IMPORTANT NOTE: The sfnetworks package is currently under active development, and breaking changes will be made in the coming months. The current version as described here is released as v0.1.0, but will soon be succeeded by a new version, with different classes and functions. Feel free to already open issues containing ideas of what functionalities you think should be included in the new package version (please label such issues with the feature label).
sfnetworks is an R package for creating, analyzing and visualizing spatial networks. The package can be seen as the connecting edge between the package sf, focused on analyzing spatial vector data, and the package igraph, which is the R implementation of the igraph library, focused on analyzing graph structures. The functionalities sfnetworks make use of several functions from these two packages. Furhtermore, the package enables to convert sf objects to igraph objects.
The package is still in development. Currently, it has the following functionalities
- Creating spatial networks
Spatial networks can be created using the function sfn_asnetwork, which is a generic function with S3 methods for a handful of classes. The returned value is always an object of class sfn_network.
The sfn_network class is a S3 class for spatial networks. An object of class sfn_network is a list with two elements:
- The edges element: an sf object with linestring geometry, representing the edges of the spatial network
- The nodes element: an sf object with point geometry, representing the nodes of the spatial network
The edges element always has five core columns:
- EdgeID: a unique identification number for each edge
- nodeID_source: the unique identification number of the source node of each edge
- nodeID_target: the unique identification number of the target node of each edge
- length: the length of each edge in meters
- geometry: the sf geometry list column
Besides the core columns, the edges element can have additional attribute columns (e.g. streetname, maximum speed, et cetera), which are copied from the input data.
The nodes element always has two core columns:
- nodeID: the unique identification number of each node
- geometry: the sf geometry list column
The sfn_asnetwork.sf method converts objects of class sf with a linestring geometry to objects of class sfn_network. The lines of the inputted object form the edges of the network. The nodes are the startpoints and endpoints of the lines. In networks, startpoints and/or endpoints are usually shared by two or more lines. These shared points will be one single node in the nodes element of the sfn_network object.
Let's show an example using the roxel dataset, which is an sf object representing the road network of the Roxel neighborhood in Münster, Germany. This dataset is attached as LazyData to the sfnetworks package, and can be loaded as follows.
# Loading the sfnetworks package
library(sfnetworks)
# Loading the tidyverse package
# To use for coding inside the vignette
library(tidyverse)
# Loading the roxel dataset
roxel## Simple feature collection with 851 features and 2 fields
## geometry type: LINESTRING
## dimension: XY
## bbox: xmin: 7.522594 ymin: 51.94151 xmax: 7.546705 ymax: 51.9612
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 10 features:
## name type geometry
## 1 Havixbecker Strasse residential LINESTRING (7.533722 51.955...
## 2 Pienersallee secondary LINESTRING (7.532442 51.954...
## 3 Schulte-Bernd-Strasse residential LINESTRING (7.532709 51.952...
## 4 <NA> path LINESTRING (7.540063 51.944...
## 5 Welsingheide residential LINESTRING (7.537673 51.947...
## 6 <NA> footway LINESTRING (7.543791 51.947...
## 7 <NA> footway LINESTRING (7.54012 51.9447...
## 8 <NA> path LINESTRING (7.53822 51.9454...
## 9 <NA> track LINESTRING (7.540063 51.944...
## 10 <NA> track LINESTRING (7.5424 51.94599...
The sf object - with linestring geometry - can then be converted to a spatial network of class sfn_network.
# Convert sf object to sfn_network object
sfn_asnetwork(roxel)## A spatial network with 851 edges and 701 nodes
##
## --> Edges element; an sf object with linestring geometry
## Simple feature collection with 851 features and 6 fields
## geometry type: LINESTRING
## dimension: XY
## bbox: xmin: 7.522594 ymin: 51.94151 xmax: 7.546705 ymax: 51.9612
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## name type edgeID nodeID_source nodeID_target
