Symlink Tool Technical Vignette"

In vmTools: Version Management Tools on the File System

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

# for debugging
# knitr::opts_chunk$set(
#   echo = TRUE,
#   message = TRUE,
#   warning = TRUE,
#   error = TRUE
# )

# Chunk options
.opt_width <- options(width = 450)

# save the built-in output hook
hook_output <- knitr::knit_hooks$get("output")
# flags to determine output
flag_eval_chunk <- if (vmTools:::is_windows_admin() | .Platform$OS.type %in% c("unix", "linux")) TRUE else FALSE

# set a new output hook to truncate text output
# - set a chunk option as e.g. : `{r chunk_name, out.lines = 15}`
# if the output is too long, it will be truncated like:
# 
# top output
# ...
# bottom output
knitr::knit_hooks$set(output = function(x, options) {
  if (!is.null(n <- options$out.lines)) {
    x <- vmTools:::split_line_breaks(x)
    if (length(x) > n) {
      # truncate the output
      # x <- c(head(x, n), "....\n")
      x <- c(head(x, n/2), '....', tail(x, n/2 + 1))
    }
    x <- paste(x, collapse = "\n")
  }
  hook_output(x, options)
})

if(!flag_eval_chunk){

 knitr::asis_output("
> **Note:** This vignette demonstrates symbolic link creation, which requires administrator privileges on Windows.
> 
> On systems without these privileges, code chunks are not evaluated, but all code is shown.
> 
> To fully run this vignette, use a Unix-based system or Windows with administrator rights.
")

} 

# Defining a couple vignette utilities

print_tree <- function(x) {vmTools:::dir_tree(x)}

# print a symlink's target from the file system 
print_symlink <- function(symlink_type){
  print(grep(symlink_type, system(paste("ls -alt", root_input), intern = TRUE), value = TRUE))
}

#' Get output directory for results to save in.
#'
#' Returns a path to save results in of the form "YYYY_MM_DD.VV".
#'
#' @param root path to root of output results
#' @param date character date in form of "YYYY_MM_DD" or "today". "today" will be interpreted as today's date.
get_output_dir <- function(root, date) {
  if (date == "today") {
    date <- format(Sys.Date(), "%Y_%m_%d")
  }
  cur.version <- get_latest_output_date_index(root, date = date)

  dir.name <- sprintf("%s.%02i", date, cur.version + 1)
  return(dir.name)
}

#' get the latest index for given an output dir and a date
#'
#' directories are assumed to be named in YYYY_MM_DD.VV format with sane
#' year/month/date/version values.
#'
#' @param dir path to directory with versioned dirs
#' @param date character in be YYYY_MM_DD format
#'
#' @return largest version in directory tree or 0 if there are no version OR
#' the directory tree does not exist
get_latest_output_date_index <- function(root, date) {
  currentfolders <- list.files(root)

  # subset to date
  pat <- sprintf("^%s[.]\\d{2}$", date)
  date_dirs <- grep(pat, currentfolders, value = T)

  if (length(date_dirs) == 0) {
    return(0)
  }

  # get the index after day
  date_list <- strsplit(date_dirs, "[.]")

  inds <- unlist(lapply(date_list, function(x) x[2]))
  if (is.na(max(inds, na.rm = T))) inds <- 0

  return(max(as.numeric(inds)))
}

print_public_methods <- function(SLT){
  output <- capture.output(print(SLT))
  idx_private <- which(output %like% "Private")
  idx_clone <- which(output %like% "clone")
  idx_custom <- which(output %like% "startup guidance messages") # can't get this to print - frustrating
  # idx_custom <- which(output %like% "foo")
  idx_keep <- c(1:idx_private - 1, idx_custom)
  idx_keep <- setdiff(idx_keep, idx_clone)
  cat(paste0(output[idx_keep], collapse = "\n"))
}

What is the SymLink Tool?

