Package {command}

command: Process command line arguments

Description

Process command line arguments, allowing scripts to behave like functions, with well-defined inputs and outputs. Helps make data analysis workflows more modular, and therefore more transparent, flexible, and reliable.

Details

• cmd_assign() Process command line arguments

• cmd_assign_collect() Collect unnamed command line arguments into one character vector

• use_renv() Activate a renv project before loading packages

• extract_shell() Turn a cmd_assign() call into a shell command

• extract_make() Turn a cmd_assign() call into a Makefile rule

• shell_script() Create a shell script

• makefile() Create a Makefile

• Quick Start How to use cmd_assign()

• Modular Workflows for Data Analysis Safe, flexible data analysis workflows

• Using command with renv Running pipeline scripts with renv

Author(s)

Maintainer: John Bryant john@bayesiandemography.com

Authors:

• John Bryant john@bayesiandemography.com

Other contributors:

• Bayesian Demography Limited [copyright holder]

See Also

Useful links:

• https://bayesiandemography.github.io/command/

• https://github.com/bayesiandemography/command

• Report bugs at https://github.com/bayesiandemography/command/issues

Assign Values Passed at the Command Line or Interactively

Description

Assign names to values in the global environment. The values are typically supplied through the command line, but can be supplied interactively.

cmd_assign_quiet() is identical to cmd_assign() except that it does not print progress messages to the console.

Usage

cmd_assign(...)

cmd_assign_quiet(...)

Arguments

Details

cmd_assign() is a user-friendly alternative to commandArgs(), designed specifically for data analysis workflows.

cmd_assign() can be used to record and enforce assumptions about inputs and outputs. This provides scripts with better-defined interfaces, making them clearer and safer.

Value

cmd_assign() is called for its side effect, which is to create objects in the global environment. However, cmd_assign() also invisibly returns a named list of objects.

Types of session

cmd_assign() behaves differently depending on whether it is called interactively, or from within an R script being run from the command line.

For instance, if the code

cmd_assign(use_log = TRUE)

is run interactively, it creates an object called use_log with value TRUE.

But if the same code is run from within a script that has been invoked using the command

⁠Rscript tidy_data.R --use_log=FALSE⁠

it creates an object called use_log with value FALSE.

cmd_assign() is typically called interactively when a workflow is being developed, and through the command line when the workflow is in production.

Matching names and values

When used in a script called from the command line, cmd_assign() first matches named command line arguments, and then matches unnamed command line arguments, in the order in which they are supplied.

If, for instance, the script person.R contains the lines

cmd_assign(
  .data = "raw_data.csv",
   max_age = 85,
   .out = "person.rds"
)

and if person.R is run from the command line using

Rscript person.R \
  raw_data.csv \
  person.rds \
  --max_age=100

then cmd_assign() assigns cmd_assign() names to command line arguments in the following order:

The values specified in the call to cmd_assign() are not defaults

The values specified in the call to cmd_assign() are not defaults in the sense of values to fall back on when no value is supplied. For instance, if we take a script containing the call

cmd_assign(
  n_draw = 10,
  .out = "simulation.rds"
)

and run it from the command line without passing a value for n_draw, then cmd_assign() will throw an error. It will not set n_draw to 10.

Defaults can hide bugs and make code less predictable.

Coercing values passed at the command line

Values passed at the command line start out as text strings. cmd_assign() coerces these strings to have the same class as the corresponding values in the call to cmd_assign(). For instance, if script fit.R contains the lines

cmd_assign(
  .data = "data/individuals.rds",
  impute = TRUE,
  date = as.Date("2025-01-01"),
  .out = "output/fitted.rds"
)

and if fit.R is run from the command line using

Rscript src/fit.R \
  data/individuals.rds \
  output/fitted.rds \
  --impute=TRUE \
  --date=2025-01-01

then cmd_assign() will create the following length-1 vectors

The dot-name convention

When naming arguments to cmd_assign(), it is typically good practice to use names starting with dots for file paths, and names without dots for ordinary objects, as in

cmd_assign(
  .households = "data/households.csv",
  n_record = 500,
  sex = "Female",
  .out = "out/subsample_female.rds"
)

Function cmd_assign() does not in fact distinguish between names with dots and names without dots. But the dot-name convention is still useful because of the way it calls attention to the distinction between file paths and ordinary objects. For instance, a dot-name is a reminder that the object being named is a reference to an input or output, rather than the input or output itself.

In contrast, functions extract_make(), extract_shell(), makefile(), and shell_script() do treat names with dots differently from names without dots. See the help for extract_make(), extract_shell(), makefile(), and shell_script() for details.

