Domain Registry

Overview

The Domain Registry is the extensible system that replaces the legacy switch() block for domain generation. It defines, per domain, exactly how data should be generated: what counts to use, what arguments to pass, and which generator function to call.

Migrating a domain to the registry makes it:

• Self-describing — all logic lives in one place per domain
• Independently testable
• Overridable without touching core generation code

Inspecting the Registry

registry <- get_domain_registry()

cat("Registry-backed domains:", paste(names(registry), collapse = ", "), "\n")

# Each entry has four contract fields
cat("Raw_AE entry fields:", paste(names(registry$Raw_AE), collapse = ", "), "\n")
# > "dataset"  "required_inputs"  "count_fn"  "generate_fn"

The Entry Contract

Every valid entry must have exactly these four fields:

Field	Type	Description
$dataset	character(1)	The Raw_* name
$required_inputs	character(n)	Context keys the entry reads
$count_fn	function(counts, snapshot_idx)	Picks the right count for this snapshot
$generate_fn	function(context)	All generation logic; returns a data.frame

ae_entry <- registry$Raw_AE

cat("dataset        :", ae_entry$dataset, "\n")
cat("required_inputs:", paste(ae_entry$required_inputs, collapse = ", "), "\n")

# count_fn body
print(body(ae_entry$count_fn))

# generate_fn body (first 10 lines)
bdy <- deparse(body(ae_entry$generate_fn))
cat(paste(head(bdy, 10), collapse = "\n"), "\n  ...\n")

How the Registry Fits the Pipeline

Inside run_domain_generation_loop() the per-domain dispatch is:

migrated_data <- generate_domain_from_registry(data_type, context)
if (!is.null(migrated_data)) {
  data[[data_type]] <- migrated_data
  next   # skip the legacy dispatcher
}
# ... dispatch_legacy_domain_generator() for domains not yet in the registry

The context list passed to every generate_fn contains:

data, previous_data, combined_specs,
n, start_date, end_date,
snapshot_idx, snapshot_count, snapshot_width, study_id

You can verify the registry path end-to-end by generating a small study:

config <- create_study_config(
  study_id          = "REG-DEMO-001",
  participant_count = 50,
  site_count        = 5
) |>
  set_temporal_config(
    start_date     = "2023-01-01",
    snapshot_count = 3,
    snapshot_width = "months"
  ) |>
  add_dataset_config("Raw_AE",    enabled = TRUE) |>
  add_dataset_config("Raw_LB",    enabled = TRUE) |>
  add_dataset_config("Raw_VISIT", enabled = TRUE)

raw_data <- generate_raw_data_from_config(config, verbose = TRUE)

cat("Datasets in snapshot 1:", paste(names(raw_data[[1]]), collapse = ", "), "\n")
cat("Raw_AE rows  :", nrow(raw_data[[1]]$Raw_AE),   "\n")
cat("Raw_LB rows  :", nrow(raw_data[[1]]$Raw_LB),   "\n")
cat("Raw_VISIT rows:", nrow(raw_data[[1]]$Raw_VISIT), "\n")

Adding a New Domain to the Registry

To migrate a domain from the legacy dispatcher:

1. Write a well-typed entry list
2. Register it via raw_data_generator() override or by editing get_domain_registry() directly
3. Remove the corresponding case from dispatch_legacy_domain_generator()

# Example: migrating a hypothetical Raw_Biomarker domain
new_entry <- list(
  dataset         = "Raw_Biomarker",
  required_inputs = c("data", "previous_data", "combined_specs", "n", "start_date"),
  count_fn        = function(counts, snapshot_idx) counts$subject_count[snapshot_idx],
  generate_fn     = function(context) {
    spec          <- context$combined_specs
    data          <- context$data
    previous_data <- context$previous_data
    n             <- context$n

    if ("Raw_Biomarker" %in% names(previous_data)) {
      dataset          <- previous_data$Raw_Biomarker
      previous_row_num <- nrow(dataset)
    } else {
      dataset          <- NULL
      previous_row_num <- 0
    }

    n_new <- n - previous_row_num
    if (n_new <= 0) return(dataset)

    args <- list(
      subjid  = list(n_new, external_subjid = data$Raw_SUBJ$subjid),
      default = list(n_new, context$start_date)
    )

    as.data.frame(add_new_var_data(dataset, spec$Raw_Biomarker, args, spec$Raw_Biomarker))
  }
)

Using the Config Helper Path

Equivalent generation using the explicit config helper:

config <- create_study_config(
  study_id          = "REGISTRY-EXAMPLE-001",
  participant_count = 80,
  site_count        = 8
) |>
  set_temporal_config(
    start_date     = "2012-01-01",
    snapshot_count = 2,
    snapshot_width = "months"
  )

for (ds in c("Raw_STUDY", "Raw_SITE", "Raw_SUBJ", "Raw_ENROLL",
             "Raw_VISIT", "Raw_AE", "Raw_LB")) {
  config <- add_dataset_config(config, ds, enabled = TRUE)
}

raw_data            <- generate_study_data(config)
raw_data_via_config <- generate_raw_data_from_config(config)

cat("Snapshots:", length(raw_data), "\n")
cat("Snapshot keys:", paste(names(raw_data), collapse = ", "), "\n")

snapshot_1 <- raw_data[[1]]
cat("Datasets in snapshot 1:", paste(names(snapshot_1), collapse = ", "), "\n")
cat("Raw_AE rows:", nrow(snapshot_1$Raw_AE), "\n")
cat("Raw_LB rows:", nrow(snapshot_1$Raw_LB), "\n")

# Wrap into a study object for helper functions
study <- create_longitudinal_study_data(
  study_id = "REGISTRY-EXAMPLE-001",
  raw_data = raw_data,
  config   = list(
    participants = 80,
    sites        = 8,
    snapshots    = 2,
    interval     = "1 month",
    domains      = c("AE", "LB", "VISIT"),
    study_type   = "standard"
  )
)

summarize_longitudinal_study(study)