The SymLink Tool is an R6 object-oriented tool that helps a researcher manage pipeline outputs in a standard way. It falls under the category of 'standard tooling'. It will:

1. Create 'best' symlinks to the best version of your outputs.
   1. Safely update the 'best' symlink when you, the researcher, want to 'promote' a new version.
2. Maintain a log of 'best' promotions & demotions within each versioned output folder.
   1. Provide reports on the state of all these logs. E.g. 'what is currently best?'
3. Maintain a central log of which versions of your pipeline have ever been 'best'.
   1. The central log can also track versioned folder deletion.

Assumptions

This assumes you have a large set of versioned output folders for your pipeline.

If you're already using a database to manage your output versions, you probably don't need this.

If you have a big mess of folders you're having difficulty tracking, this tool may help you out!

---

SymLink Tool Intro

SLT (short for SymLink Tool) is an R6 object generator, or "R6ClassGenerator".

When you want to make a new 'instance' of a tool, call the new method on the tool's class.

• This is called "instantiation".
• Think of SLT like a template, the Platonic ideal of a hammer. Think of slt as your favorite, particular hammer.

Let's start by calling $new() with no arguments.

• We expect an error, with some helpful messages about what arguments the tool template expects, in order to make the type of tool we need.

library(vmTools)
library(data.table)
slt <- try(SLT$new())

OK, now that we know what the tool expects, let's feed it this information and try using it in earnest.

SymLink Tool Use

In my pipeline, I divert outputs to two folders:

1. 'to_model', which is prepped data going into ST-GPR
2. 'modeled', which is post-processed after ST-GPR

I want to have the same version_name of my pipeline outputs in both roots so I can correlate pre and post modeled data.

If you need to handle roots independently, then you should instantiate different versions of the tool to handle each independent root, giving each instance of the tool a unique name e.g. slt_input and slt_output.

• A version_name is simply a string like "2024_02_02_new_covariates" that's important to you, the modeler, to tell you when and why the pipeline was run. There is no requirement for this to include a date, but it's good practice.

In addition, the tool needs a location for a central log. I'll set that one level above both my output folders, since the central log will be shared between them.

• The tool manages outputs at the version_name level, not the folder level, so one 'best' promotion affects folders in both my output roots.

# a safe temporary directory every user has access to, that we'll clean up later
root_base   <- file.path(tempdir(), "slt")
root_input  <- file.path(root_base, "to_model")
root_output <- file.path(root_base, "modeled")

PATHS       <- list(
  log_cent       = file.path(root_base, "log_symlinks_central.csv"),
  log_2024_02_02 = file.path(root_input, "2024_02_02", "logs/log_version_history.csv"),
  log_2024_02_10 = file.path(root_input, "2024_02_10", "logs/log_version_history.csv")
)

Try to make the tool naively.

slt <- try(SLT$new(
  user_root_list = list(
    root_input  = root_input,
    root_output = root_output
  )
  , user_central_log_root = root_base
))

# We need to ensure all output folders exist first
dir.create(root_input, recursive = TRUE, showWarnings = FALSE)
dir.create(root_output, recursive = TRUE, showWarnings = FALSE)

# Now everything should work
suppressWarnings({ # idiosyncratic and benign cluster message

  slt <- SLT$new(
    user_root_list = list(
      root_input  = root_input,
      root_output = root_output
    )
    , user_central_log_root = root_base
  )

})

# multithreading can cause github actions issues
.opt_mccores <- options(mc.cores = 1)

What do we have in our root_base folder?

print_tree(root_base)

We should now have a central log, and two output folder.

• When you instantiate the tool, the central log is made automatically.
• We must make the folders, or the tool will stop.

Mark Best

You can mark any output folder as 'best', and give it a 'best' symlink in each output root.

• Only one version_name can be 'best', and the SLT will demote the current 'best' version if you promote a new version_name.
• This will create a log in the version_name folder, and an entry in the central log.

NOTE:

• The version_name log record records both 'demote' and 'promote' actions.
• The central log only records 'promote' actions.

