dodgr_times {dodgr}

Calculate matrix of pair-wise travel times between points.

Description

Calculate matrix of pair-wise travel times between points.

Usage

dodgr_times(graph, from = NULL, to = NULL, shortest = FALSE, heap = "BHeap")

Arguments

graph

data.frame or equivalent object representing the network graph (see Notes). For dodgr street networks, this may be a network derived from either sf or silicate ("sc") data, generated with weight_streetnet. Note, however, that networks derived from sf data will generally not produce reliable estimates of times. Accurate estimates can only be guaranteed by using networks derived from silicate ("sc") data.

from

Vector or matrix of points from which route distances are to be calculated (see Notes)

to

Vector or matrix of points to which route distances are to be calculated (see Notes)

shortest

If TRUE, calculate times along the shortest rather than fastest paths.

heap

Type of heap to use in priority queue. Options include Fibonacci Heap (default; FHeap), Binary Heap (BHeap), ⁠Trinomial Heap (⁠TriHeap⁠), Extended Trinomial Heap (⁠TriHeapExt⁠, and 2-3 Heap (⁠Heap23').

Value

square matrix of distances between nodes

Note

graph must minimally contain three columns of from, to, dist. If an additional column named weight or wt is present, shortest paths are calculated according to values specified in that column; otherwise according to dist values. Either way, final distances between from and to points are calculated by default according to values of dist. That is, paths between any pair of points will be calculated according to the minimal total sum of weight values (if present), while reported distances will be total sums of dist values.

For street networks produced with weight_streetnet, distances may also be calculated along the fastest routes with the shortest = FALSE option. Graphs must in this case have columns of time and time_weighted. Note that the fastest routes will only be approximate when derived from sf-format data generated with the osmdata function osmdata_sf(), and will be much more accurate when derived from sc-format data generated with osmdata_sc(). See weight_streetnet for details.

The from and to columns of graph may be either single columns of numeric or character values specifying the numbers or names of graph vertices, or combinations to two columns specifying geographical (longitude and latitude) coordinates. In the latter case, almost any sensible combination of names will be accepted (for example, ⁠fromx, fromy⁠, ⁠from_x, from_y⁠, or ⁠fr_lat, fr_lon⁠.)

from and to values can be either two-column matrices or equivalent of longitude and latitude coordinates, or else single columns precisely matching node numbers or names given in graph$from or graph$to. If to is NULL, pairwise distances are calculated from all from points to all other nodes in graph. If both from and to are NULL, pairwise distances are calculated between all nodes in graph.

Calculations in parallel (parallel = TRUE) ought very generally be advantageous. For small graphs, calculating distances in parallel is likely to offer relatively little gain in speed, but increases from parallel computation will generally markedly increase with increasing graph sizes. By default, parallel computation uses the maximal number of available cores or threads. This number can be reduced by specifying a value via ⁠RcppParallel::setThreadOptions (numThreads = <desired_number>)⁠. Parallel calculations are, however, not able to be interrupted (for example, by Ctrl-C), and can only be stopped by killing the R process.

See Also

Other distances: dodgr_distances(), dodgr_dists(), dodgr_dists_categorical(), dodgr_dists_nearest(), dodgr_flows_aggregate(), dodgr_flows_disperse(), dodgr_flows_si(), dodgr_isochrones(), dodgr_isodists(), dodgr_isoverts(), dodgr_paths()

Examples

# A simple graph
graph <- data.frame (
    from = c ("A", "B", "B", "B", "C", "C", "D", "D"),
    to = c ("B", "A", "C", "D", "B", "D", "C", "A"),
    d = c (1, 2, 1, 3, 2, 1, 2, 1)
)
dodgr_dists (graph)

# A larger example from the included [hampi()] data.
graph <- weight_streetnet (hampi)
from <- sample (graph$from_id, size = 100)
to <- sample (graph$to_id, size = 50)
d <- dodgr_dists (graph, from = from, to = to)
# d is a 100-by-50 matrix of distances between `from` and `to`

## Not run: 
# a more complex street network example, thanks to @chrijo; see
# https://github.com/UrbanAnalyst/dodgr/issues/47

xy <- rbind (
    c (7.005994, 51.45774), # limbeckerplatz 1 essen germany
    c (7.012874, 51.45041)
) # hauptbahnhof essen germany
xy <- data.frame (lon = xy [, 1], lat = xy [, 2])
essen <- dodgr_streetnet (pts = xy, expand = 0.2, quiet = FALSE)
graph <- weight_streetnet (essen, wt_profile = "foot")
d <- dodgr_dists (graph, from = xy, to = xy)
# First reason why this does not work is because the graph has multiple,
# disconnected components.
table (graph$component)
# reduce to largest connected component, which is always number 1
graph <- graph [which (graph$component == 1), ]
d <- dodgr_dists (graph, from = xy, to = xy)
# should work, but even then note that
table (essen$level)
# There are parts of the network on different building levels (because of
# shopping malls and the like). These may or may not be connected, so it may
# be necessary to filter out particular levels
index <- which (!(essen$level == "-1" | essen$level == "1")) # for example
library (sf) # needed for following sub-select operation
essen <- essen [index, ]
graph <- weight_streetnet (essen, wt_profile = "foot")
graph <- graph [which (graph$component == 1), ]
d <- dodgr_dists (graph, from = xy, to = xy)

## End(Not run)