References

• Command-Line Programs Introduction to Rscript

See Also

• cmd_assign_collect() Collect unnamed command line arguments into one character vector

• extract_shell() Turn a cmd_assign() call into a shell command

• extract_make() Turn a cmd_assign() call into a Makefile rule

• shell_script() Create a shell script

• makefile() Create a Makefile

• use_renv() Activate a renv project before loading packages

• Quick Start How to use cmd_assign()

• Modular Workflows for Data Analysis Safe, flexible data analysis workflows.

• Using command with renv Running pipeline scripts with renv

• Base R function commandArgs() uses a more general, lower-level approach to processing command line arguments. (commandArgs() is called internally by cmd_assign().)

• littler Alternative to Rscript

Examples

if (interactive()) {
  cmd_assign(.data = "mydata.csv",
             n_iter = 2000,
             .out = "results.rds")
}

Assign Varying Set of Values Passed at the Command Line or Interactively

Description

Assign names to values in the global environment, like cmd_assign(), but without specifying the number of values in advance. Values for unnamed command line arguments are collected into a single character vector.

cmd_assign_collect_quiet() is identical to cmd_assign_collect() except that it does not print progress messages to the console.

Usage

cmd_assign_collect(...)

cmd_assign_collect_quiet(...)

Arguments

Value

cmd_assign_collect() is called for its side effect, which is to create objects in the global environment. However, cmd_assign_collect() also invisibly returns a named list of objects.

Matching names and values

cmd_assign_collect(), like cmd_assign(), starts by processing named command line arguments (i.e. arguments of the form --n_draw=100). Unlike cmd_assign(), however, cmd_assign_collect() puts all the values that are left over into a single character vector. This vector can be any length, including zero. cmd_assign_collect() must contain a single name-value pair that is not used for named command line arguments. cmd_assign_collect() assigns the name from this name-value to the character vector.

If, for instance, the script combine.R contains

cmd_assign_collect(
  .files = c("data/a.csv", "data/b.csv", "out/combined.rds"),
  use_log = TRUE
)

and if combine.R is run from the command line using

Rscript combine.R \
  data/a.csv \
  data/c.csv \
  out/combined.rds \
  --use_log=TRUE

then use_log will be assigned the value TRUE and .files will be assigned the value c("data/a.csv", "data/c.csv", "out/combined.rds").

Processing the vector of unnamed arguments

The R script in which cmd_assign_collect() is called has responsibility for processing the character vector of values from unnamed command line arguments. The call to cmd_assign_collect() above, for instance, might be followed by statements such as

.inputs <- head(.files, -1)
.out <- tail(.files, 1)
inputs <- lapply(.inputs, read.csv)

Use cmd_assign() where possible

cmd_assign_collect() is more flexible than cmd_assign(), in that it does not require the number, name, or type of values to be specified in advance. But if these things are in fact known in advance, then it is good practice to document and test for them, which is what cmd_assign() does.

See Also

• cmd_assign() Process command line arguments with one value per argument

• extract_shell() Turn a cmd_assign() or cmd_assign_collect() call into a shell command

• extract_make() Turn a cmd_assign() or cmd_assign_collect() call into a Makefile rule

• shell_script() Create a shell script

• makefile() Create a Makefile

• Quick Start How to use cmd_assign()

Examples

if (interactive()) {
  cmd_assign_collect(
    files = c("data/x.csv", "data/y.csv"),
    use_log = TRUE
  )
}

Turn a 'cmd_assign' Call Into a Makefile Rule

Description

Extract a call to cmd_assign() or cmd_assign_collect() from an R script, and turn it into a Makefile rule.

Usage

extract_make(path_file, dir_make = NULL)

Arguments

Details

The rule may need editing before it is added to a Makefile.

Value

extract_make() is typically called for its side effect, which is to print a Makefile rule. However, extract_make() invisibly returns a text string with the rule.

The components of a Makefile rule

A Makefile rule produced by extract_make() normally looks something like this:

out/model.rds: src/model.R \
  data/cleaned.rds
       Rscript $^ $@ --use_log=TRUE

In this rule

• out/model.rds is the "target", i.e. the file that the rule creates;

• src/model.R and data/timeseries.rds are "prerequisites", i.e. files that are used to create the target;

• ⁠\⁠ is a "line continuation character";

• ⁠ ⁠ at the start of the third line is a tab, telling make that the recipe for creating the target from the starts here;

• Rscript is a call to utils::Rscript();

• ⁠$^⁠ is an automatic variable meaning "all the prerequisites" and ⁠$@⁠ is an automatic variable meaning "the target", so that ⁠Rscript $^ $@⁠ expands to ⁠Rscript src/model.R data/cleaned.rds out/model.rds⁠; and

• --use_log=TRUE is a named argument that Rscript passes to src/model.R

Using extract_make() to build a data analysis workflow

• Step 1. Write the R file that carries out the step in analysis (eg tidying data, fitting a model, making a graph.) This file will contain a call to cmd_assign() or cmd_assign_collect(), and will appear in the first line of the Makefile rule. When writing and testing the file, use cmd_assign() or cmd_assign_collect() interactively.

