fix: update hypotf to follow the IEEE 754-2019 standard

PR-URL: #6511
Reviewed-by: Athan Reines <[email protected]>

Author: anandkaranubc

lib/node_modules/@stdlib/math/base/special/hypotf/README.md

@@ -52,7 +52,7 @@ h = hypotf( -0.0, -0.0 );
 // returns +0.0
 ```
 
-If either argument is `NaN`, the function returns `NaN`.
+If either argument is `NaN` and the other argument is not `+-Infinity`, the function returns `NaN`.
 
 ```javascript
 var h = hypotf( NaN, 12.0 );

lib/node_modules/@stdlib/math/base/special/hypotf/docs/repl.txt

@@ -2,7 +2,8 @@
 {{alias}}( x, y )
     Computes the hypotenuse avoiding overflow and underflow (single-precision).
 
-    If either argument is `NaN`, the function returns `NaN`.
+    If either argument is `NaN` and the other argument is not `+-Infinity`,
+    the function returns `NaN`.
 
     Parameters
     ----------

lib/node_modules/@stdlib/math/base/special/hypotf/lib/main.js

@@ -24,7 +24,7 @@ var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' );
 var isnanf = require( '@stdlib/math/base/assert/is-nanf' );
 var isInfinitef = require( '@stdlib/math/base/assert/is-infinitef' );
 var PINF = require( '@stdlib/constants/float32/pinf' );
-var sqrt = require( '@stdlib/math/base/special/sqrt' );
+var sqrtf = require( '@stdlib/math/base/special/sqrtf' );
 
 
 // MAIN //
@@ -50,12 +50,14 @@ var sqrt = require( '@stdlib/math/base/special/sqrt' );
 */
 function hypotf( x, y ) {
 	var tmp;
-	if ( isnanf( x ) || isnanf( y ) ) {
-		return NaN;
-	}
+
+	// If one of the arguments is `+-infinity`, return `+infinity` even if the other argument is `NaN` (IEEE 754-2019)...
 	if ( isInfinitef( x ) || isInfinitef( y ) ) {
 		return PINF;
 	}
+	if ( isnanf( x ) || isnanf( y ) ) {
+		return NaN;
+	}
 	x = float64ToFloat32( x );
 	y = float64ToFloat32( y );
 	if ( x < 0.0 ) {
@@ -73,7 +75,7 @@ function hypotf( x, y ) {
 		return 0.0;
 	}
 	y = float64ToFloat32( y / x );
-	return float64ToFloat32( x * float64ToFloat32( sqrt( float64ToFloat32( 1.0 + float64ToFloat32(y*y) ) ) ) ); // eslint-disable-line max-len
+	return float64ToFloat32( x * sqrtf( float64ToFloat32( 1.0 + float64ToFloat32(y*y) ) ) ); // eslint-disable-line max-len
 }
 
 

lib/node_modules/@stdlib/math/base/special/hypotf/src/main.c

@@ -37,12 +37,14 @@ float stdlib_base_hypotf( const float x, const float y ) {
 	float tmp;
 	float a;
 	float b;
-	if ( stdlib_base_is_nanf( x ) || stdlib_base_is_nanf( y ) ) {
-		return 0.0f / 0.0f; // NaN
-	}
+
+	// If one of the arguments is `+-infinity`, return `+infinity` even if the other argument is `NaN` (IEEE 754-2019)...
 	if ( stdlib_base_is_infinitef( x ) || stdlib_base_is_infinitef( y ) ) {
 		return STDLIB_CONSTANT_FLOAT32_PINF;
 	}
+	if ( stdlib_base_is_nanf( x ) || stdlib_base_is_nanf( y ) ) {
+		return 0.0f / 0.0f; // NaN
+	}
 	a = x;
 	b = y;
 	if ( a < 0.0f ) {

lib/node_modules/@stdlib/math/base/special/hypotf/test/test.js

@@ -73,10 +73,22 @@ tape( 'the function returns `+infinity` if either argument is `+-infinity`', fun
 	h = hypotf( NINF, NINF );
 	t.strictEqual( h, PINF, 'returns expected value' );
 
+	h = hypotf( NaN, PINF );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
+	h = hypotf( PINF, NaN );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
+	h = hypotf( NINF, NaN );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
+	h = hypotf( NaN, NINF );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
 	t.end();
 });
 
-tape( 'the function returns `NaN` if either argument is `NaN`', function test( t ) {
+tape( 'the function returns `NaN` if either argument is `NaN` but not `+-infinity`', function test( t ) {
 	var h;
 
 	h = hypotf( NaN, 3.14 );

lib/node_modules/@stdlib/math/base/special/hypotf/test/test.native.js

@@ -82,10 +82,22 @@ tape( 'the function returns `+infinity` if either argument is `+-infinity`', opt
 	h = hypotf( NINF, NINF );
 	t.strictEqual( h, PINF, 'returns expected value' );
 
+	h = hypotf( NaN, PINF );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
+	h = hypotf( PINF, NaN );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
+	h = hypotf( NINF, NaN );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
+	h = hypotf( NaN, NINF );
+	t.strictEqual( h, PINF, 'returns expected value' );
+
 	t.end();
 });
 
-tape( 'the function returns `NaN` if either argument is `NaN`', opts, function test( t ) {
+tape( 'the function returns `NaN` if either argument is `NaN` but not `+-infinity`', opts, function test( t ) {
 	var h;
 
 	h = hypotf( NaN, 3.14 );
@@ -145,6 +157,8 @@ tape( 'the function computes the hypotenuse', opts, function test( t ) {
 });
 
 tape( 'the function computes the hypotenuse (canonical inputs)', opts, function test( t ) {
+	var delta;
+	var tol;
 	var h;
 
 	h = hypotf( 3.0, 4.0 );
@@ -153,8 +167,11 @@ tape( 'the function computes the hypotenuse (canonical inputs)', opts, function
 	h = hypotf( 6.0, 8.0 );
 	t.strictEqual( h, 10.0, 'returns expected value' );
 
+	// NOTE: the tolerance here is larger than for the JavaScript implementation due to compiler optimizations which may be performed resulting in result divergence. For discussion, see https://github.com/stdlib-js/stdlib/pull/2298#discussion_r1624765205
 	h = hypotf( 5.0, 12.0 );
-	t.strictEqual( h, 13.0, 'returns expected value' );
+	delta = absf( h - 13.0 );
+	tol = EPS * absf( 13.0 );
+	t.strictEqual( delta <= tol, true, 'within tolerance. h: '+h+'. Expected: 13.0. Delta: '+delta+'. Tol: '+tol+'.' );
 
 	t.end();
 });