normalize_height causes "NULL value passed as symbol address"

[Jean-Romain]

Discussed in #579

^{Originally posted by kulpojke May 5, 2022}
I am using the lidR package to do some tree segmentation on a set of 561 km² tiles using the LAScatalog engine and the futures library for parallelization. I have successfully created the DTM, but when I attempt to normalize I get an error.

I have read this thread already and it did not solve my problem.

I am running inside of a container (rocker/tidyverse:4.2) with the following:

R version 4.2.0 (2022-04-22)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.4 LTS

future_1.25.0
lidR_4.0.0

Here is the code:

# import libraries
library(lidR)
library(future)

plan(multisession)

# define some paramaters, currently using ft :(
chunk_size <-1000            # chunk size 
buff <- 50                   # buffer size
resolution <-2               # raster resolution

# (1) path to data 
path <- '/data'

# make las catalog
ctg <- readLAScatalog(path)

# set disk storage option by assigning a filename template to a function
opt_output_files(ctg) <- paste0(tempdir(), "/{XCENTER}_{YCENTER}_{ID}")

# create index files (lax) to index points within laz files. speeds up reads.
lidR:::catalog_laxindex(ctg)

# set the chunk buffer ofr ctg
opt_chunk_buffer(ctg) <- buff

# set the chunk size for ctg
opt_chunk_size(ctg) <- chunk_size

# make dtm
opt_output_files(ctg) <- paste0(tempdir(), "/{XCENTER}_{YCENTER}_{ID}_dtm")
dtm <- rasterize_terrain(ctg, resolution, tin())

# in my Rmd I plot the dtm here and it looks right

# normalise ctg
opt_output_files(ctg) <- paste0(tempdir(), "/{XCENTER}_{YCENTER}_{ID}_norm")
ctg_norm <- normalize_height(ctg, dtm)

During the normalization I get the following warning, then error:

Processing [-----------------]   0% (0/1152) eta:  ?sWarning: There are 3 points flagged 'withheld'.
Processing [-----------------]   0% (1/1152) eta:  4hError: NULL value passed as symbol address

After reading this thread. I tried disabling multisession and using set_lidr_threads,

# import libraries
library(lidR)
library(future)

#plan(multisession)
get_lidr_threads()

but the problem persists. Any insights would be greatly appreciated.

[Jean-Romain]

After reproducing with data of mine I get

Error : external pointer is not valid

Which is likely to be very similar to your error. SpatRaster are not serializable and I think this is what caused the error. However I do not have error with sequential processing because only one worker is implied and no serialization is involved. Are you sure you really disabled plan(multissesion). Did you start a fresh session or reset the strategy with plan(sequential)?

Edit: in another run I also go Error : NULL value passed as symbol address (with multisession)

---

Side note: plan(multisession) and s/get_lidr_threads() are doing very different things.

[Jean-Romain]

Fixed with a workaround. If user is using a parallel strategy SpatRaster are converted to RasterLayer. There is not way to work with SpatRaster

[Ben-Shamgochian]

Fixed with a workaround. If user is using a parallel strategy SpatRaster are converted to RasterLayer. There is not way to work with SpatRaster

Hi, I'm currently experiencing the same issue when attempting to normalize height and I don't quite understand this work around. Do you mind explaining in a little more depth?

Thanks!