← SASnRMimicking the behavior of PROC CONTENTS

Introduction

\r\n

In SAS, PROC CONTENTS is the go-to tool for describing a dataset — it returns a metadata listing with variable name, position, type, length and dataset-level facts like the number of observations.
We use it to audit newly-received data, compare structure between production and QC, and document expected variables against a spec.
R does not ship with a single function that returns the same shape. Between str(), names(), class(), sapply() and nrow(), the raw material is all there — but we have to assemble it ourselves when we want a clean, tabular result similar to PROC CONTENTS out=.
In this lesson we write a small helper function that takes a data frame and returns a one-row-per-variable tibble (or data frame) with columns: memname, name, varnum, type, length, nobs.
We build the helper twice — once in tidyverse and once in base R — and compare the output against the real PROC CONTENTS result from SAS.
After running, we confirm the R output has one row per variable in the input data frame, with the expected variable names in name and a sensible value in type (character, numeric, etc.).

Mimicking the behavior of PROC CONTENTS

SAS (Base SAS)+

*==============================================================================;
* Step 1: create a small DM-like dataset                                       ;
*==============================================================================;
data dm;
  infile cards truncover dlm="|";
  length domain $2 usubjid $5 sex $3;
  input usubjid $ sex $ age;
  domain = "DM";
  label usubjid = "Unique Subject Identifier";
  cards;
1001|M|72
1002|F|65
1003|F|45
1004|M|40
1005|F|35
1006|F|6
1007|F|5
;
run;

*==============================================================================;
* Step 2: run PROC CONTENTS and capture the metadata in a dataset              ;
*==============================================================================;
proc contents data = dm out = meta01 noprint;
run;

*------------------------------------------------------------------------------;
* Step 3: print the metadata columns we care about                            ;
*------------------------------------------------------------------------------;
proc print data = meta01 noobs;
  var memname name varnum type length nobs;
run;

To get a dataset-level metadata listing in SAS, we run PROC CONTENTS on the target dataset. By default it writes a listing to the output window.
When we want the metadata as a SAS dataset (so we can filter or join it downstream), we add the out= option — PROC CONTENTS then writes one row per variable to the named output dataset.
We add noprint so PROC CONTENTS does not also dump the listing to the output window when all we want is the meta01 dataset.
The key columns in the out= dataset are memname (dataset name), name (variable name), varnum (position), type (1 = numeric, 2 = character), length (storage length in bytes) and nobs (number of observations in the parent dataset).
We then print the subset of columns we care about with PROC PRINT, which gives us a clean tabular view — exactly the shape we aim to reproduce in R.
One gotcha: the type column in meta01 uses 1/2 encoding, not the words “numeric” and “character”. In the R versions below we emit readable class names (character, numeric) instead, which is friendlier for a listing.

R (tidyverse)+

library(tidyverse)

#==============================================================================
# Step 1: create the same small DM-like dataset
#==============================================================================
dm <- tribble(
  ~usubjid, ~sex, ~age,
  "1001",   "M",  72,
  "1002",   "F",  65,
  "1003",   "F",  45,
  "1004",   "M",  40,
  "1005",   "F",  35,
  "1006",   "F",   6,
  "1007",   "F",   5
) %>%
  mutate(domain = "DM") %>%
  relocate(domain, .before = usubjid)

#==============================================================================
# Step 2: define a helper that mimics PROC CONTENTS out=
#==============================================================================
proc_contents_out <- function(data) {
  memname <- deparse(substitute(data))

  tibble(
    memname = memname,
    name    = names(data),
    varnum  = seq_along(data),
    type    = data %>% map_chr(~ class(.x)[1]),
    length  = data %>% map_dbl(~ {
      if (is.character(.x)) {
        suppressWarnings(max(nchar(.x), na.rm = TRUE))
      } else if (is.numeric(.x)) {
        8
      } else {
        NA_real_
      }
    }),
    nobs    = nrow(data)
  )
}

#------------------------------------------------------------------------------
# Step 3: call the helper and inspect the result
#------------------------------------------------------------------------------
meta_dm <- proc_contents_out(dm)
print(meta_dm)

To mimic PROC CONTENTS in tidyverse, we wrap the logic in a small function that takes a data frame and returns a tibble with one row per variable.
deparse(substitute(data)) captures the name the caller passed in (e.g. dm), so memname in the output matches the SAS convention of the parent dataset name.
names(data) and seq_along(data) give us name and varnum — R counts columns from 1, same as SAS variable position.
For type, we use map_chr(~ class(.x)[1]) — that grabs the first class of each column (character, numeric, integer, Date, etc.). This is friendlier than the SAS 1/2 numeric code.
For length, we emulate the SAS storage-length idea: for character columns we take the max nchar (the shortest storage length that could hold any value), and for numeric columns we use 8 — the default numeric length in SAS. For other classes we return NA_real_.
nrow(data) gives us nobs — the same observation count PROC CONTENTS reports.
After running, we expect meta_dm to have four rows (domain, usubjid, sex, age) with varnum 1 through 4, type showing character for the first three and numeric for age, and nobs = 7.
The big win over the built-in str() is that the output is a real tibble — we can filter, join to a spec, export to Excel, or feed into a comparison step.

Happy learning!

Search lessons

Introduction

Mimicking the behavior of PROC CONTENTS