Infer relationship types between variables in two datasets using sampling

Description

This function compares selected variables from two data frames and infers their relational structure (e.g., one-to-one, many-to-one). It uses random sampling—either automatic or user-defined—to estimate match behavior, uniqueness patterns, and missingness characteristics. The goal is to help diagnose potential join keys or detect unrelated fields without performing full-table comparisons.

Usage

joinless(
  x,
  y,
  x_vars = NULL,
  y_vars = NULL,
  conf = 0.95,
  error = 0.05,
  n_x = NULL,
  n_y = NULL,
  max_vars = 20,
  ignore = character(0),
  missingness_tol = 0.1,
  type_coerce = TRUE,
  seed = NULL,
  verbose = FALSE,
  info = FALSE
)

Arguments

Details

Relationship inference is determined using:

• Match rate: proportion of keys in x found in y

• Key uniqueness: frequency distribution of non-missing values

Based on these, relationships are classified as:

• "one-one"

• "many-one"

• "one-many"

• "many-many"

• "unrelated" (very low or zero match rate)

• "null" (missingness above tolerance)

• "error_type" (incompatible types and coercion disabled)

Value

A data frame summarizing the inferred relationship between every variable pair. If info = FALSE, the output contains:

• x_var: variable name in x

• y_var: variable name in y

• relation_type: inferred relationship

If info = TRUE, additional columns include:

• n_used: sample size used

• match_rate: proportion of sampled values from x found in y

• null_rate_x, null_rate_y: missingness/problematic rates

• type_x, type_y: underlying storage types

• notes: diagnostic messages

Examples

df1 <- data.frame(
  id    = 1:5,
  value = 1:5
)

df2 <- data.frame(
  id    = 3:7,
  value = 3:7
)

joinless(df1, df2, x_vars = "id", y_vars = "id")
joinless(df1, df2, conf = 0.99, error = 0.02, info = TRUE)
joinless(df1, df2, ignore = "unrelated")

Infer relationship types between one dataset and multiple counterparts

Description

This function is a convenience wrapper around joinless() that compares a single dataset x against multiple datasets supplied in a list ys. Internally it calls joinless() once per dataset and row-binds the results, adding an extra column that identifies the target dataset.

Usage

joinless_multiple(
  x,
  ys,
  x_vars = NULL,
  y_vars = NULL,
  dataset_names = NULL,
  ...
)

Arguments

Details

For each dataset in ys, the function:

• optionally restricts the variables in x via x_vars,

• optionally restricts the variables in that dataset via y_vars,

• calls joinless() with the provided settings,

• tags the output with a dataset name.

When y_vars is not NULL, the function intersects y_vars with the column names of each dataset in ys. This means that:

• Variables listed in y_vars but missing in a given dataset are silently dropped for that dataset.

• If none of the variables in y_vars exist in a particular dataset, that dataset is skipped and a warning is emitted.

This behavior avoids producing "error_type" rows solely due to missing columns in some of the target datasets.

If all datasets are skipped (e.g., because none contain the requested y_vars), the function returns an empty data frame.

Value

A data frame that row-binds the result of joinless() for all target datasets that were processed. It contains all columns returned by joinless() plus an additional column:

• dataset: identifier of the target dataset (one per element of ys).

If all datasets are skipped, an empty data frame is returned.

Examples

df_base <- data.frame(id = 1:5, value = 1:5)
df_a <- data.frame(id = 3:7, value = 3:7)
df_b <- data.frame(id_alt = 1:5, value = 11:15)

# Compare the same key from df_base against multiple datasets
res <- joinless_multiple(
  x = df_base,
  ys = list(a = df_a, b = df_b),
  x_vars = "id",
  y_vars = c("id", "id_alt"),
  info = TRUE
)