• Step 2. Once the R file is working correctly, call extract_make(), and add the rule to your Makefile.

When using extract_make(), it is a good idea to set the current working directory to the project directory.

Location of the Makefile

The Makefile normally sets at the top of the project, so that the project folder looks something like this:

Makefile
- data/
- src/
- out/
report.qmd

Identifying file arguments

To identify arguments that refer to file names, extract_make() uses the following rules:

• if the call includes arguments whose names start with a dot, then these arguments are assumed to refer to file names;

• otherwise, find arguments whose values actually are file names (as determined by file.exists()), or that look like they could be.

References

• Project Management with Make Makefiles in data analysis workflows

• GNU make Definitive guide

• Command-Line Programs Introduction to Rscript

See Also

• extract_shell() Shell script equivalent of extract_make()

• makefile() Create a Makefile from calls to cmd_assign()

• cmd_assign() Process command line arguments

• cmd_assign_collect() Process command line arguments when the number of arguments is not known in advance

• Quick Start How to use cmd_assign() and cmd_assign_collect()

• Modular Workflows for Data Analysis Safe, flexible data analysis workflows

• littler Alternative to Rscript

Examples

library(withr)
with_tempdir({

  ## create 'src'  directory
  dir.create("src")

  ## put an R script containing a call to
  ## 'cmd_assign' in the 'src' directory
  writeLines(c("cmd_assign(x = 1, .out = 'out/results.rds')",
               "results <- x + 1",
               "saveRDS(results, file = .out)"),
             con = "src/results.R")

  ## call 'extract_make()'
  extract_make(path_file = "src/results.R",
               dir_make = ".")

})

Turn a 'cmd_assign' Call Into a Shell Command

Description

Extract a call to cmd_assign() or cmd_assign_collect() from an R script, and turn it into a shell command.

Usage

extract_shell(path_file, dir_shell = NULL)

Arguments

Value

extract_shell() is typically called for its side effect, which is to print a shell command. However, extract_shell() invisibly returns a text string with the command.

The components of a shell command

The shell command produced by extract_shell() normally looks something like this:

Rscript src/model.R \
  data/cleaned.rds \
  out/model.rds \
  --use_log=TRUE

In this command

• Rscript is a call to utils::Rscript();

• ⁠\⁠ is a "line continuation character";

• data/cleaned.rds and out/model.rds are unnamed arguments that Rscript passes to src/model.R; and

• --use_log=TRUE is a named argument that Rscript passes to src/model.R

Using extract_shell() to build a data analysis workflow

• Step 1. Write an R script that carries out a step in analysis (eg tidying data, fitting a model, making a graph.) This script will contain a call to cmd_assign() or cmd_assign_collect(), and will be the first argument passed to Rscript in the shell command. When writing and testing the script, use cmd_assign() or cmd_assign_collect() interactively.

• Step 2. Once the R script is working correctly, call extract_shell(), and add the command to your shell script.

Location of the shell script

The shell script normally sits at the top level of the project, so that the project folder looks something like this:

workflow.sh
- data/
- src/
- out/
report.qmd

Identifying file arguments

To identify arguments that refer to file names, extract_shell() uses the following rules:

• if the call includes arguments whose names start with a dot, then these arguments are assumed to refer to file names;

• otherwise, find arguments whose values actually are file names (as determined by file.exists()), or that look like they could be.

References

• Episodes 1–3 of The Unix Shell Introduction to the command line

• Command-Line Programs Introduction to Rscript

• littler Alternative to Rscript

See Also

• extract_make() Makefile equivalent of extract_shell()

• shell_script() Create a shell script from calls to cmd_assign()

• cmd_assign() Process command line arguments

• cmd_assign_collect() Process command line arguments when the number of arguments is not known in advance

• Quick Start How to use cmd_assign() and cmd_assign_collect()

• Modular Workflows for Data Analysis Safe, flexible data analysis workflows

Examples

library(withr)

with_tempdir({

  ## create 'src' directory
  dir.create("src")

  ## add an R script containing a call to 'cmd_assign'
  writeLines(c("cmd_assign(x = 1, .out = 'out/results.rds')",
               "results <- x + 1",
               "saveRDS(results, file = .out)"),
             con = "src/results.R")

  ## call 'extract_shell()'
  extract_shell(path_file = "src/results.R",
                dir_shell = ".")

})

Create a Makefile

Description

Create a Makefile for a data analysis workflow. The Makefile can include rules extracted from existing R files.

Usage

makefile(
  path_files = NULL,
  dir_make = NULL,
  name_make = "Makefile",
  overwrite = FALSE,
  quiet = FALSE
)

Arguments

Details

The Makefile is a draft and needs editing before use. makefile() cannot infer every rule, and a file path that contains a space or a character special to Make must be corrected in the dependency line.

