Split polyline with polyline

Question

I want to split line with line. (I want do like QGIS algorithm "split lines with lines)

function "gsection" in a package of stplanr do this.

library(stplanr)
data(routes_fast)
result <- gsection(routes_fast)
class(result)

But the function return SpatialLines class.

I want get SpatialLiensdataframe class,and keep "ID" etc.

What should I do?

Answer 1

I had to remind myself what gsection() does despite packaging it up into stplanr (the orginal code was written by Barry Rowlingson). It is used primarily in my work as a helper function for overline() but I decided to export it in case it's of use/interest to others. Great to see it is!

The function does not return data for a reason: individual segments have different numbers of overlapping routes.

However, it is useful to be able to query data from which the segments come, so let's work through some code, building on your reproducible example, to see what's going on:

library(stplanr)

## Loading required package: sp

length(routes_fast) # too many to visualise segments

## [1] 42

r = routes_fast[3:4,] # take 2 lines to see what's going on s = gsection(r) # split into overlapping sections class(r) # has data, as you say

## [1] "SpatialLinesDataFrame"
## attr(,"package")
## [1] "sp"

class(s) # does not have data!

## [1] "SpatialLines"
## attr(,"package")
## [1] "sp"

length(r) # 2 lines, as expected

## [1] 2

length(s) # 3 segments with same number of overlaps

## [1] 3

As you can see from the output of the above code chunk, there are more segments than there are routes. So surely each segment can be allocated it's own route? No.

This is illustrated below. The 3rd line from the resulting segments s (coloured grey) is the result of the overlap between both lines in r. So what data values would you expect it to have?

library(tmap) # for awesome plotting abilities qtm(routes_fast[3:4,], line.lwd = 20, line.alpha = 0.3) + qtm(routes_fast[3,], line.lwd = 5) + qtm(s[1,], line.col = "white") + qtm(s[2,], line.col = "black") + qtm(s[3,], line.col = "grey", line.lwd = 2)

There are different ways to answer this question. The default way in sp::over() is to take the first overlap. But this is not what we want as over()eturns a match even if the lines touch but do not have any shared distance (take a look inside the results to see what I mean):

result_data = over(x = s, y = r) result_data

##      plan        start           finish length time waypoint
## 1 fastest Gledhow Lane Harehills Avenue   2241  475       43
## 2 fastest Gledhow Lane Harehills Avenue   2241  475       43
## 3 fastest Gledhow Lane Harehills Avenue   2241  475       43

result_list = over(x = s, y = r, returnList = T)

result_dataeturns the first matching row from data in lines touching each segment - in this case that's simply routes_fast@data[3,]epeated 3 times, not very useful!

Assuming you're happy with the first match of lines that actually share lengths, you could use the (undocumented) minDimension argument of over(), described in vignette("over"):

over(x = s, y = r, minDimension = 1)

##      plan        start           finish length time waypoint
## 1 fastest Gledhow Lane Harehills Avenue   2241  475       43
## 2 fastest Gledhow Lane      Ekota Place   1864  270       37
## 3 fastest Gledhow Lane Harehills Avenue   2241  475       43

I think adding an argument return_data to the function would useful and plan to do so before the next release of stplanr. It should probably say something about how many overlapping lines each segment has as an additional output.

Many thanks for your instigating these investigations in any case: very useful.