fix: compound assignment operations on quantities now behave the same as on the underying representation types

Resolves #137

src/core/include/mp-units/framework/quantity.h

@@ -377,40 +377,50 @@ public:
   }
 
   // compound assignment operators
-  template<typename FwdQ, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
+  template<typename FwdQ, auto R2, typename Rep2, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
-    requires requires(rep a, rep b) {
+    requires detail::QuantityConvertibleTo<quantity<R2, rep>, quantity> && requires(rep a, Rep2 b) {
       { a += b } -> std::same_as<rep&>;
     }
-  friend constexpr decltype(auto) operator+=(FwdQ&& lhs, const quantity& rhs)
+  friend constexpr decltype(auto) operator+=(FwdQ&& lhs, const quantity<R2, Rep2>& rhs)
   {
-    lhs.numerical_value_is_an_implementation_detail_ += rhs.numerical_value_is_an_implementation_detail_;
+    if constexpr (equivalent(unit, get_unit(R2)))
+      lhs.numerical_value_is_an_implementation_detail_ += rhs.numerical_value_is_an_implementation_detail_;
+    else
+      lhs.numerical_value_is_an_implementation_detail_ += rhs.in(lhs.unit).numerical_value_is_an_implementation_detail_;
     return std::forward<FwdQ>(lhs);
   }
 
-  template<typename FwdQ, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
+  template<typename FwdQ, auto R2, typename Rep2, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
-    requires requires(rep a, rep b) {
+    requires detail::QuantityConvertibleTo<quantity<R2, rep>, quantity> && requires(rep a, Rep2 b) {
       { a -= b } -> std::same_as<rep&>;
     }
-  friend constexpr decltype(auto) operator-=(FwdQ&& lhs, const quantity& rhs)
+  friend constexpr decltype(auto) operator-=(FwdQ&& lhs, const quantity<R2, Rep2>& rhs)
   {
-    lhs.numerical_value_is_an_implementation_detail_ -= rhs.numerical_value_is_an_implementation_detail_;
+    if constexpr (equivalent(unit, get_unit(R2)))
+      lhs.numerical_value_is_an_implementation_detail_ -= rhs.numerical_value_is_an_implementation_detail_;
+    else
+      lhs.numerical_value_is_an_implementation_detail_ -= rhs.in(lhs.unit).numerical_value_is_an_implementation_detail_;
     return std::forward<FwdQ>(lhs);
   }
 
-  template<typename FwdQ, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
+  template<typename FwdQ, auto R2, typename Rep2, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
-    requires(!treat_as_floating_point<rep>) && requires(rep a, rep b) {
+    requires detail::QuantityConvertibleTo<quantity<R2, rep>, quantity> && (!treat_as_floating_point<rep>) &&
-      { a %= b } -> std::same_as<rep&>;
+             requires(rep a, Rep2 b) {
-    }
+               { a %= b } -> std::same_as<rep&>;
-  friend constexpr decltype(auto) operator%=(FwdQ&& lhs, const quantity& rhs)
+             }
+  friend constexpr decltype(auto) operator%=(FwdQ&& lhs, const quantity<R2, Rep2>& rhs)
 
   {
     MP_UNITS_EXPECTS_DEBUG(rhs != zero());
-    lhs.numerical_value_is_an_implementation_detail_ %= rhs.numerical_value_is_an_implementation_detail_;
+    if constexpr (equivalent(unit, get_unit(R2)))
+      lhs.numerical_value_is_an_implementation_detail_ %= rhs.numerical_value_is_an_implementation_detail_;
+    else
+      lhs.numerical_value_is_an_implementation_detail_ %= rhs.in(lhs.unit).numerical_value_is_an_implementation_detail_;
     return std::forward<FwdQ>(lhs);
   }
 
   template<typename FwdQ, typename Value, std::derived_from<quantity> Q = std::remove_cvref_t<FwdQ>>
-    requires(!Quantity<Value>) && requires(rep a, const Value b) {
+    requires(!Quantity<Value>) && requires(rep a, Value b) {
       { a *= b } -> std::same_as<rep&>;
     }
   friend constexpr decltype(auto) operator*=(FwdQ&& lhs, const Value& v)

test/static/chrono_test.cpp

@@ -137,8 +137,8 @@ static_assert(std::chrono::nanoseconds(quantity{1 * ns}) == 1ns);
 static_assert(std::chrono::nanoseconds(quantity{1 * s}) == 1s);
 
