Create line_merge function about stplanr HOT 6 CLOSED

Update: the following code is moving towards a solution:

# Aim: prototype geographical joining function

library(stplanr)

# Create data in projected CRS with units of m
rf = routes_fast[5:1,] # wrong order
l = flowlines[1:5,]
proj4string(rf) = CRS("+init=epsg:4326")
rf = spTransform(rf, CRS("+init=epsg:27700"))
l = spTransform(l, CRS("+init=epsg:27700"))

# Test mismatch
plot(l); plot(l[2,], lwd = 5, add = T)
plot(rf[2,], add = T) # does not match

# Merge based on ods
l_od_coords = line2df(l)
rf_od_coords = line2df(rf) 
sid1 = as.matrix(l_od_coords[-1])
sid2 = as.matrix(rf_od_coords[-1])

# install.packages("Hmisc")
m = Hmisc::find.matches(sid1, sid2, maxmatch = nrow(sid1),
                    tol = c(200, 200, 200, 200))
rf_new = rf[m$matches,]
l_new = l[1:length(l) %in% m$matches,]

# test 1
n = 1
plot(l[n,]); plot(rf[n,], add = T) # fail

```r
plot(l_new[n,]); plot(rf_new[n,], add = T) # works

# test 2
n = 3
plot(l_new[n,]); plot(rf_new[n,], add = T) # succeed

from stplanr.

Also thinking it should be called line_match now as it's matching not merging.

from stplanr.

Problem: what do you do about lines like l[1,] that have no match?

Suggestion: Print the line numbers in l1 with no match in l2.

from stplanr.

Do you just want to use origin and destination coordinates to match lines?

Normally when I'm matching lines I typically use Hausdorff Distance in PostGIS or the gDistance function in rgeos because it accounts for differences in the paths of the lines (rather than just origin-destination).

from stplanr.

Aha that sounds good, although in our dataset it is the orgin and destination specifically that we're matching to. Interesting that there doesn't seem to be an implementation for sp objects in R, based on a rudimentary web search yet there is clearly a GEOS algorithm for this. Any plans for this related to sf/rgeos that you know of @edzer? Wondering where the best place for an R implementation to live would be and guessing rgeos.

Ah OK looking at the docs it seems hausdorff=TRUE in gDistance() would do the trick, e.g.

library(stplanr)

## Loading required package: sp

r = routes_fast[1:3,]
plot(r,
     col = c("red", "yellow", "green"),
     lwd = 3:1)

rgeos::gDistance(r, r, byid = T, hausdorff = T)

##            1          2          3
## 1 0.00000000 0.01569878 0.01209752
## 2 0.01569878 0.00000000 0.02130649
## 3 0.01209752 0.02130649 0.00000000

But why is the difference between 1 and 2, which seem very similar, 0.016 whereas 1 and 3, which seem very different, is only 0.012?

from stplanr.

Looks pretty trivial to add to CPL_geos_binop

from stplanr.