GH-35676: [MATLAB] Add an InferNulls name-value pair for controllin…

…g null value inference during construction of `arrow.array.Array` (#35827)

### Rationale for this change
This change lets users control toggle the automatic null-value detection behavior. By default, values MATLAB considers to be missing  (e.g. `NaN` for `double`, `<missing>` for `string`, and `NaT` for `datetime`) will be treated as `null` values. Users can toggle this behavior on and off using the `InferNulls` name-value pair. 

**Example**
```matlab
>> matlabArray = [1 NaN 3]'

matlabArray =

     1
     NaN
     3
      
% Treat NaN as a null value
 >> arrowArray1 = arrow.array.Float64Array(maltabArray, InferNulls=true)

arrowArray1 = 

[
  1,
  null,
  3
]

% Don't treat NaN as a null value 
 >> arrowArray2 = arrow.array.Float64Array(maltabArray, InferNulls=false)
   
arrowArray2 = 

[
  1,
  nan,
  3
]

```
We've only added this nv-pair to `arrow.array.Float64Array` for now. We'll add this nv-pair to the other types in a followup changelist.

### What changes are included in this PR?

1. Added `InferNulls` name-value pair to `arrow.array.Float64Array`.
2. Added common validation function `arrow.args.validateTypeAndShape` to remove duplicate validation code among the numeric classes.
3. Added a function called `arrow.args.parseValidElements` that the `arrow.array.<Type>Array` classes will be able to share for generating the logical mask of valid elements.

### Are these changes tested?

Yes, we added a test pointed called `InferNulls` to  the test class`tFloat64Array.m`.

### Are there any user-facing changes?

Yes, users can now control how `NaN` values are treated when creating an `arrow.array.Float64Array`.

### Future Directions

1. Add a name-value pair to allow users to specify the valid elements themselves.
2. Extend null support to other numeric types.
3. We've been working on adding error-handling support to `mathworks/libmexclass`. We have a prototype to do this using status-like and result-like objects already pushed to a [branch](https://github.com/mathworks/libmexclass/tree/33). Once this branch is merged with the `main` branch of `mathworks/libmexclass`, we'll port it over.

### Notes

Thank you @ kevingurney for all the help with this PR! 
* Closes: #35676

Lead-authored-by: Sarah Gilmore <sgilmore@mathworks.com>
Co-authored-by: Kevin Gurney <kgurney@mathworks.com>
Signed-off-by: Sutou Kouhei <kou@clear-code.com>

matlab/src/matlab/+arrow/+args/parseValidElements.m

@@ -0,0 +1,28 @@
+% Licensed to the Apache Software Foundation (ASF) under one or more
+% contributor license agreements.  See the NOTICE file distributed with
+% this work for additional information regarding copyright ownership.
+% The ASF licenses this file to you under the Apache License, Version
+% 2.0 (the "License"); you may not use this file except in compliance
+% with the License.  You may obtain a copy of the License at
+%
+%   http://www.apache.org/licenses/LICENSE-2.0
+%
+% Unless required by applicable law or agreed to in writing, software
+% distributed under the License is distributed on an "AS IS" BASIS,
+% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+% implied.  See the License for the specific language governing
+% permissions and limitations under the License.
+
+function validElements = parseValidElements(data, inferNulls)
+    % Returns a logical vector of the validElements in data. If inferNulls
+    % is true, calls ismissing on data to determine which elements are 
+    % null.
+
+    if inferNulls
+        % TODO: consider making validElements empty if everything is valid.
+        validElements = ~ismissing(data);
+    else
+        % TODO: consider making this an empty array if everything is valid
+        validElements = true([numel(data) 1]);
+    end
+end