## 1 Havixbecker Strasse residential 1 316 305
## 2 Pienersallee secondary 2 279 263
## 3 Schulte-Bernd-Strasse residential 3 283 287
## 4 <NA> path 4 568 489
## 5 Welsingheide residential 5 463 455
## length geometry
## 1 28.85976 [m] LINESTRING (7.533722 51.955...
## 2 107.70459 [m] LINESTRING (7.532442 51.954...
## 3 54.36237 [m] LINESTRING (7.532709 51.952...
## 4 155.23052 [m] LINESTRING (7.540063 51.944...
## 5 208.66369 [m] LINESTRING (7.537673 51.947...
##
## --> Nodes element; an sf object with point geometry
## Simple feature collection with 701 features and 1 field
## geometry type: POINT
## dimension: XY
## bbox: xmin: 7.522622 ymin: 51.94151 xmax: 7.546705 ymax: 51.9612
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## nodeID geometry
## 1 316 POINT (7.533722 51.95556)
## 2 305 POINT (7.533461 51.95576)
## 3 279 POINT (7.532442 51.95422)
## 4 263 POINT (7.53209 51.95328)
## 5 283 POINT (7.532709 51.95209)
One who says 'mapping' and 'open source', automatically says 'OpenStreetMap'. The project allows free access to their map and all of the underlying map data, and is therefore extremly popular by those who analyze spatial data. The R implementation of automatically downloading and importingOpenStreetMap data is the package osmdata. Using this package, a user can download the OpenStreetMap data for an area inside a given bounding box, and also query for a specific type of data. Road infrastructure in OpenStreetMap has the tag highway. When working with spatial road networks, this is the data of interest.
Let's show an example of importing a road network with the osmdata package, using a bounding box for the neighborhood of Handorf in Münster, Germany, and then converting it to an spatial network of class sfn_network with the sfnetworks package.
# Load the osmdata package
library(osmdata)## Data (c) OpenStreetMap contributors, ODbL 1.0. http://www.openstreetmap.org/copyright
# Define the bounding box for Handorf
handorf_bbox = c(7.696474, 51.9816336, 7.7100898, 51.994452)
# Download and import the OpenStreetMap data
handorf = opq(bbox = handorf_bbox) %>%
add_osm_feature(key = 'highway') %>%
osmdata_sf()
handorf## Object of class 'osmdata' with:
## $bbox : 51.9816336,7.696474,51.994452,7.7100898
## $overpass_call : The call submitted to the overpass API
## $meta : metadata including timestamp and version numbers
## $osm_points : 'sf' Simple Features Collection with 1238 points
## $osm_lines : 'sf' Simple Features Collection with 285 linestrings
## $osm_polygons : 'sf' Simple Features Collection with 2 polygons
## $osm_multilines : 'sf' Simple Features Collection with 0 multilinestrings
## $osm_multipolygons : 'sf' Simple Features Collection with 0 multipolygons
As can be seen, the data comes as an object of class osmdata, which is a list containing several sf objects with different geometry types. The point geometries with the highway tag in OpenStreetMap are things like bus stops, zebra crossings, lampposts, stop signs, speed bumps, et cetera. In other words, they are not roads, neither road intersections. That is, they will not be included in the spatial network of class sfn_network. The sf object with the line geometry are all roads and pathes, and will form the edges element in the sfn_network object.However, squares, roundabouts and roads that have the same start as endpoint, are often stored as polygons. The sfn_asnetwork.osmdata method therefore first transforms the polygons of the osmdata object into lines, then exracts all the lines as an sf object with line geometry, and converts that to spatial network of class sfn_network.
