fix: return NaN when the number of indexed elements is zero in stats/strided/distances/dcosine-similarity

PR-URL: #9497
Reviewed-by: Athan Reines <kgryte@gmail.com>

Author: nakul-krishnakumar

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/README.md

@@ -140,7 +140,7 @@ var z = dcosineSimilarity.ndarray( 3, x, 2, 1, y, -1, y.length-1 );
 
 ## Notes
 
--   If `N <= 0`, both functions return `0.0`.
+-   If `N <= 0`, both functions return `NaN`.
 
 </section>
 

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/docs/repl.txt

@@ -9,7 +9,7 @@
     Indexing is relative to the first index. To introduce an offset, use a typed
     array view.
 
-    If `N <= 0`, the function returns `0`.
+    If `N <= 0`, the function returns `NaN`.
 
     Parameters
     ----------

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/lib/ndarray.js

@@ -53,7 +53,7 @@ function dcosineSimilarity( N, x, strideX, offsetX, y, strideY, offsetY ) {
 	var dot;
 
 	if ( N <= 0 ) {
-		return 0.0;
+		return NaN;
 	}
 	dot = ddot( N, x, strideX, offsetX, y, strideY, offsetY );
 	xnrm = dnrm2( N, x, strideX, offsetX );

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/src/main.c

@@ -56,7 +56,7 @@ double API_SUFFIX( stdlib_strided_dcosine_similarity_ndarray )( const CBLAS_INT
 	double dot;
 
 	if ( N <= 0 ) {
-		return 0.0;
+		return 0.0 / 0.0;
 	}
 	dot = c_ddot_ndarray( N, X, strideX, offsetX, Y, strideY, offsetY );
 	xnrm = c_dnrm2_ndarray( N, X, strideX, offsetX );

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/test/test.dcosine_similarity.js

@@ -61,7 +61,7 @@ tape( 'the function calculates the cosine similarity of two strided arrays', fun
 	t.end();
 });
 
-tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `0`', function test( t ) {
+tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `NaN`', function test( t ) {
 	var sim;
 	var x;
 	var y;
@@ -70,10 +70,10 @@ tape( 'if provided an `N` parameter less than or equal to `0`, the function retu
 	y = new Float64Array( [ 1.0, -2.0, 3.0 ] );
 
 	sim = dcosineSimilarity( 0, x, 1, y, 1 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 
 	sim = dcosineSimilarity( -4, x, 1, y, 1 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 	t.end();
 });
 

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/test/test.dcosine_similarity.native.js

@@ -70,7 +70,7 @@ tape( 'the function calculates the cosine similarity of two strided arrays', opt
 	t.end();
 });
 
-tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `0`', opts, function test( t ) {
+tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `NaN`', opts, function test( t ) {
 	var sim;
 	var x;
 	var y;
@@ -80,11 +80,11 @@ tape( 'if provided an `N` parameter less than or equal to `0`, the function retu
 
 	// Test against textbook algorithm for computing cosine similarity:
 	sim = dcosineSimilarity( 0, x, 1, y, 1 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 
 	// Test against textbook algorithm for computing cosine similarity:
 	sim = dcosineSimilarity( -4, x, 1, y, 1 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 	t.end();
 });
 

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/test/test.ndarray.js

@@ -61,7 +61,7 @@ tape( 'the function calculates the cosine similarity of two strided arrays', fun
 	t.end();
 });
 
-tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `0`', function test( t ) {
+tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `NaN`', function test( t ) {
 	var sim;
 	var x;
 	var y;
@@ -71,11 +71,11 @@ tape( 'if provided an `N` parameter less than or equal to `0`, the function retu
 
 	// Test against textbook algorithm for computing cosine similarity:
 	sim = dcosineSimilarity( 0, x, 1, 0, y, 1, 0 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 
 	// Test against textbook algorithm for computing cosine similarity:
 	sim = dcosineSimilarity( -4, x, 1, 0, y, 1, 0 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 	t.end();
 });
 

lib/node_modules/@stdlib/stats/strided/distances/dcosine-similarity/test/test.ndarray.native.js

@@ -70,7 +70,7 @@ tape( 'the function calculates the cosine similarity of two strided arrays', opt
 	t.end();
 });
 
-tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `0`', opts, function test( t ) {
+tape( 'if provided an `N` parameter less than or equal to `0`, the function returns `NaN`', opts, function test( t ) {
 	var sim;
 	var x;
 	var y;
@@ -80,11 +80,11 @@ tape( 'if provided an `N` parameter less than or equal to `0`, the function retu
 
 	// Test against textbook algorithm for computing cosine similarity:
 	sim = dcosineSimilarity( 0, x, 1, 0, y, 1, 0 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 
 	// Test against textbook algorithm for computing cosine similarity:
 	sim = dcosineSimilarity( -4, x, 1, 0, y, 1, 0 );
-	t.strictEqual( isAlmostSameValue( sim, 0.0, 1 ), true, 'returns expected value' );
+	t.strictEqual( isAlmostSameValue( sim, NaN, 1 ), true, 'returns expected value' );
 	t.end();
 });