NOTE:

• Each 'mark' action needs the user to provide a log entry comment as a named list.

First we'll create two version_name folders to play with in each root.

dir.create(file.path(root_input, "2024_02_02"), recursive = TRUE, showWarnings = FALSE)
dir.create(file.path(root_output, "2024_02_02"), recursive = TRUE, showWarnings = FALSE)
dir.create(file.path(root_input, "2024_02_10"), recursive = TRUE, showWarnings = FALSE)
dir.create(file.path(root_output, "2024_02_10"), recursive = TRUE, showWarnings = FALSE)

print_tree(root_base)

Then we'll mark one as best.

slt$mark_best(version_name = "2024_02_02", user_entry = list(comment = "testing mark_best"))

Look at the folder structure again.

print_tree(root_base)

Where does the 'best' folder point to?

print_symlink("best")

Now let's mark the other one as best, and see what happens to the symlinks.

# The tool is chatty by default at the console, but it's easy to make it quite if it's part of a pipeline.
suppressMessages({
  slt$mark_best(version_name = "2024_02_10", user_entry = list(comment = "testing mark_best"))
})

print_tree(root_base)

print_symlink("best")

Inspect Logs

In this trio, we see the central log, and each versioned folder's log of best promotion events.

• The central log only records promotions, not demotions.
• The versioned logs record both promotions and demotions.
• Note each log comes with a creation date - this is the log creation date, not the folder creation date (that information isn't available on the cluster).

data.table::fread(PATHS$log_cent)

data.table::fread(PATHS$log_2024_02_02)

data.table::fread(PATHS$log_2024_02_10)

Reports

You've probably also noticed a report. This also shows the current state of all tool-created symlinks, built from the version_name folder logs (not the central log).

• This report runs automatically any time you mark a folder.
• It only shows symlinks made by this tool, so if you make other symlinks, they won't be included.
  • Other reports are possible (see 'Other Features' section.)
• It runs separately for each root, and pulls the last log entry.

data.table::fread(file.path(root_input, "report_key_versions.csv"))

data.table::fread(file.path(root_output, "report_key_versions.csv"))

Create New Folders

This tool was designed to allow the researcher to use it 'mid-stream' during a modeling round. I.e. you may mark existing output folders as best, and all the logging takes care of itself.

In addition, the researcher may choose to have this tool manage folder creation. This is useful if you want to ensure that all your output folders are managed by the same tool, and that the tool is aware of all the version_name folders that exist. Further, there are more reports you can run against your version_name logs that are more informative if you create all your folders with the tool. (See 'Other Features')

• This records the creation date and time of each folder, so you can 'round up' folders that are younger or older than some date.
• The Linux filesystem does not record the creation date of folders.

# Let's use a programmatic example to build a new `version_name.`
version_name_input  <- get_output_dir(root_input,  "today")
version_name_output <- get_output_dir(root_output, "today")
if(!version_name_input == version_name_output) {
  stop("version_name_input and version_name_output must be the same")
}
version_name_today  <- intersect(version_name_input, version_name_output)

# This creates folders safely, and will not overwrite existing folders if called twice.
slt$make_new_version_folder(version_name = version_name_today)
slt$make_new_version_folder(version_name = version_name_today)

Now we can see the new folders, with their logs and creation date-time stamps.

• Creating new folders using the method above will index new version as YYYY_MM_DD.VV if reruns on the same day are necessary.
• e.g. 2024_02_10.01, 2024_02_10.02, 2024_02_10.03, etc.
• Note this behavior is outside the tool, for demonstration purposes only.
• The tool assumes the research team determines its own folder-naming conventions.

print_tree(root_base)

data.table::fread(file.path(root_input, version_name_today, "logs/log_version_history.csv"))

This folder can be marked best just as the others, and prior best will be demoted.

suppressMessages({
  slt$mark_best(version_name = version_name_today, user_entry = list(comment = "testing mark_best"))
})
data.table::fread(PATHS$log_cent)

data.table::fread(PATHS$log_2024_02_10)

Unmark

Now let's say you review your results and no version of outputs should be 'best'. You can run unmark() to remove the 'best' status.