# Convert osmdata object to sfn_network object
sfn_asnetwork(handorf)## A spatial network with 287 edges and 466 nodes
##
## --> Edges element; an sf object with linestring geometry
## Simple feature collection with 287 features and 37 fields
## geometry type: LINESTRING
## dimension: XY
## bbox: xmin: 7.692907 ymin: 51.98082 xmax: 7.716578 ymax: 51.99502
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## osm_id name access area bicycle bridge created_by
## 1 4863860 Dorbaumstraße <NA> <NA> <NA> <NA> <NA>
## 2 4863864 Handorfer Straße <NA> <NA> <NA> <NA> <NA>
## 3 23373682 Dorbaumstraße <NA> <NA> <NA> <NA> <NA>
## 4 23373691 <NA> <NA> <NA> <NA> <NA> JOSM
## 5 23373694 Kötterstraße <NA> <NA> <NA> <NA> <NA>
## cycleway description foot hgv highway lanes lanes.backward
## 1 <NA> <NA> <NA> <NA> tertiary <NA> <NA>
## 2 <NA> <NA> <NA> <NA> tertiary <NA> <NA>
## 3 <NA> <NA> <NA> <NA> tertiary <NA> <NA>
## 4 <NA> <NA> yes <NA> cycleway <NA> <NA>
## 5 <NA> <NA> <NA> <NA> residential <NA> <NA>
## lanes.forward layer lit maxspeed noexit oneway place_numbers ref
## 1 <NA> <NA> <NA> <NA> <NA> <NA> 1-29 K 16
## 2 <NA> <NA> <NA> <NA> <NA> <NA> 1-49 K 16
## 3 <NA> <NA> <NA> <NA> <NA> <NA> 30- K 16
## 4 <NA> <NA> <NA> <NA> <NA> <NA> <NA> <NA>
## 5 <NA> <NA> <NA> <NA> <NA> <NA> 1-63 <NA>
## segregated service smoothness source source.ref surface
## 1 <NA> <NA> <NA> survey: gps by bike <NA> <NA>
## 2 <NA> <NA> <NA> <NA> <NA> <NA>
## 3 <NA> <NA> <NA> survey: gps by bike <NA> <NA>
## 4 <NA> <NA> <NA> <NA> <NA> <NA>
## 5 <NA> <NA> <NA> <NA> <NA> <NA>
## tracktype tunnel turn.lanes.backward turn.lanes.forward width edgeID
## 1 <NA> <NA> <NA> <NA> <NA> 1
## 2 <NA> <NA> <NA> <NA> <NA> 2
## 3 <NA> <NA> <NA> <NA> <NA> 3
## 4 <NA> <NA> <NA> <NA> <NA> 4
## 5 <NA> <NA> <NA> <NA> <NA> 5
## nodeID_source nodeID_target length geometry
## 1 76 200 360.44650 [m] LINESTRING (7.698971 51.988...
## 2 76 65 179.47556 [m] LINESTRING (7.698971 51.988...
## 3 200 463 773.02914 [m] LINESTRING (7.702385 51.991...
## 4 199 200 4.73605 [m] LINESTRING (7.702332 51.991...
## 5 188 306 298.13865 [m] LINESTRING (7.701638 51.985...
##
## --> Nodes element; an sf object with point geometry
## Simple feature collection with 466 features and 1 field
## geometry type: POINT
## dimension: XY
## bbox: xmin: 7.692907 ymin: 51.9812 xmax: 7.716578 ymax: 51.99477
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## nodeID geometry
## 1 76 POINT (7.698971 51.98864)
## 2 200 POINT (7.702385 51.99106)
## 3 65 POINT (7.698562 51.98707)
## 4 463 POINT (7.711681 51.99474)
## 5 199 POINT (7.702332 51.99109)
OpenStreetMap data comes with a lot of attribute columns. For users who are not interested in these attributes, but only in the geometry of the network, the sfn_asnetwork function has a parameter attributes, which is by default set to TRUE, but can be changed to FALSE. If that is the case, all attribute columns from the input object will be dropped, and the resulting sfn_network object will only contain the core columns.
# Convert osmdata object to sfn_network, with attrbutes set to FALSE
sfn_asnetwork(handorf, attributes = FALSE)## A spatial network with 287 edges and 466 nodes
##
## --> Edges element; an sf object with linestring geometry
## Simple feature collection with 287 features and 4 fields
## geometry type: LINESTRING
## dimension: XY
## bbox: xmin: 7.692907 ymin: 51.98082 xmax: 7.716578 ymax: 51.99502
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## edgeID nodeID_source nodeID_target length
