ksCompare

Smart data-frame comparison in the spirit of SAS PROC COMPARE.

ksCompare is a tidyverse-native R package for comparing two data frames with the rigor expected of clinical / pharma QC workflows. It combines strict cell-by-cell comparison with review helpers, report generation, and pipeline-friendly assertions.

Why ksCompare exists

In clinical-trial programming, double programming is still one of the most credible ways to QC derived datasets: one programmer builds the production result, another independently recreates it, and the two outputs are compared as evidence that they truly match. In SAS-based workflows, PROC COMPARE has long been the standard tool for that final check because it turns "these datasets look similar" into something auditable and reviewable.

R has many general data-manipulation tools, but far less support for this specific QC step as a first-class workflow. ksCompare is an attempt to bring the spirit of PROC COMPARE into R and push it further: compare structure and values, surface unmatched rows, support tolerances and SAS-aware semantics, and produce interactive outputs that make dataset QC easier to review, explain, and document.

Installation

Install the package from r-universe:

install.packages(
  "ksCompare",
  repos = c(
    "https://crow16384.r-universe.dev",
    "https://cloud.r-project.org"
  )
)

Some features rely on suggested packages:

• htmltools + reactable for self-contained HTML reports.
• openxlsx2 for Excel export.
• haven and arrow for direct file-path input.
• pointblank and arsenal for workflow integrations.

What it does

• Per-column tolerances: absolute, relative, and ULP.
• Intelligent type reconciliation, including haven_labelled and SAS special missings (.A–.Z, ._).
• Automatic key inference (by = "auto") and explicit duplicate-key strategies (first, last, keep_all, all_pairs, error).
• Triage helpers: ks_glance(), ks_tidy(), ks_cause_summary(), ks_row_diff_summary(), and ks_unmatched_rows().
• Opt-in diff-pattern detection via find_patterns = TRUE.
• Rich console output (cli) plus self-contained HTML and Excel reports.
• Pipeline gates (ks_assert_clean(), ks_pointblank_step()).
• SAS-parity surface: as_outbase() / as_outcomp() / as_outdif() / as_outnoequal() and ks_sysinfo().

ks_compare() accepts both in-memory data frames and file paths for .sas7bdat, .xpt, .parquet, .feather, .rds, .RData, .csv, and .tsv inputs.

Quick start

library(ksCompare)

base <- data.frame(
  id = 1:4,
  age = c(34, 41, 28, 55),
  arm = c("A", "A", "B", "B")
)
comp <- data.frame(
  id = 1:4,
  age = c(34, 41, 27, 55),
  arm = c("A", "A", "b", "B")
)

cmp <- ks_compare(base, comp, by = "id")
cmp
ks_glance(cmp)
ks_tidy(cmp)
ks_cause_summary(cmp)
ks_row_diff_summary(cmp)

If you want recurring diff shapes summarized, turn pattern detection on:

ks_compare(base, comp, by = "id", find_patterns = TRUE)$pattern_summary

Reports

ks_report_html(cmp, "report.html")
ks_report_xlsx(cmp, "report.xlsx")

The HTML report is self-contained and does not require Quarto, Pandoc, or internet access.

SAS PROC COMPARE migration

# PROC COMPARE BASE=a COMPARE=b ID id; OUT=outbase; OUTDIF; OUTNOEQUAL;
cmp <- ks_compare(a, b, by = "id")
as_outbase(cmp)
as_outdif(cmp)
as_outnoequal(cmp)
ks_sysinfo(cmp)

See vignette("from-proc-compare") for a full mapping.

Pipeline gates

ks_compare(target, qc, by = "USUBJID") |>
  ks_assert_clean(max_value_diffs = 0L)

Learn more

• vignette("getting-started")
• vignette("from-proc-compare")
• vignette("smart-features")
• vignette("reports")
• vignette("pipeline-gates")
• Source repository: https://github.com/al-garik/ksCompare

Optional integrations

Optional integrations include ks_pointblank_step() for pointblank pipelines and as_ks_comparison() for teams moving from arsenal::comparedf().

License

MIT. See LICENSE.