To create a Makefile in the files directory, set files to ".".

To obtain the contents of the Makefile without creating a file on disk, creating the file on disk, set name_make to NULL.

Supplying a value for files is optional for makefile(), but compulsory for shell_script(). The output from makefile() includes some general-purpose Makefile commands, while the output from shell_script() is generated entirely from files.

Value

makefile() is called for its side effect, which is to create a file. However, makefile() also returns a string with the contents of the Makefile.

References

• Project Management with Make Makefiles in data analysis workflows

• GNU make Definitive guide

• Command-Line Programs Introduction to Rscript

See Also

• Creating a Makefile More on makefile()

• extract_make() Turn a cmd_assign() call into a Makefile rule

• shell_script() Shell script equivalent of makefile()

• cmd_assign() Process command line arguments

• Modular Workflows for Data Analysis Safe, flexible data analysis workflows

• littler Alternative to Rscript

Examples

library(withr)

with_tempdir({

  ## create 'src'  directory
  dir.create("src")

  ## put R scripts containing calls to
  ## 'cmd_assign' in the 'src' directory
  writeLines(c("cmd_assign(x = 1, .out = 'out/results.rds')",
               "results <- x + 1",
               "saveRDS(results, file = .out)"),
             con = "src/results.R")
  writeLines(c("cmd_assign(x = 1, .out = 'out/more_results.rds')",
               "more_results <- x + 2",
               "saveRDS(more_results, file = .out)"),
             con = "src/more_results.R")

  ## call 'makefile()'
  makefile(path_files = "src",
           dir_make = ".")

  ## Makefile has been created
  list.files()

  ## print contents of Makefile
  cat(readLines("Makefile"), sep = "\n")

})

Create a Shell Script

Description

Create a shell script for a data analysis workflow consisting of commands extracted from existing R files.

Usage

shell_script(
  path_files,
  dir_shell = NULL,
  name_shell = "workflow.sh",
  overwrite = FALSE,
  quiet = FALSE
)

Arguments

Details

shell_script() tries to put the commands in the order they need to run, by matching each script's output path with the input paths of the other scripts. It can get this order wrong, so the shell script may still need editing.

To create a shell script in the files directory, set files to ".".

To obtain the contents of the shell script without creating a file on disk, creating the file on disk, set name_shell to NULL.

Supplying a value for files is compulsory for shell_script(), but optional for makefile(). The output from shell_script() is generated entirely from files while the output from makefile() also includes some general-purpose Makefile commands.

Value

shell_script() is called for its side effect, which is to create a file. However, shell_script() also returns a string with the contents of the shell script.

References

• Episodes 1–3 of The Unix Shell Introduction to the command line

• Command-Line Programs Introduction to Rscript

See Also

• Creating a Shell Script More on shell_script()

• extract_shell() Turn a cmd_assign() call into a shell command

• makefile() Makefile equivalent of shell_script()

• cmd_assign() Process command line arguments

• Modular Workflows for Data Analysis Safe, flexible data analysis workflows

• littler Alternative to Rscript

Examples

library(withr)

with_tempdir({

  ## create 'src'  directory
  dir.create("src")

  ## put R scripts containing calls to
  ## 'cmd_assign' in the 'src' directory
  writeLines(c("cmd_assign(x = 1, .out = 'out/results.rds')",
               "results <- x + 1",
               "saveRDS(results, file = .out)"),
             con = "src/results.R")
  writeLines(c("cmd_assign(.results = 'out/results.rds', .out = 'out/more_results.rds')",
               "results <- readRDS(.results)",
               "more_results <- results + 2",
               "saveRDS(more_results, file = .out)"),
             con = "src/more_results.R")

  ## the shell script runs 'results.R' first, because
  ## 'more_results.R' reads the file that 'results.R' writes

  ## call 'shell_script()'
  shell_script(path_files = "src",
               dir_shell = ".")

  ## shell script has been created
  list.files()

  ## print contents of shell script
  cat(readLines("workflow.sh"), sep = "\n")

})

Activate an renv Project for the Current Session

Description

Make sure renv is activated, even if a script is run from Rscript or littler.

Usage

use_renv(project = NULL, quiet = TRUE)

Arguments

Details

Package renv makes workflows reproducible by locking down the versions of R packages. Normal renv mechanisms can fail, however, if a script is run from Rscript or littler. Adding

command::use_renv()

to the top of R scripts fixes the problem.

For more details, see the article Using command with renv.

Value

The project root path (invisibly) if renv was activated or was already active; otherwise NULL.

See Also

• cmd_assign() Process command line arguments

• cmd_assign_collect() Process command line arguments when the number of arguments is not known in advance

• Using command with renv

Examples

command::use_renv() ## if renv not used here, no effect

## see article for full example