## 1 1 76 200 360.44650 [m]
## 2 2 76 65 179.47556 [m]
## 3 3 200 463 773.02914 [m]
## 4 4 199 200 4.73605 [m]
## 5 5 188 306 298.13865 [m]
## geometry
## 1 LINESTRING (7.698971 51.988...
## 2 LINESTRING (7.698971 51.988...
## 3 LINESTRING (7.702385 51.991...
## 4 LINESTRING (7.702332 51.991...
## 5 LINESTRING (7.701638 51.985...
##
## --> Nodes element; an sf object with point geometry
## Simple feature collection with 466 features and 1 field
## geometry type: POINT
## dimension: XY
## bbox: xmin: 7.692907 ymin: 51.9812 xmax: 7.716578 ymax: 51.99477
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## nodeID geometry
## 1 76 POINT (7.698971 51.98864)
## 2 200 POINT (7.702385 51.99106)
## 3 65 POINT (7.698562 51.98707)
## 4 463 POINT (7.711681 51.99474)
## 5 199 POINT (7.702332 51.99109)
Important to note here that the quality of OpenStreetMap data is not always very high. It often happens that lines are not cut at intersections, lines don't touch when they should, et cetera. The best way is to manually fix these topology errors. However, in large networks this will be a very labour intensive operation. The v.clean (see here) tool in GRASS GIS is then an alternative. It is highly recommended to first run this tool on the OpenStreetMap data before analysing it. This cleaning operation was also performed on the roxel dataset that is attached to this package.
Besides the methods for objects of class sf and objects of class osmdata, the sfn_asnetwork generic function also has a method for objects of class SpatialLinesDataFrame, which is the class for spatial data with linestring geometry in the sp package. Finally, there exists a method for objects of class sfn_route, which is a class in the sfnetworks package for routes in a spatial network. This class will be introduced later in this vignette.
- Analyzing spatial networks
The sfnetworks has a few built-in functions to analyze spatial networks. For further analysis, there is the possibility to convert the network of class sfn_network into a graph object of class igraph, after which the vast range of analytic functions from the igraph package can be used.
With the function sfn_shortestpath, the shortest path between two nodes in a spatial network can be calculated. The function returns an object of class sfn_route, a S3 class specially designed for routes in a spatial network. A route is different from a network in the sense that the nodes and edges in a route have a specific order. In an object of class sfn_route, this order is defined in the route element. The route element consists of one feature that contains both the nodes and edges in a geometrycollection (class sfc_GEOMETRYCOLLECTION), in the right order. The features in the edges element and/or the nodes element can now be ordered in any way the user wants (e.g. from longest edge to shortest edge), without disturbing the right order of the route. Finally, an object of class sfn_route also contains a weights element, representing the total weight (i.e. sum of the edges weights) of the route.
To summarize, an object of class sfn_route is a list with four elements:
- The route element: an sf object of one feature with a geometrycollection geometry, containing both the nodes and edges of the route, in the right order
- The edges element: an sf object with linestring geometry, representing the edges of which the route is made up
- The nodes element: an sf object with point geometry, representing the nodes of which the route is made up
- The weights element: a value representing the total weight of the route (i.e. sum of the edges weights)
Just as in the class sfn_network, the edges elements has five core columns (edgeID, nodeID_source, nodeID_target, length and geometry), with optionally additional attributes columns. The nodes element has two core columns (nodeID, geometry).
As inputs, the sfn_shortestpath function takes the spatial network of class sfn_network (parameter x), the nodeID of the node where the shortest path should calculated from (parameter start) and the nodeID of the node where the shortest path should lead to (parameter end).
Additionally, the weight of each edge should be given (parameter weights). This can be done by specifying the name of the column in the edges element of the inputted sfn_network object which values should be used as edge weights. By default, the length column - one of the core columns of the edges element - is used for the edge weights. That mean that literally the shortest path is calculated. However, if someone for example has a column in their data with the duration of each edge, this column could be used for edge weights, and the fastest path is calculated. Another example could be a column with the average slope of each edge, which would lead to calculating the flattest path.
Let's show an example with the roxel dataset. Below, the shortest path between the node with nodeID 195 (in the southwest of Münster Roxel) and the node with nodeID 676 (in the northeast of Münster Roxel). For the weights, the default value ('length') will be used.
