name: libaom/prefer-clamp-over-nested-min-max-macros description: When constraining values to a range, use clamp(), clamp64(), fclamp(), or clip_pixel() instead of nested AOMMIN/AOMMAX macros, ensuring low <= high for dynamic bounds rubric: correctness severity: blocking

Rule. When constraining integer, 64-bit, floating-point, or pixel values between lower and upper limits, use standard inline helper functions clamp() (for int), clamp64() (for int64_t), fclamp() (for double/float), or clip_pixel() / clip_pixel_highbd() from aom_dsp/aom_dsp_common.h rather than nesting or chaining AOMMIN and AOMMAX macros. Additionally, when bounds are computed dynamically and could invert (low > high), ensure low <= high holds or sanitize the upper/lower bounds with AOMMAX/AOMMIN prior to calling clamp().

Why. AOMMIN and AOMMAX macros evaluate their arguments multiple times, risking redundant computation and accidental side effects with complex expressions. The inline helper functions evaluate arguments exactly once, provide type safety, and improve readability across the codebase. Furthermore, clamp(val, low, high) is implemented assuming low <= high; if dynamic bounds invert (low > high), calling clamp() will produce values outside the intended upper bound.

Examples

Bad:

// Nested macro calls cause redundant evaluations and side effects
const int val = AOMMAX(0, AOMMIN(QINDEX_RANGE - 1, input_val));
const double fval = AOMMAX(0.0, AOMMIN(1.0, rate_factor));

// Chained macros evaluate expressions multiple times
mv_limits->row_min = AOMMAX(mv_limits->row_min, -top_margin);
mv_limits->row_min = AOMMIN(mv_limits->row_min, mv_limits->row_max);

// If dynamic bounds can invert (-top_margin > row_max), clamp returns -top_margin (> row_max)
mv_limits->row_min = clamp(mv_limits->row_min, -top_margin, mv_limits->row_max);

Good:

// Use inline helpers for concise, single-evaluation bounding
const int val = clamp(input_val, 0, QINDEX_RANGE - 1);
const double fval = fclamp(rate_factor, 0.0, 1.0);

// Sanitize dynamic bounds to guarantee low <= high before clamping
bottom_margin = AOMMAX(bottom_margin, -top_margin);
mv_limits->row_min = AOMMAX(mv_limits->row_min, -top_margin);
mv_limits->row_max = AOMMIN(mv_limits->row_max, bottom_margin);