catalog_apply() & readTLSLAS() gives warnings (minor issue)

[zoeschindler]

Hello,
I am sorry if I should have posted this in the discussion, I was not sure where to post this. Right now, I was writing some functions for catalog_apply() and somehow, when I use readTLSLAS() instead of readLAS(), I get warnings:

1: Argument 'select' is not used with a LAScluster. Use opt_select() with the LAScatalog instead.
2: Argument 'filter' is not used with a LAScluster. Use opt_filter() with the LAScatalog instead.

As soon as I switch back to readLAS() (I also tried readALSLAS(), it also gives warnings), everything works without a warning. I am not 100% sure, but the output looks similar to me. But I also did not give any filter or select arguments to the readTLS() function. I guess it's just a minor issue, it just took me some time to realize what the problem was. I just wanted to inform you guys, I am not too sure if it's relevant at all. I tried this on an example too (just switching the function how to read in the data) and it gives the same warnings. Maybe this is also intentional or I should ignore this because I already read in my catalog with readTLSLAScatalog()?

[Jean-Romain]

Please show your code so I can spot where and why you have this. This looks like a legit bug report.

[zoeschindler]

Okay, this is the code from the example page. The issue is in the filter_noise.LAScluster() function.

filter_noise.LAS = function(las, sensitivity)
{
  p95 <- grid_metrics(las, ~quantile(Z, probs = 0.95), 10)
  las <- merge_spatial(las, p95, "p95")
  las <- filter_poi(las, Z < p95*sensitivity)
  las$p95 <- NULL
  return(las)
}

filter_noise.LAScluster = function(las, sensitivity)
{
  las <- readTLSLAS(las)  # HERE, WHEN CHANGING THIS
  if (is.empty(las)) return(NULL)
  
  las <- filter_noise.LAS(las, sensitivity)
  las <- filter_poi(las, buffer == 0)
  return(las)
}

filter_noise.LAScatalog = function(las, sensitivity)
{
  opt_select(las) <-  "*"
 
  options <- list(
    need_output_file = TRUE,
    need_buffer = TRUE,
    automerge = TRUE)
  
  output  <- catalog_apply(las, filter_noise.LAScluster, sensitivity = sensitivity, .options = options)
  return(output)
}

path_points <- "H:/data/testing.las"
points_name <- "banana"

ctg <- readTLSLAScatalog(path_points)
opt_chunk_buffer(ctg) <- 0.25
opt_chunk_size(ctg) <- 20
opt_output_files(ctg) <- paste0(dirname(path_points), "/filtered/", points_name, "_{ID}")
filter_noise.LAScatalog(ctg, 1.2)

[Jean-Romain]

Ho ok I got it. Use readLAS() in the LAScluster version.

The idea behind the LAScatalog processing engine is to provide tools for users. And you never know if users are processing ALS TLS or whatever. So, as a developer you should never call the ALS or TLS versions of readLAS(). It is the responsibility of the user to choose. Inside the function you are a developer, outside your are a user.

The object LAScluster already carries the information needed (because you used readTLSLAScatalog()) and readLAS() will do the job transparently.

That being said you're right I should add something to trigger a warning or an error and better document this point.

[zoeschindler]

Thanks for looking into this and answering! :)

[Jean-Romain]

I added an error. Thanks