# Calculate the shortest path between node 195 and node 676
roxel %>%
sfn_asnetwork() %>%
sfn_shortestpath(start = 195, end = 676)## A route with 43 edges and 44 nodes
##
## --> Route element; a sf geometrycollection
## Simple feature collection with 1 feature and 0 fields
## geometry type: GEOMETRYCOLLECTION
## dimension: XY
## bbox: xmin: 7.525638 ymin: 51.94252 xmax: 7.543782 ymax: 51.95921
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## geometry
## 1 GEOMETRYCOLLECTION (POINT (...
##
## --> Edges element; an sf object with linestring geometry
## Simple feature collection with 43 features and 6 fields
## geometry type: LINESTRING
## dimension: XY
## bbox: xmin: 7.525638 ymin: 51.94252 xmax: 7.543782 ymax: 51.95921
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## name type edgeID nodeID_source nodeID_target
## 1 Lise-Meitner-Strasse residential 87 159 195
## 2 Lise-Meitner-Strasse residential 550 152 159
## 3 Lise-Meitner-Strasse residential 549 146 152
## 4 Lise-Meitner-Strasse residential 552 139 146
## 5 Lise-Meitner-Strasse residential 551 124 139
## length geometry
## 1 112.53259 [m] LINESTRING (7.529177 51.942...
## 2 46.89715 [m] LINESTRING (7.528951 51.942...
## 3 36.21925 [m] LINESTRING (7.528777 51.943...
## 4 54.09753 [m] LINESTRING (7.528463 51.943...
## 5 63.72230 [m] LINESTRING (7.528095 51.944...
##
## --> Nodes element; an sf object with point geometry
## Simple feature collection with 44 features and 1 field
## geometry type: POINT
## dimension: XY
## bbox: xmin: 7.52564 ymin: 51.94252 xmax: 7.543782 ymax: 51.95921
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## First 5 features:
## nodeID geometry
## 1 195 POINT (7.530772 51.94274)
## 2 159 POINT (7.529177 51.94252)
## 3 152 POINT (7.528951 51.94292)
## 4 146 POINT (7.528777 51.94322)
## 5 139 POINT (7.528463 51.94367)
##
## --> Weight element; the total weight of the route
## 2744.92736183062
Instead of passing a column name to the weights parameter, it is also possible to give a numeric vector of the same length as the edges element in the inputted network. The values of this vector will then be used as weights. Furthermore, one can define if the network should be considered a directed network (parameter directed). That is, if each edge can only be travelled from source to target, and not in the opposite direction. By default, this is set to FALSE, meaning that each edge can be travelled in both directions.
Let's show another example, that may not make a lot of sense, but does show how the functions works. The parameter weights will now be a numeric vector of random numbers, and the parameter directed will be set to TRUE. Keep in mind that setting directed to TRUE while working with a network that is not designed as a directional network, like the roxel dataset, is never a good idea. Here, it is just done to show the functionalities of the sfn_shortestpath function.
# Create a vector with random numbers
# Length equal to the number of edges in the sfn_network
vec = roxel %>%
sfn_asnetwork() %>%
pluck('edges') %>%
nrow() %>%
runif()
head(vec)## [1] 0.02358652 0.27596723 0.32328779 0.48491675 0.23858638 0.73370355
# Calculate the shortest path between node 90 and node 51, with directed to FALSE
roxel %>%
sfn_asnetwork() %>%
sfn_shortestpath(start = 90, end = 51, weights = vec, directed = TRUE)## A route with 2 edges and 3 nodes