• There are no longer any 'best' symlinks, and the version_name log shows demotion.
• The central log does not show the demotion. The presence/absence of the best symlink, and the version_name log are the acid test of what's current.
• The tool-symlink report should now be empty.

suppressMessages({
  slt$unmark(version_name = version_name_today, user_entry = list(comment = "testing unmark_best"))
})
print_tree(root_base)

data.table::fread(PATHS$log_cent)

data.table::fread(file.path(root_input, version_name_today, "logs/log_version_history.csv"))

data.table::fread(file.path(root_input, "report_key_versions.csv"))

Deletion

Mark remove

We love our 'best' version of outputs as long as it's best, but time passes and we get new 'best' versions.

When it's time to remove those old folders, we can use this tool to do that safely, in two stages.

1. Mark a folder to 'remove'.
2. Delete that marked folder and update the central log.
3. This way you have a history of what was deleted, and when.

Think of the two-step process a bit like git add and git commit.

• First, you mark a folder with mark_remove, which puts it the the 'deletion staging area'.
• Second, if you're sure you want to delete it, you run delete_version_folders() to actually delete the folder and update the central log.

Let's demonstrate on the folder we made programmatically with today's date.

suppressMessages({
  slt$mark_remove(version_name = version_name_today, user_entry = list(comment = "testing mark_remove"))
})

And let's look at the folder structure.

• since many folders can be marked remove_, we combine 'remove_' with the version_name to make the folder name unique.

print_tree(root_base)

print_symlink("remove")

And finally look at the log.

• It should have a 'promote_remove' action in the last row.

data.table::fread(file.path(root_input, version_name_today, "logs/log_version_history.csv"))

The report should also update to show the new symlink.

data.table::fread(file.path(root_input, "report_key_versions.csv"))

Delete the Folder

First, let's naively try to delete a folder we haven't marked as ready for removal.

• We should be blocked from doing this operation, since we haven't marked the folder as 'remove'.

Note:

• this will try to delete folders in both output roots.

slt$delete_version_folders(
  version_name = "2024_02_02",
  user_entry   = list(comment = "testing delete_version_folders")
)

Now let's delete the folder we have marked as ready for removal.

• This will delete the folder and update the central log.
• This function defaults to requiring user input
  • For each root, it will ask if you're sure you want to delete the folder.
  • This allow the user to choose what to keep.
  • E.g. keep outputs, but delete the inputs to free up disk space.

slt$delete_version_folders(
  version_name       = version_name_today,
  user_entry         = list(comment = "testing delete_version_folders"),
  require_user_input = FALSE
)

Let's look at the folder structure.

• The folder and remove_ symlink should be gone.

print_tree(root_base)

Since we no longer have a version_name log, we can't look at it. But we can look at the central log.

• The last two rows should show:
  • the 'promote_remove' action
  • the 'delete_remove_folder' action

data.table::fread(PATHS$log_cent)

Let's look at the report.

• We should expect to not see the version_name in the report, since it's been deleted.

data.table::fread(file.path(root_input, "report_key_versions.csv"))

---

Other Features

Other available features will be covered briefly, and will assume the reader has already read the Symlink Tool Intro section.

Mark Keep

It's likely you'll have other output versions you want to keep, but not as 'best'. You can mark these as 'keep'.

• This will produce symlinks of the form keep_<version_name>

suppressMessages(
  slt$mark_keep(version_name = "2024_02_10", user_entry = list(comment = "testing mark_keep"))
)

print_tree(root_base)

print_symlink("keep")

Reports pt 2

In addition to the report_key_versions.csv file, there are other reports available. These will show the status of the last log row for each version_name folder in each root folder.

You can view things like:

• All your folders that are not currently symlinked (marked).
• All symlinked folders.
  • Includes symlinks not recognized by this tool.

