terminal, or (2) the final point at d < dlim that connects with a subsequent point which is > dlim. In finding that set of "iso-vertices", the algorithms collects candidate points, for example dA. If that point connects to a subsequent point, dB, with dist(from, dB) > dist(from, dA), then dA is dropped from the candidate list and replaced by dB. The result is that you end up with far fewer of those points than the total number of points actually traversed in searching for them. So you can not use an isodistance algorithm to generate the kind of full network coverage it seems like you're wanting.
But what (i think?) you seem to want nevetheless seems reasonable, so i'll have a go at generating it.
demcortillas
Hello everyone, I hope you are all well
My problem is roughly to build a dodgr_isodistance from a point and convert it into a polygon.
When I plot the lines, the result only shows the "main routes" or what I would expect to be "directed" but not the streets that are in a normal CBD (many streets in a small area). Something like in the following image:
My question is the following: How can I occupy the entire network without segregating via wt_profile? A function that does apply to what I'm doing is:
But that function takes a long time, so I am discarding that idea. Thank you very much for your cooperation.