Getting started with wbcmd

If you work with neuroimaging data in R, you’ve probably typed out a system() call to wb_command at some point. Maybe you wrapped it in a helper function. Maybe that helper grew. wbcmd is the version of that helper that validates your inputs, finds your binary, and gets out of the way.

Finding the binary

Install Connectome Workbench, then load the package:

library(wbcmd)
wb_sitrep()

wb_sitrep() checks whether the binary is reachable and prints what it finds. On macOS and Linux, wbcmd searches common installation paths and the system PATH automatically.

If the binary lives somewhere unusual, tell wbcmd where:

set_wb_path("/opt/workbench/bin_linux64/wb_command")

The full lookup order is:

  1. R option wbcmd.path
  2. Environment variable WB_PATH
  3. Platform-specific defaults (e.g. /Applications/wb_view.app/... on macOS)
  4. System PATH

Set WB_PATH in your .Renviron and you never think about it again.

The direct route

wb_cmd() is the universal entry point. It takes the subcommand name and arguments exactly as you would type them on the command line:

wb_cmd("-version")
wb_cmd("-file-information", "data.dscalar.nii")

Every wrapper function in wbcmd calls wb_cmd() internally. If a wrapper doesn’t exist for the command you need, wb_cmd() always works.

Wrappers

The wrappers handle argument construction, file validation, and output path management. They error before the command runs if an input file is missing, and they create output directories on the fly.

Parcellate CIFTI data

cifti_parcellate(
  "data.dtseries.nii",
  "atlas.dlabel.nii",
  direction = "COLUMN",
  cifti_out = "parcellated.ptseries.nii"
)

Compute a correlation matrix

cifti_correlation(
  "data.dtseries.nii",
  "corr.dconn.nii",
  fisher_z = TRUE
)

Reduce across maps

cifti_reduce(
  "data.dtseries.nii",
  direction = "ROW",
  operation = "MEAN",
  cifti_out = "mean.dscalar.nii"
)

Surface curvature

surface_curvature(
  "lh.midthickness.surf.gii",
  mean_out = "lh.curvature.func.gii"
)

Map a volume to a surface

volume_to_surface_mapping(
  "bold.nii.gz",
  "lh.midthickness.surf.gii",
  metric_out = "lh.bold.func.gii",
  method = "ribbon-constrained",
  inner_surface = "lh.white.surf.gii",
  outer_surface = "lh.pial.surf.gii"
)

Math expressions

cifti_math(), metric_math(), and volume_math() all take a symbolic expression and a named list of file paths:

cifti_math(
  "(x + y) / 2",
  cifti_out = "average.dscalar.nii",
  var = list(
    x = "subj01.dscalar.nii",
    y = "subj02.dscalar.nii"
  )
)

Getting help

Every wrapper function includes a “Connectome Workbench Help” section in its documentation, pulled directly from wb_command at build time. You can also call wb_help() interactively:

wb_help("-cifti-parcellate")

Using wbcmd with ciftiTools

ciftiTools reads and writes CIFTI files as R objects. wbcmd runs Workbench operations on those files. They complement each other — ciftiTools gives you the data in R, wbcmd gives you the processing pipeline.

Both packages need the same binary. Point them to the same path and they share the setup:

wb_path <- "/opt/workbench/bin_linux64/wb_command"

wbcmd::set_wb_path(wb_path)
ciftiTools::ciftiTools.setOption("wb_path", wb_path)

A typical workflow reads data with ciftiTools, processes files with wbcmd, then reads the results back:

library(ciftiTools)
library(wbcmd)

xii <- read_cifti("data.dtseries.nii", brainstructures = "all")

cifti_parcellate(
  "data.dtseries.nii",
  "atlas.dlabel.nii",
  direction = "COLUMN",
  cifti_out = "parcellated.ptseries.nii"
)

parcellated <- read_cifti("parcellated.ptseries.nii")

ciftiTools also has its own run_wb_cmd() function that takes a pre-formatted command string. If you already use that, wbcmd’s wb_cmd() does the same job with the addition of path detection, argument quoting, and verbose logging.

Controlling verbosity

By default, wbcmd prints the full command before running it. Turn that off for cleaner output:

options(wbcmd.verbose = FALSE)

Or set WB_VERBOSE=FALSE in your .Renviron.