Detect sensitive columns by name

Description

Uses a broad, configurable regex library to match likely PII columns. You can extend it with extra_patterns (they get ORed in) or replace everything with a single override_regex.

Usage

detect_sensitive_columns(x_names, extra_patterns = NULL, override_regex = NULL)

Arguments

Value

Character vector of names from x_names that matched.

Examples

detect_sensitive_columns(c("id","email","home_phone","zip","notes"))
detect_sensitive_columns(names(mtcars), extra_patterns = c("^vin$", "passport"))

Save a fake dataset to disk

Description

Save a data.frame to CSV, RDS, or Parquet based on the file extension.

Usage

export_fake(x, path)

Arguments

Value

(Invisibly) the path written.

Generate Fake Data from Real Dataset Structure

Description

Generate Fake Data from Real Dataset Structure

Usage

generate_fake_data(
  data,
  n = 30,
  category_mode = c("preserve", "generic", "custom"),
  numeric_mode = c("range", "distribution"),
  column_mode = c("keep", "generic", "custom"),
  custom_levels = NULL,
  custom_names = NULL,
  seed = NULL,
  verbose = FALSE,
  sensitive = NULL,
  sensitive_detect = TRUE,
  sensitive_strategy = c("fake", "drop"),
  normalize = TRUE
)

Arguments

Value

A data.frame of n rows with attributes:

• name_map (named chr: original -> output)

• column_mode (chr)

• sensitive_columns (chr; original names)

• dropped_columns (chr; original names that were dropped)

Generate fake data from a DB schema data.frame

Description

Generate fake data from a DB schema data.frame

Usage

generate_fake_from_schema(sch_df, n = 30, seed = NULL)

Arguments

Value

A base data.frame with n rows and one column per schema entry. Column classes follow the schema type values (integer, numeric, character, logical, Date, POSIXct); missingness is injected when nullable is TRUE.

Generate a Fake POSIXct Column

Description

Create synthetic timestamps either by mimicking an existing POSIXct vector (using its range and NA rate) or by sampling uniformly between start and end.

Usage

generate_fake_posixct_column(
  like = NULL,
  n = NULL,
  start = NULL,
  end = NULL,
  tz = "UTC",
  na_prop = NULL
)

Arguments

Value

A POSIXct vector of length n.

Generate fake data with privacy controls

Description

Generates a synthetic copy of data, then optionally detects/handles sensitive columns by name. Detection uses the ORIGINAL column names and maps to output via attr(fake, "name_map") if present.

Usage

generate_fake_with_privacy(
  data,
  n = 30,
  level = c("low", "medium", "high"),
  seed = NULL,
  sensitive = NULL,
  sensitive_detect = TRUE,
  sensitive_strategy = c("fake", "drop"),
  normalize = TRUE,
  sensitive_patterns = NULL,
  sensitive_regex = NULL
)

Arguments

Details

Generate fake data with privacy controls

Value

data.frame with attributes: sensitive_columns, dropped_columns, name_map

Create a copy-paste prompt for LLMs

Description

Create a copy-paste prompt for LLMs

Usage

generate_llm_prompt(
  fake_path,
  schema_path = NULL,
  notes = NULL,
  write_file = TRUE,
  path = dirname(fake_path),
  filename = "README_FOR_LLM.txt"
)

Arguments

Value

The prompt string (invisibly returns the file path if written).

Create a fake-data bundle for LLM workflows

Description

Generates fake data, writes files (CSV/RDS/Parquet), writes a scrubbed JSON schema, and optionally writes a README prompt and a single ZIP file containing everything.

Usage

llm_bundle(
  data,
  n = 30,
  level = c("medium", "low", "high"),
  formats = c("csv", "rds"),
  path = tempdir(),
  filename = "fake_bundle",
  seed = NULL,
  write_prompt = TRUE,
  zip = FALSE,
  prompt_filename = "README_FOR_LLM.txt",
  zip_filename = NULL,
  sensitive = NULL,
  sensitive_detect = TRUE,
  sensitive_strategy = c("fake", "drop"),
  normalize = FALSE
)

Arguments

Details

Tips Avoid using angle brackets in examples; prefer plain tokens like NAME or FILE_NAME. If you truly want bracket glyphs, use Unicode ⟨name⟩ ⟩name⟨.