matlab/src/matlab/+arrow/+args/validateTypeAndShape.m

@@ -0,0 +1,36 @@
+% Licensed to the Apache Software Foundation (ASF) under one or more
+% contributor license agreements.  See the NOTICE file distributed with
+% this work for additional information regarding copyright ownership.
+% The ASF licenses this file to you under the Apache License, Version
+% 2.0 (the "License"); you may not use this file except in compliance
+% with the License.  You may obtain a copy of the License at
+%
+%   http://www.apache.org/licenses/LICENSE-2.0
+%
+% Unless required by applicable law or agreed to in writing, software
+% distributed under the License is distributed on an "AS IS" BASIS,
+% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+% implied.  See the License for the specific language governing
+% permissions and limitations under the License.
+
+function validateTypeAndShape(data, type)
+    % Validates data has the expected type and is a vector or empty 2D
+    % matrix. If data is numeric, validates is real and nonsparse.
+
+    arguments
+        data
+        type(1, 1) string
+    end
+
+    % If data is empty, only require it's shape to be 2D to support 0x0 
+    % arrays. Otherwise, require data to be a vector.
+    %
+    % TODO: Consider supporting nonvector 2D arrays. We chould reshape them
+    % to column vectors if needed.
+
+    expectedShape = "vector";
+    if isempty(data)
+        expectedShape = "2d";
+    end
+    validateattributes(data, type, [expectedShape, "nonsparse", "real"]);
+end

matlab/src/matlab/+arrow/+array/Float32Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "single", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "single", "vector"); end
+            arrow.args.validateTypeAndShape(data, "single");
             obj@arrow.array.Array("Name", "arrow.array.proxy.Float32Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/Float64Array.m

@@ -25,16 +25,11 @@
         function obj = Float64Array(data, opts)
             arguments
                 data
-                opts.DeepCopy = false
+                opts.DeepCopy(1, 1) logical = false
+                opts.InferNulls(1, 1) logical = true
             end
-
-            validateattributes(data, "double", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "double", "vector"); end
-            % Extract missing (i.e. null) values.
-            % TODO: Determine a more robust approach to handling "detection" of null values.
-            %       For example - add a name-value pair to allow clients to choose which values
-            %       should be considered null (if any).
-            validElements = ~isnan(data);
+            arrow.args.validateTypeAndShape(data, "double");
+            validElements = arrow.args.parseValidElements(data, opts.InferNulls);
             obj@arrow.array.Array("Name", "arrow.array.proxy.Float64Array", "ConstructorArguments", {data, opts.DeepCopy, validElements});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end
@@ -50,3 +45,4 @@
         end
     end
 end
+

matlab/src/matlab/+arrow/+array/Int16Array.m

@@ -26,8 +26,8 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "int16", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "int16", "vector"); end
+
+            arrow.args.validateTypeAndShape(data, "int16");
             obj@arrow.array.Array("Name", "arrow.array.proxy.Int16Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/Int32Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "int32", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "int32", "vector"); end
+            arrow.args.validateTypeAndShape(data, "int32");
             obj@arrow.array.Array("Name", "arrow.array.proxy.Int32Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/Int64Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "int64", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "int64", "vector"); end
+            arrow.args.validateTypeAndShape(data, "int64");
             obj@arrow.array.Array("Name", "arrow.array.proxy.Int64Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/Int8Array.m

@@ -26,8 +26,8 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "int8", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "int8", "vector"); end
+
+            arrow.args.validateTypeAndShape(data, "int8");
             obj@arrow.array.Array("Name", "arrow.array.proxy.Int8Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/UInt16Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "uint16", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "uint16", "vector"); end
+            arrow.args.validateTypeAndShape(data, "uint16");
             obj@arrow.array.Array("Name", "arrow.array.proxy.UInt16Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/UInt32Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "uint32", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "uint32", "vector"); end
+            arrow.args.validateTypeAndShape(data, "uint32");
             obj@arrow.array.Array("Name", "arrow.array.proxy.UInt32Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/UInt64Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "uint64", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "uint64", "vector"); end
+            arrow.args.validateTypeAndShape(data, "uint64");
             obj@arrow.array.Array("Name", "arrow.array.proxy.UInt64Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/src/matlab/+arrow/+array/UInt8Array.m

