wrapper for algorithm logic

[mdsumner]

Here's a trip_sda filter function and example that works with the same interface as argosfilter. We necessarily have to remove locations that don't make sense (out of order, duplicate times) - internally argosfilter swallows these with an eps values.

  library(argosfilter)
data(seal)
## fix the data first (there is only one trip here, no multiples in argosfilter)
seal <- seal[order(seal$dtime), ]
bad <- c(FALSE, !diff(unclass(seal$dtime)) > 0)
seal <- seal[!bad, ]

# The algorithm first removes all locations with
# location class Z (-9), which are the points for which the location
# process failed. 
# (there aren't any Zs here, but for example) 
bad2 <- grepl("z", lc, ignore.case = TRUE)
#> Error in grepl("z", lc, ignore.case = TRUE): object 'lc' not found
seal <- seal[!bad2, ]
#> Error in `[.data.frame`(seal, !bad2, ): object 'bad2' not found
lat<-seal$lat
lon<-seal$lon
dtime<-seal$dtime
lc<-seal$lc
# filter by speed only
mfilter<-vmask(lat,lon,dtime,2)
# filter data using sdafilter
cfilter<-sdafilter(lat, lon, dtime, lc)

trip_sda <- function(lat, lon, dtime, vmax = 2, ang = c(15, 25), 
                     distlim = c(2500,5000), id = "1") {
  ## basic checks
  badlat <- any(is.na(lat))
  badlon <- any(is.na(lon))
  badtime <- any(is.na(dtime))
  baddt <- any(!diff(unclass(dtime)) > 0)
  if (badlat) stop("missing lat value/s")
  if (badlon) stop("missing lon value/s")
  if (badtime) stop("missing dtime value/s")
  if (baddt) stop("dtime not in order and/or contains duplicates")

  if (length(id) > 1) {
    if (!length(id) == length(lat)) stop("length of id must be 1 (for a single trip) or the same length as lat, lon, dtime (defining multiple trips)")
  }
  df <- data.frame(lon = lon, lat = lat, time = dtime, id = id)
  sp::coordinates(df) <- c("lon", "lat")
  sp::proj4string(df) <- "+init=epsg:4326"
  df <- trip::trip(df, c("time", "id"))
  ## trip works in km and km/h (don't judge me)
  trip::sda(df, smax = vmax * 3.6, ang = ang, distlim = distlim/1000)
}

sda <- trip_sda(lat, lon, dtime, vmax = 2, id = 1)
op <- par(mfrow = c(2, 1))
plot(dtime, lon, cex = 0.4, pch = 19);lines(dtime[sda], lon[sda], col = "firebrick", lwd = 5);lines(dtime[cfilter == "not"], lon[cfilter == "not"])
plot(dtime, lat, cex = 0.4, pch = 19);lines(dtime[sda], lat[sda], col = "firebrick", lwd = 5);lines(dtime[cfilter == "not"], lat[cfilter == "not"])

par(op)

^{Created on 2019-04-03 by the reprex package (v0.2.1)}

More ideal would be to remove any algorithm logic from trip (speed, angle filters, time spent, dupe time fudges, the sims, etc and make then complietely generic in a function with no formal classes, with lon, lat, time, [group] function inputs that any package could use.

[mdsumner]

Ping @jmlondon fyi

[jmlondon]

Thanks for this, @mdsumner. Even more ideal (at least for my own selfish workflow) would be an sda filter that I could pass an sf points object (in any projection) and get back an sf object with records to remove indicated.

[mdsumner]

Use st_cordinates to get the raw x,y values. I'd be interested in seeing the way you call argosfilter now