Value

List with paths: $data_paths (named), $schema_path, $readme_path (optional), $zip_path (optional), and $fake (data.frame).

Build an LLM bundle directly from a database table

Description

Reads just the schema from table on conn, synthesizes n fake rows, writes a schema JSON, fake dataset(s), and a README prompt, and optionally zips them into a single archive.

Usage

llm_bundle_from_db(
  conn,
  table,
  n = 30,
  level = c("medium", "low", "high"),
  formats = c("csv", "rds"),
  path = tempdir(),
  filename = "fake_from_db",
  seed = NULL,
  write_prompt = TRUE,
  zip = FALSE,
  zip_filename = NULL,
  sensitive_strategy = c("fake", "drop")
)

Arguments

Value

Invisibly, a list with useful paths:

• schema_path – schema JSON

• files – vector of written fake-data files

• zip_path – zip archive path (if zip = TRUE)

Examples

if (requireNamespace("DBI", quietly = TRUE) &&
    requireNamespace("RSQLite", quietly = TRUE)) {
  con <- DBI::dbConnect(RSQLite::SQLite(), ":memory:")
  on.exit(DBI::dbDisconnect(con), add = TRUE)
  DBI::dbWriteTable(con, "cars", head(cars, 20), overwrite = TRUE)
  out <- llm_bundle_from_db(
    con, "cars",
    n = 100, level = "medium",
    formats = c("csv","rds"),
    path = tempdir(), filename = "db_bundle",
    seed = 1, write_prompt = TRUE, zip = TRUE
  )
}

Prepare Input Data: Coerce to data.frame and (optionally) normalize values

Description

Converts common tabular objects to a base data.frame, and if normalize = TRUE it applies light, conservative value normalization:

• Converts common date/time strings to POSIXct (best-effort across several formats)

• Converts percent-like character columns (e.g. "85%") to numeric (85)

• Maps a configurable set of "NA-like" strings to NA, while keeping common survey responses like "not applicable" or "prefer not to answer" as real levels

• Normalizes yes/no character columns to an ordered factor c("no","yes")

Usage

prepare_input_data(
  data,
  normalize = TRUE,
  na_strings = c("", "NA", "N/A", "na", "No data", "no data"),
  keep_as_levels = c("not applicable", "prefer not to answer", "unsure"),
  percent_detect_threshold = 0.6,
  datetime_formats = c("%m/%d/%Y %H:%M:%S", "%m/%d/%Y %H:%M",
    "%Y-%m-%d %H:%M:%S", "%Y-%m-%d %H:%M", "%Y-%m-%dT%H:%M:%S",
    "%Y-%m-%dT%H:%M", "%m/%d/%Y", "%Y-%m-%d")
)

Arguments

Value

A base data.frame.

Extract a table schema from a DB connection

Description

Returns a data frame describing the columns of a database table.

Usage

schema_from_db(conn, table, level = c("medium", "low", "high"))

Arguments

Value

A data.frame with column metadata (e.g., name, type).

Examples

if (requireNamespace("DBI", quietly = TRUE) &&
    requireNamespace("RSQLite", quietly = TRUE)) {
  con <- DBI::dbConnect(RSQLite::SQLite(), ":memory:")
  on.exit(DBI::dbDisconnect(con), add = TRUE)
  DBI::dbWriteTable(con, "mtcars", mtcars[1:3, ])
  sc <- schema_from_db(con, "mtcars")
  head(sc)
}

Validate a fake dataset against the original

Description

Compares classes, NA/blank proportions, and simple numeric ranges.

Usage

validate_fake(original, fake, tol = 0.15)

Arguments

Value

data.frame summary by column

Zip a set of files for easy sharing

Description

Zip a set of files for easy sharing

Usage

zip_llm_bundle(files, zipfile)

Arguments

Value

The path to the created zip file.