##
## --> Route element; a sf geometrycollection
## Simple feature collection with 1 feature and 0 fields
## geometry type: GEOMETRYCOLLECTION
## dimension: XY
## bbox: xmin: 7.525749 ymin: 51.94671 xmax: 7.527087 ymax: 51.94805
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## geometry
## 1 GEOMETRYCOLLECTION (POINT (...
##
## --> Edges element; an sf object with linestring geometry
## Simple feature collection with 2 features and 6 fields
## geometry type: LINESTRING
## dimension: XY
## bbox: xmin: 7.525749 ymin: 51.94671 xmax: 7.527087 ymax: 51.94805
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## name type edgeID nodeID_source nodeID_target length
## 1 Pienersallee secondary 830 90 56 165.271693 [m]
## 2 Pienersallee secondary 831 56 51 9.464045 [m]
## geometry
## 1 LINESTRING (7.527087 51.948...
## 2 LINESTRING (7.525816 51.946...
##
## --> Nodes element; an sf object with point geometry
## Simple feature collection with 3 features and 1 field
## geometry type: POINT
## dimension: XY
## bbox: xmin: 7.525749 ymin: 51.94671 xmax: 7.527087 ymax: 51.94805
## epsg (SRID): 4326
## proj4string: +proj=longlat +datum=WGS84 +no_defs
## nodeID geometry
## 1 90 POINT (7.527087 51.94805)
## 2 56 POINT (7.525816 51.94679)
## 3 51 POINT (7.525749 51.94671)
##
## --> Weight element; the total weight of the route
## 0.893741030711681
The function sfn_length is a generic function with S3 methods for objects of class sfn_network or class sfn_route. It uses the st_length function from the sf package to calculate the total length of respectively a spatial network or a route in a spatial network.
# Calculate the total length of a spatial network
roxel %>%
sfn_asnetwork() %>%
sfn_length()## 50522.78 [m]
# Calculate the total length of a route in a spatial network
roxel %>%
sfn_asnetwork() %>%
sfn_shortestpath(start = 195, end = 676) %>%
sfn_length()## 2744.927 [m]
The function sfn_length is a generic function with S3 methods for objects of class sfn_network or class sfn_route. It uses the st_transform function from the sf package to transform or convert the coordinate reference system of a spatial network or a route in a spatial network. When doing this, the st_transform function is applied to all the geographic elements of the input object (nodes and edges for an sfn_network object and route, nodes and edges for an sfn_route object).
# Original sfn_network
roxel %>%
sfn_asnetwork() %>%
summary()## value
## Number of edges 851
## Number of nodes 701
## Total length (m) 50523
## CRS EPSG 4326
# Transform from WGS84 (EPSG 4326) to Gauss-Kruger Zone 2 (EPSG 31466)
roxel %>%
sfn_asnetwork() %>%
sfn_transform(crs = 31466) %>%
summary()## value
## Number of edges 851
## Number of nodes 701
## Total length (m) 50530
## CRS EPSG 31466
The function sfn_network2graph converts a spatial network of class sfn_network to a graph object of class igraph. After this conversion, one can apply the wide range of functions for graph analysis from the igraph package. When converting to an igraph object, the geometry of the edges and nodes is not preserved. The resulting graph is simply a set of weighted connections between unlocated nodes. By default, this weights are the lengths of the edges. However, just as in the sfn_shortestpath function - that actually uses a conversion to a graph inside the function -, the user can set any column from the edges element as weight column, or give a numeric vector of the same length as the edges element.
# Convert sfn_network to igraph object
roxel %>%
sfn_asnetwork() %>%
sfn_network2graph()## IGRAPH 2f92861 UNW- 701 851 --
## + attr: name (v/c), weight (e/n)
## + edges from 2f92861 (vertex names):
## [1] 316--305 279--263 283--287 568--489 463--455 677--674 572--562
## [8] 489--455 568--670 646--615 364--369 357--356 361--332 681--701
## [15] 273--303 512--565 561--531 183--169 206--188 262--219 290--274
## [22] 322--319 205--241 418--432 432--378 611--632 628--643 673--694
## [29] 352--365 314--281 81 --87 7 --13 86 --62 218--270 596--617
## [36] 112--233 32 --57 37 --18 14 --16 24 --12 21 --8 19 --15
## [43] 10 --4 18 --17 16 --26 12 --9 89 --106 85 --99 58 --77
## [50] 117--141 163--157 398--396 411--425 233--235 194--202 172--178
## + ... omitted several edges
Let's show some examples of measures that can now be calculated with the igraph package. For a full description of the functions in this package, please see its website.
