After review

src/include/units/bits/constexpr_math.h

@@ -23,40 +23,41 @@
 #pragma once
 
 #include <gsl/gsl_assert>
+#include <units/bits/math_concepts.h>
 #include <units/bits/pow.h>
 #include <units/bits/ratio_maths.h>
 
 namespace units::detail {
 
 struct decimal_fp {
-  double significand;
+  double significant;
-  std::intmax_t mantissa;
+  std::intmax_t exponent;
 };
 
 [[nodiscard]] constexpr decimal_fp to_decimal(double v) noexcept
 {
   if (v == 0) {
-    return {.significand = 0.0, .mantissa = 0};
+    return {.significant = 0.0, .exponent = 0};
   }
 
-  double significand = abs(v);
+  double significant = abs(v);
-  std::intmax_t mantissa = 0;
+  std::intmax_t exponent = 0;
 
-  while (significand < 1) {
+  while (significant < 1) {
-    significand *= 10.0;
+    significant *= 10.0;
-    --mantissa;
+    --exponent;
   }
 
-  while (significand >= 10) {
+  while (significant >= 10) {
-    significand /= 10.0;
+    significant /= 10.0;
-    ++mantissa;
+    ++exponent;
   }
 
   if (v < 0) {
-    significand = -significand;
+    significant = -significant;
   }
 
-  return {.significand = significand, .mantissa = mantissa};
+  return {.significant = significant, .exponent = exponent};
 }
 
 /* approximate natural log as https://math.stackexchange.com/a/977836
@@ -66,8 +67,8 @@ struct decimal_fp {
 {
   Expects(v > 0);
 
-  // lookup table to speed up convergence for all significand values
+  // lookup table to speed up convergence for all significant values
-  // significand values of 7 and greater benefit mostly as they now converge in 5 terms compared to O(10)-O(100)
+  // significant values of 7 and greater benefit mostly as they now converge in 5 terms compared to O(10)-O(100)
   // required without the table
   //
   // using python:
@@ -177,14 +178,14 @@ struct decimal_fp {
   };
   decimal_fp x = to_decimal(v);
 
-  // dividing the significand by nearest lower value in [1.0, 1.1, 1.2, ..., 9.9] will greatly improve convergence
+  // dividing the significant by nearest lower value in [1.0, 1.1, 1.2, ..., 9.9] will greatly improve convergence
-  x.significand *= 10;
+  x.significant *= 10;
-  const auto isignificand = static_cast<std::size_t>(x.significand);
+  const auto isignificant = static_cast<std::size_t>(x.significant);
-  x.significand /= static_cast<double>(isignificand);
+  x.significant /= static_cast<double>(isignificant);
-  const double result = static_cast<double>(x.mantissa - 1) * log_table[9] + log_table[isignificand - 1];
+  const double result = static_cast<double>(x.exponent - 1) * log_table[9] + log_table[isignificant - 1];
 
-  // 1.0 <= significand < 1.1 converges rapidly
+  // 1.0 <= significant < 1.1 converges rapidly
-  const double y = (x.significand - 1) / (x.significand + 1);
+  const double y = (x.significant - 1) / (x.significant + 1);
   const double y_squared = y * y;
   double sum = 0;
   // 5 terms are needed for convergence to machine precision in the worst case scenario
@@ -201,7 +202,8 @@ struct decimal_fp {
  larger Factor values improve convergence for all values but reduce the precision
 */
 template<std::size_t N, std::intmax_t Factor = 256>
-[[nodiscard]] constexpr double constexpr_exp(double v) noexcept requires requires { Factor > 0; }
+  requires gt_zero<Factor>
+[[nodiscard]] constexpr double constexpr_exp(double v) noexcept
 {
   if constexpr (N == 0) {
     return 1.0;

src/include/units/bits/math_concepts.h

@@ -0,0 +1,35 @@
+// The MIT License (MIT)
+//
+// Copyright (c) 2018 Mateusz Pusz
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+
+#pragma once
+
+#include <cstdint>
+
+namespace units::detail {
+
+template<std::intmax_t N>
+concept gt_zero = (N > 0);
+
+template<std::intmax_t N>
+concept non_zero = (N != 0);
+
+}  // namespace units::detail