 // operators
-static_assert((1 * s += 1s) == 2 * s);
+static_assert((1 * s += quantity{1s}) == 2 * s);
-static_assert((2 * s -= 1s) == 1 * s);
+static_assert((2 * s -= quantity{1s}) == 1 * s);
 static_assert(quantity{1s} + 1 * s == 2 * s);
 static_assert(quantity{1s} + 1 * min == 61 * s);
 static_assert(1 * s + quantity{1s} == 2 * s);

test/static/quantity_test.cpp

@@ -456,7 +456,7 @@ static_assert((1 * m *= 2 * one).numerical_value_in(m) == 2);
 static_assert((2 * m /= 2 * one).numerical_value_in(m) == 1);
 static_assert((7 * m %= 2 * m).numerical_value_in(m) == 1);
 
-// different types
+// different representation types
 static_assert((2.5 * m += 3 * m).numerical_value_in(m) == 5.5);
 static_assert((123 * m += 1 * km).numerical_value_in(m) == 1123);
 static_assert((5.5 * m -= 3 * m).numerical_value_in(m) == 2.5);
@@ -467,12 +467,26 @@ static_assert((2.5 * m *= 3 * one).numerical_value_in(m) == 7.5);
 static_assert((7.5 * m /= 3 * one).numerical_value_in(m) == 2.5);
 static_assert((3500 * m %= 1 * km).numerical_value_in(m) == 500);
 
+// convertible quantity types
+static_assert((isq::length(1 * m) += isq::height(1 * m)).numerical_value_in(m) == 2);
+static_assert((isq::length(2 * m) -= isq::height(1 * m)).numerical_value_in(m) == 1);
+static_assert((isq::length(7 * m) %= isq::height(2 * m)).numerical_value_in(m) == 1);
+
+// different representation types with truncation
+// clang-format off
+static_assert((3 * m += 2.5 * m).numerical_value_in(m) == []{ auto v = 3; v += 2.5; return v; }());
+static_assert((3 * m -= 1.5 * m).numerical_value_in(m) == []{ auto v = 3; v -= 1.5; return v; }());
+static_assert((2 * m *= 2.5).numerical_value_in(m) == []{ auto v = 2; v *= 2.5; return v; }());
+static_assert((10 * m /= 2.5).numerical_value_in(m) == []{ auto v = 10; v /= 2.5; return v; }());
+static_assert((2 * m *= 2.5 * one).numerical_value_in(m) == []{ auto v = 2; v *= 2.5; return v; }());
+static_assert((10 * m /= 2.5 * one).numerical_value_in(m) == []{ auto v = 10; v /= 2.5; return v; }());
+// clang-format on
+
 // static_assert((std::uint8_t{255} * m %= 256 * m).numerical_value_in(m) == [] {
 //   std::uint8_t ui(255);
 //   return ui %= 256;
 // }());  // UB
-// TODO: Fix
+static_assert((std::uint8_t{255}* m %= 257 * m).numerical_value_in(m) == [] {
-static_assert((std::uint8_t{255}* m %= 257 * m).numerical_value_in(m) != [] {
   std::uint8_t ui(255);
   return ui %= 257;
 }());
@@ -490,15 +504,18 @@ static_assert((22 * m /= 3.33 * one).numerical_value_in(m) == 6);
 template<template<auto, typename> typename Q>
 concept invalid_compound_assignments = requires() {
   // truncating not allowed
-  requires !requires(Q<isq::length[m], int> l) { l += 2.5 * m; };
-  requires !requires(Q<isq::length[m], int> l) { l -= 2.5 * m; };
   requires !requires(Q<isq::length[km], int> l) { l += 2 * isq::length[m]; };
   requires !requires(Q<isq::length[km], int> l) { l -= 2 * isq::length[m]; };
   requires !requires(Q<isq::length[km], int> l) { l %= 2 * isq::length[m]; };
   requires !requires(Q<isq::length[km], int> l) { l %= 2 * percent; };
   requires !requires(Q<isq::length[km], int> l) { l %= 2. * percent; };
 
-  // TODO: accept non-truncating argument
+  // compound assignment with a non-convertible quantity not allowed
+  requires !requires(Q<isq::height[m], int> l) { l += 2 * isq::length[m]; };
+  requires !requires(Q<isq::height[m], int> l) { l -= 2 * isq::length[m]; };
+  requires !requires(Q<isq::height[m], int> l) { l %= 2 * isq::length[m]; };
+
+  // dimensionless quantities with a unit different than `one`
   requires !requires(Q<isq::length[km], int> l) { l *= 1 * (km / m); };
   requires !requires(Q<isq::length[km], int> l) { l /= 1 * (km / m); };
   requires !requires(Q<isq::length[km], int> l) { l %= 1 * (km / m); };