Skip to contents

Capturing NetLogo output manually

Jan Salecker

2024-03-30

Source: vignettes/articles/manual-output.Rmd
manual-output.Rmd

Capturing output manually

While nlrx provides the metrics, metrics.turtles, metrics.patches and metrics.links slots of the exerpiment class to capture global and agent related output from NetLogo models, this might not be sufficient under certain circumstances. For example, the metrics slot cannot track and collect nested list output or similar complex data structures such as NetLogo arrays or matrices. Another example would be spatial output such as shapefiles or grids, generated with the GIS extension. Additionally, you might already have implemented complex routines in your model which write model output to disk.

The main question for this vignette to answer is: How can you link such self-written output to your nlrx experiment. Here we show you one basic example how this works. In this example we write ascii raster files using the GIS extension in NetLogo. However, the same workflow can of course be applied to other types of output, for example text files written with file-type primitives or the csv extension.

Step 1: Creating an ID widget on your model interface

In order to link self-written output to our nlrx simulations within the R session, we need to transfer the current nlrx siminputrow and seed to our NetLogo model. That way, when executing run_nl_all(), within each simulation the NetLogo model exactly “knows” which parameter row (siminputrow) is currently simulated and the corresponding random seed. To transfer this information we just need to create a string input widget on our model interface. In the example below, we created such a widget at the bottom of the Wolf Sheep model called nlrx_id.

Step 2: Define idrunnum of the experiment

The idrunnum field of the experiment is a built-in feature that allows to transfer the current experiment name, siminputrow and random seed to a defined NetLogo gui parameter widget. In our case we want to use our newly created nlrx_id string input field, so we just set idrunnum = nlrx_id.

# Attach experiment
nl@experiment <- experiment(expname="wolf-sheep",
                            outpath=outpath,
                            repetition=1,
                            tickmetrics="true",
                            idsetup="setup",
                            idgo="go",
                            idrunnum="nlrx_id",
                            runtime=50,
                            evalticks=seq(40,50),
                            metrics=c("count sheep"),
                            variables = list('initial-number-sheep' = list(min=50, max=150, qfun="qunif"),
                                             'initial-number-wolves' = list(min=50, max=150, qfun="qunif")),
                            constants = list("model-version" = "\"sheep-wolves-grass\"",
                                             "grass-regrowth-time" = 30,
                                             "sheep-gain-from-food" = 4,
                                             "wolf-gain-from-food" = 20,
                                             "sheep-reproduce" = 4,
                                             "wolf-reproduce" = 5,
                                             "show-energy?" = "false"))

Step 3: Use nlrx_id within NetLogo model to tag self-written output

We can now use the nlrx_id field within the NetLogo model to tag our self-written output. One way to do this is to tag the filenames of self-written output. Here is an example where raster grids are created and written at the end of each model run:

Here, we first create our filename by using the string from the nlrx_id field (which contains the current experiment name, siminputrow and random seed divided by an underscore) and add the current tick (containing leading zeros by adding 1000 to the tick count and removing the first number afterwards).

Of course you could also use different approaches to utilize the information within the nlrx_id field for tagging your output.

Step 4: Linking self-written output to nlrx collected output

Finally, we want to read in our self-written raster files, calculate some landscape metrics and add those to the results table that we received from the run_nl_all() function.

We basically have two types of output now: The tibble from our model executions, and a folder with self-written output (here defined as a subfolder of our modelpath).

## Output from nlrx simulations:
results <- run_nl_all(nl)
## Self-writte output directory:
ascdir <- file.path(dirname(nl@modelpath), "output")

We can now loop over the files in that folder, read the content and use the strplit function on the filename to identify the experiment name, the siminputrow, the random seed and the tick count. As an example application, we then calculate some landscapemetrics using the landscapemetrics package. Finally we put everything into a results tibble:

results.lsm <- purrr::map_dfr(list.files(ascdir, pattern = "asc", full.names = TRUE), function(x) {
  x.split <- strsplit(x, "_")[[1]]
  x.tick <- as.numeric(strsplit(x.split[[length(x.split)]], "\\.")[[1]][[1]])
  x.siminputrow <- as.numeric(x.split[[length(x.split) - 1]][[1]])
  x.seed <- as.numeric(x.split[[length(x.split) - 2]][[1]])
  x.raster <- raster(x)
  
  ## netlogo asc files use NaN as default nodata value in the asc file header
  ## this leads to problems when reading the raster because it sets ´zeros to NA
  ## here we set NAs back to zertos manually:
  x.raster <- reclassify(x.raster, cbind(NA, 0))
  
  ## Calculate landscape metrics:
  metrics <- c("lsm_l_ed", "lsm_l_shdi", "lsm_l_lsi", "lsm_l_lpi", "lsm_l_area_mn")
  x.metrics <- landscapemetrics::calculate_lsm(x.raster, what=metrics) %>% 
    dplyr::select(metric, value) %>% 
    tidyr::pivot_wider(names_from=metric, values_from = value)
  
  x.final <- tibble::tibble(siminputrow = x.siminputrow,
                            `[step]` = x.tick,
                            `random-seed` = x.seed)
  
  x.final <- cbind(x.final, x.metrics)
  return(x.final)
})

Now we have the same identifier columns in the nlrx reported output and the self-written output tibble which allows us to join both tibbles together:

## Combine results with lsm and store:
results <- results %>% left_join(results.lsm, by = c("siminputrow", "random-seed", "[step]"))

## Attach output to nl object:
setsim(nl, "simoutput") <- results
saveRDS(nl, file = file.path(outpath, "my_final_nl_object.rds"))