@@ -26,8 +26,7 @@
                 data
                 opts.DeepCopy = false
             end
-            validateattributes(data, "uint8", ["2d", "nonsparse", "real"]);
-            if ~isempty(data), validateattributes(data, "uint8", "vector"); end
+            arrow.args.validateTypeAndShape(data, "uint8");
             obj@arrow.array.Array("Name", "arrow.array.proxy.UInt8Array", "ConstructorArguments", {data, opts.DeepCopy});
             % Store a reference to the array if not doing a deep copy
             if (~opts.DeepCopy), obj.MatlabArray = data; end

matlab/test/arrow/array/hNumericArray.m

@@ -117,17 +117,17 @@ function ErrorIfComplex(tc, MakeDeepCopy)
             tc.verifyError(fcn, "MATLAB:expectedReal");
         end
 
-        function ErrorIfNotTwoDimensional(tc, MakeDeepCopy)
+        function ErrorIfNonVector(tc, MakeDeepCopy)
             data = tc.MatlabArrayFcn([1 2 3 4 5 6 7 8 9]);
             data = reshape(data, 3, 1, 3);
             fcn = @() tc.ArrowArrayConstructor(tc.MatlabArrayFcn(data), DeepCopy=MakeDeepCopy);
-            tc.verifyError(fcn, "MATLAB:expected2D");
+            tc.verifyError(fcn, "MATLAB:expectedVector");
         end
 
-        function ErrorIfNonVector(tc, MakeDeepCopy)
-            data = tc.MatlabArrayFcn([1 2; 3 4]);
+        function ErrorIfEmptyArrayIsNotTwoDimensional(tc, MakeDeepCopy)
+            data = tc.MatlabArrayFcn(reshape([], [1 0 0]));
             fcn = @() tc.ArrowArrayConstructor(data, DeepCopy=MakeDeepCopy);
-            tc.verifyError(fcn, "MATLAB:expectedVector");
+            tc.verifyError(fcn, "MATLAB:expected2D");
         end
     end
 end

matlab/test/arrow/array/tFloat64Array.m

@@ -40,12 +40,29 @@ function ErrorIfSparse(testCase, MakeDeepCopy)
 
         function ValidBasic(testCase, MakeDeepCopy)
             % Create a MATLAB array with one null value (i.e. one NaN).
+            % Verify NaN is considered a null value by default.
             matlabArray = [1, NaN, 3]';
             arrowArray = arrow.array.Float64Array(matlabArray, DeepCopy=MakeDeepCopy);
             expectedValid = [true, false, true]';
             testCase.verifyEqual(arrowArray.Valid, expectedValid);
         end
 
+        function InferNulls(testCase, MakeDeepCopy)
+            matlabArray = [1, NaN, 3];
+
+            % Verify NaN is treated as a null value when InferNulls=true.
+            arrowArray1 = arrow.array.Float64Array(matlabArray, InferNulls=true, DeepCopy=MakeDeepCopy);
+            expectedValid1 = [true false true]';
+            testCase.verifyEqual(arrowArray1.Valid, expectedValid1);
+            testCase.verifyEqual(toMATLAB(arrowArray1), matlabArray');
+
+            % Verify NaN is not treated as a null value when InferNulls=false.
+            arrowArray2 = arrow.array.Float64Array(matlabArray, InferNulls=false, DeepCopy=MakeDeepCopy);
+            expectedValid2 = [true true true]';
+            testCase.verifyEqual(arrowArray2.Valid, expectedValid2);
+            testCase.verifyEqual(toMATLAB(arrowArray2), matlabArray');
+        end
+
         function ValidNoNulls(testCase, MakeDeepCopy)
             % Create a MATLAB array with no null values (i.e. no NaNs).
             matlabArray = [1, 2, 3]';
@@ -79,6 +96,5 @@ function ValidEmpty(testCase, MakeDeepCopy)
             arrowArray = arrow.array.Float64Array(matlabArray, DeepCopy=MakeDeepCopy);
             testCase.verifyEqual(arrowArray.Valid, expectedValid);
         end
-
     end
 end