src/include/units/bits/root.h

@@ -24,6 +24,7 @@
 
 #include <gsl/gsl_assert>
 #include <units/bits/constexpr_math.h>
+#include <units/bits/math_concepts.h>
 #include <units/bits/pow.h>
 #include <units/bits/ratio_maths.h>
 #include <cmath>
@@ -31,7 +32,8 @@
 namespace units::detail {
 
 template<std::intmax_t N, typename F>
-[[nodiscard]] constexpr std::intmax_t iroot_impl(std::intmax_t v, F const& pow_function) noexcept requires requires { N > 0; }
+  requires gt_zero<N>
+[[nodiscard]] constexpr std::intmax_t iroot_impl(std::intmax_t v, F const& pow_function) noexcept
 {
   if constexpr (N == 1) {
     return v;
@@ -61,13 +63,15 @@ template<std::intmax_t N, typename F>
 // Factor = 32 needs quite a few more terms to converge
 // https://godbolt.org/z/odWq1o
 template<std::intmax_t N, std::size_t ExpOrder = 12, std::intmax_t Factor = 64>
-[[nodiscard]] constexpr std::intmax_t iroot_compile(std::intmax_t v) noexcept requires requires { N > 0; }
+  requires gt_zero<N>
+[[nodiscard]] constexpr std::intmax_t iroot_compile(std::intmax_t v) noexcept
 {
   return iroot_impl<N>(v, [](double x, double exponent) { return constexpr_pow<ExpOrder, Factor>(x, exponent); });
 }
 
 template<std::intmax_t N>
-[[nodiscard]] std::intmax_t iroot_runtime(std::intmax_t v) noexcept requires requires { N > 0; }
+  requires gt_zero<N>
+[[nodiscard]] std::intmax_t iroot_runtime(std::intmax_t v) noexcept
 {
   return iroot_impl<N>(v, [](double x, double exponent) {
     if constexpr (N == 2) {
@@ -81,7 +85,8 @@ template<std::intmax_t N>
 }
 
 template<std::intmax_t N>
-[[nodiscard]] constexpr std::intmax_t iroot(std::intmax_t v) noexcept requires requires { N > 0; }
+  requires gt_zero<N>
+[[nodiscard]] constexpr std::intmax_t iroot(std::intmax_t v) noexcept
 {
   // compile time version is much slower, use faster version at runtime
   if (std::is_constant_evaluated()) {

src/include/units/math.h

@@ -40,8 +40,8 @@ namespace units {
  * @return Quantity The result of computation
  */
 template<std::intmax_t Num, std::intmax_t Den = 1, Quantity Q>
-[[nodiscard]] inline auto pow(const Q& q) noexcept
+  requires detail::non_zero<Den>
-  requires requires { std::pow(q.count(), 1.0); Den != 0; }
+[[nodiscard]] inline auto pow(const Q& q) noexcept requires requires { std::pow(q.count(), 1.0); }
 {
   using rep = TYPENAME Q::rep;
   if constexpr (Num == 0) {

src/include/units/ratio.h

@@ -23,6 +23,7 @@
 #pragma once
 
 #include <units/bits/external/hacks.h>
+#include <units/bits/math_concepts.h>
 #include <units/bits/ratio_maths.h>
 #include <units/bits/pow.h>
 #include <units/bits/root.h>
@@ -107,7 +108,8 @@ namespace detail {
 }
 
 template<std::intmax_t N>
-[[nodiscard]] constexpr ratio root(const ratio& r) requires requires { N > 0; }
+  requires gt_zero<N>
+[[nodiscard]] constexpr ratio root(const ratio& r)
 {
   if constexpr (N == 1) {
     return r;
@@ -124,7 +126,8 @@ template<std::intmax_t N>
 }  // namespace detail
 
 template<std::intmax_t Num, std::intmax_t Den = 1>
-[[nodiscard]] constexpr ratio pow(const ratio& r) requires requires { Den != 0; }
+  requires detail::non_zero<Den>
+[[nodiscard]] constexpr ratio pow(const ratio& r)
 {
   if constexpr (Num == 0) {
     return ratio(1);