NOTE: This includes a discrepancy report that shows if logs do not conform to expected standards.

• If you find a log discrepancy that is not caught by this report and you think it should be, please contact the package maintainer.
  • There is currently no issue queue, but this may be added if there is demand.

# Show the types of reports currently available
slt$make_reports

# Run the reports
suppressMessages({
  slt$make_reports()
})
print_tree(root_base)

# View an example report - logs for folders with no active symlink
# - you can see this folder was previously marked 'best'
data.table::fread(file.path(root_input, "report_all_logs_non_symlink.csv"))

# Expect this to be absent for the vignette
try(data.table::fread(file.path(root_input, "REPORT_DISCREPANCIES.csv")))

Roundups

Let's say you have a set of folders you want to keep or remove, and you want to do it all at once.

We'll demonstrate by:

1. Making a set of dummy folders
   1. Marking some as remove_
      1. Rounding up the remove_ folders for deletion
   2. Round up the rest by date
      1. Mark these as keep_

# Make a set of dummy folders
dv1  <- get_output_dir(root_input,  "today")
slt$make_new_version_folder(dv1)
dv2  <- get_output_dir(root_input,  "today")
slt$make_new_version_folder(dv2)
dv3  <- get_output_dir(root_input,  "today")
slt$make_new_version_folder(dv3)
dv4  <- get_output_dir(root_input,  "today")
slt$make_new_version_folder(dv4)

print_tree(root_base)

roundup_remove

# Mark some as 'remove_'
suppressMessages({
  for(dv in c(dv1, dv2)){
    slt$mark_remove(dv, user_entry = list(comment = "mark_remove for roundup"))
  }
})

# Round up and delete
roundup_remove_list <- slt$roundup_remove()

if(! identical(roundup_remove_list$root_input$version_name, 
               roundup_remove_list$root_output$version_name)){

  stop("roundup_remove found different version_names in each `root` folder")
}

suppressMessages({
  for(dv in roundup_remove_list$root_input$version_name){
    slt$delete_version_folders(
      version_name       = dv,
      user_entry         = list(comment = "roundup_remove"),
      require_user_input = FALSE
    )
  }
})
print_tree(root_base)

roundup_by_date

Use the log creation date (first row) to round up folders created on, before, or after that date.

my_date <- format(Sys.Date(), "%Y_%m_%d")
roundup_date_list <- slt$roundup_by_date(
  user_date     = my_date,
  date_selector = "lte" # less than or equal to today's date
)

# mark all our dummy folders (with the ".VV" pattern) as keepers
dv_keep <- grep(
  pattern = "\\.\\d\\d"
  , x     = roundup_date_list$root_input$version_name
  , value = TRUE
)
suppressMessages({
  for(dv in dv_keep){
    slt$mark_keep(dv, user_entry = list(comment = "roundup_by_date"))
  }
})
print_tree(root_base)

Make new log

The date roundup relies on the log creation date (recall, the Linux filesystem does not record folder creation / birth dates). If you've made your own folders without the symlink tool, you can make a blank log easily. You can hand-edit the creation date if you know when the folder was made.

Note:

• The tool tries to resolve all operations in each root independently. So even though we're not creating a folder in both our roots, the tool will create as many logs as it can.

# Make a naive folder without a log
dir.create(file.path(root_output, "2024_02_10_naive"))
try(slt$make_new_log(version_name = "2024_02_10_naive"))

print_tree(root_base)

Internal State

You can audit the internal state of the tool with the print_ functions.

# Print all static fields (output truncated)
slt$return_dictionaries()

# ROOTS are likely most interesting to the user.
slt$return_dictionaries(item_names = "ROOTS")

# Show the last 'action' the tool performed
# - these fields are set as part of each 'marking' new action.
slt$return_dynamic_fields()

Clean Up

# Finally, clean up all our temporary folders
system(paste("rm -rf", root_base))

options(.opt_width)
options(.opt_mccores)

vmTools documentation built on Aug. 8, 2025, 7:28 p.m.