# Load the igraph package
library(igraph)
# Create the graph
g = roxel %>%
sfn_asnetwork() %>%
sfn_network2graph()
# Calculate network diameter
# This is the longest shortest path between two nodes in the network
igraph::diameter(g, directed = FALSE)## [1] 2876.518
# Calculate the degree centrality of the nodes
# This is the number of edges that a specific node has
# Show only the first 6 results
igraph::degree(g, v = V(g), mode = 'all') %>% head()## 316 305 279 263 283 287
## 4 3 4 3 3 4
# Calculate the betweenness centrality of the nodes
# This is the number of shortest paths that pass through a node
# Show only the first 6 results
igraph::betweenness(g, v = V(g), directed = FALSE) %>% head()## 316 305 279 263 283 287
## 13809 9777 35240 31745 7174 9081
- Visualization of a spatial network
The sfnetworks package contains plot methods for both of its classes (i.e. class sfn_network and class sfn_route).
The plot method for objects of class sfn_network plots the geometry of the edges with plot.sf and adds the geometry of the nodes on top of it.
# Plot sfn_network
roxel %>%
sfn_asnetwork() %>%
plot()
The user can change several parameters regarding the layout of the plot, for example the color in which the edges are plot, the linewidth of the edges, the color in which the nodes are plot and the symbol type and size of the nodes. Additional parameters are passed on to the plot function for sf objects.
# Plot sfn_network in different style
roxel %>%
sfn_asnetwork() %>%
plot(
col_edges = 'black',
lwd = 2,
col_nodes = 'red',
pch = 21,
cex = 1.1,
main = "Road Network of Münster Roxel"
) 
It is also possible to color the edges by an attribute from the edges element of the network. To do so, the name of the column needs to be assigned to the attribute parameter. Then, only the edges element will be plotted, as an sf object with one attribute column.
# Plot sfn_network colored by attribute
roxel %>%
sfn_asnetwork() %>%
plot(
lwd = 2,
attribute = 'type',
pal = rainbow,
key.size = lcm(3.5),
main = "Road types in Münster Roxel"
)
The plot method for objects of class sfn_route plots the geometry of the edges with plot.sf and adds the geometry of the nodes on top of it.
# Plot sfn_route
roxel %>%
sfn_asnetwork() %>%
sfn_shortestpath(start = 195, end = 676) %>%
plot()
This is not very informative though. Therefore, the plot methods for objects of class sfn_route has the parameter network. By default it is set to NULL, but the user can assign it the spatial network of class sfn_network on which the shortest path is created. Then, the network will be plotted in grayscale on the background.
# Plot sfn_route with network on background
roxel %>%
sfn_asnetwork() %>%
sfn_shortestpath(start = 195, end = 676) %>%
plot(
col_edges = "Red",
lwd = 2,
network = roxel %>% sfn_asnetwork()
)
- Future work
The package sfnetworks is still in development. The following functionalities are planned to be added to the package:
- Add a column specifying the direction of each edge to the edges element of an object of class
sfn_network. This could for example be values of 1 for an edge that can only be travelled from startpoint to endpoint, -1 for an edge that can only be travelled from endpoint to startpoint, and 0 for an edge that can be travelled in both ways. The user can assign this direction values to each edge. In the case of anosmdataobject as input, lines with aonewaylabel will be assigned a value of 1 for direction automatically. - Give the option to not only assign weights to edges, but also to nodes. For example, when calculating shortest paths based on travel time, he may have information about average waiting time on a crossing. This information can then be added as weights to the nodes, and taken into account when calculating shortest paths.
- Give the option to calculate shortest paths from one node to several other nodes. Now, it is only possible to calculate shortest paths between two nodes at a time.
- Import the v.clean function for cleaning vector topology from GRASS GIS, making use of the
rgrass7package, the bridge between R and GRASS GIS. - Enable users to not only calculate clusters of nodes based on betweenness centrality (which can already be done after converting the
sfn_networkobject to anigraphobject) but also visualize them (which can not be done yet since the convertedigraphobject has no geometry).
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent