Ops

The ops module provides compact infix operators for common interactive and pipeline work. These operators are intentionally small, but their contracts need to be precise because infix syntax hides function calls inside expressions.

Scope

R/ops.R currently exports five operators:

Operator Role
%p% Paste two character vectors with a single space
%nin% Negated %in%, preserving base R membership semantics
%match% Case-insensitive character matching that returns indices
%map% Case-insensitive character matching that returns named values
%is% Strict identity comparison via base::identical()

Design Contract

%p% Is Safer Than Plain paste()

%p% is meant for readable string assembly. It accepts character vectors without NA values. Empty strings are allowed, and the separating space is always inserted.

Length handling is intentionally stricter than paste():

  • equal lengths are allowed;
  • one side may have length 1 and be recycled;
  • other unequal lengths error.

This avoids accidental partial recycling in labels, annotations, and messages.

%nin% Mirrors Base R

%nin% is exactly !(x %in% table). It accepts any type, follows base R coercion rules, and preserves base R behavior for NA and empty vectors.

This operator should stay boring. Its value is readability, not new semantics.

%match% And %map% Are Strict Character Tools

%match% and %map% are designed for case-insensitive character matching, especially gene-symbol style workflows.

Both sides must be non-empty character vectors without NA or empty string values. Empty query or table values usually indicate dirty upstream data, so the operators fail early.

Both operators normalize with tolower(). If table contains duplicated values after normalization, they warn and use the first match.

%map% Drops Unmatched Values

%map% is not a length-preserving mapper. It returns only successful matches as a named character vector:

  • names are canonical entries from table;
  • values are original entries from x;
  • output order follows x;
  • unmatched entries are dropped.

This makes %map% useful for building compact alias-to-canonical mappings. Use %match% when unmatched positions need to be retained as NA.

%is% Is Just Identity

%is% wraps base::identical(). It accepts any object and returns one logical value. It should remain a thin readability helper rather than gaining custom comparison rules.

Review Notes

The latest review focused on four issues:

  1. %p% allowed non-length-1 unequal recycling through paste().
  2. %match% and %map% rejected NA and character(0) but allowed empty strings.
  3. %match% and %map% silently used the first table value when case normalization created duplicates.
  4. %match% documentation had a small grammar issue: “a upstream” should be “an upstream”.

The fixes made the operator contracts more explicit while preserving the existing high-level design.

Tests

The focused ops test suite lives in tests/testthat/test-ops.R.

Latest focused run:

devtools::test(filter = "ops")
[ FAIL 0 | WARN 0 | SKIP 0 | PASS 63 ]

The important tests are contract tests:

  • %p% rejects incompatible vector lengths;
  • %match% and %map% reject empty strings;
  • %match% and %map% warn on duplicated normalized table values;
  • %nin% continues to mirror base R behavior for NA, empty vectors, and type coercion;
  • %is% remains strict identity comparison.

Open Questions

  • Whether %match% and %map% should eventually use Unicode-aware case folding instead of tolower().
  • Whether %map% should have a length-preserving companion if future workflows need NA placeholders instead of dropped unmatched entries.
  • Whether duplicated normalized table values should remain a warning or become an error in stricter bioinformatics workflows.