apacheGH-37592: [MATLAB] Add NumRows property to `arrow.tabular.Rec…

…ordBatch` (apache#38215) ### Rationale for this change Currently, there is a `NumColumns` property on `arrow.tabular.RecordBatch`, but no `NumRows` property. It would be useful to be able to query the number of rows in a `RecordBatch`. This pull request adds a `NumRows` property to `arrow.tabular.RecordBatch` to mirror the design of `arrow.tabular.Table`. ### What changes are included in this PR? 1. Added new `NumRows` property to `arrow.tabular.RecordBatch` **Example** ```matlab >> matlabTable = array2table(rand(10, 5)) matlabTable = 10x5 table Var1 Var2 Var3 Var4 Var5 ________ ________ _______ _______ ________ 0.76062 0.12009 0.98898 0.29974 0.42165 0.64994 0.85116 0.71768 0.58693 0.31061 0.33593 0.87823 0.87766 0.38206 0.45742 0.031364 0.8336 0.71528 0.14987 0.3618 0.5986 0.81193 0.25784 0.21073 0.76715 0.46493 0.40281 0.39729 0.16737 0.94521 0.18738 0.16351 0.46437 0.45545 0.40774 0.67682 0.3577 0.94882 0.1295 0.022501 0.29368 0.47122 0.99682 0.46011 0.34275 0.6849 0.064717 0.89719 0.38302 0.4523 >> arrowRecordBatch = arrow.recordBatch(matlabTable); >> arrowRecordBatch.NumRows ans = int64 10 ``` ### Are these changes tested? Yes. 1. Added `NumRows` test to `tRecordBatch` test class. 3. Updated `EmptyTable` test (renamed to `EmptyRecordBatch`) in `tRecordBatch` test class. 4. Added `FromArraysNoInputs` test to mirror the `FromArraysNoInputs` test in `tTable` test class. ### Are there any user-facing changes? Yes. This pull request adds a new public `NumRows` property to the `arrow.tabular.RecordBatch` class. Users can query the number of rows in an `arrow.tabular.RecordBatch` by accessing the `NumRows` property. ### Future Directions 1. apache#38214 3. apache#38213 * Closes: apache#37592 Authored-by: Kevin Gurney <kgurney@mathworks.com> Signed-off-by: Kevin Gurney <kgurney@mathworks.com>
JerAguilon · Oct 23, 2023 · 7567717 · 7567717
1 parent 57eb420
commit 7567717
Show file tree

Hide file tree

Showing 4 changed files with 58 additions and 5 deletions.
diff --git a/matlab/src/cpp/arrow/matlab/tabular/proxy/record_batch.cc b/matlab/src/cpp/arrow/matlab/tabular/proxy/record_batch.cc
@@ -52,6 +52,7 @@ namespace arrow::matlab::tabular::proxy {
 
     RecordBatch::RecordBatch(std::shared_ptr<arrow::RecordBatch> record_batch) : record_batch{record_batch} {
         REGISTER_METHOD(RecordBatch, toString);
+        REGISTER_METHOD(RecordBatch, getNumRows);
         REGISTER_METHOD(RecordBatch, getNumColumns);
         REGISTER_METHOD(RecordBatch, getColumnNames);
         REGISTER_METHOD(RecordBatch, getColumnByIndex);
@@ -104,6 +105,14 @@ namespace arrow::matlab::tabular::proxy {
         return record_batch_proxy;
     }
 
+    void RecordBatch::getNumRows(libmexclass::proxy::method::Context& context) {
+        namespace mda = ::matlab::data;
+        mda::ArrayFactory factory;
+        const auto num_rows = record_batch->num_rows();
+        auto num_rows_mda = factory.createScalar(num_rows);
+        context.outputs[0] = num_rows_mda;
+    }
+
     void RecordBatch::getNumColumns(libmexclass::proxy::method::Context& context) {
         namespace mda = ::matlab::data;
         mda::ArrayFactory factory;

diff --git a/matlab/src/cpp/arrow/matlab/tabular/proxy/record_batch.h b/matlab/src/cpp/arrow/matlab/tabular/proxy/record_batch.h
@@ -35,6 +35,7 @@ namespace arrow::matlab::tabular::proxy {
 
         protected:
             void toString(libmexclass::proxy::method::Context& context);
+            void getNumRows(libmexclass::proxy::method::Context& context);
             void getNumColumns(libmexclass::proxy::method::Context& context);
             void getColumnNames(libmexclass::proxy::method::Context& context);
             void getColumnByIndex(libmexclass::proxy::method::Context& context);

diff --git a/matlab/src/matlab/+arrow/+tabular/RecordBatch.m b/matlab/src/matlab/+arrow/+tabular/RecordBatch.m
@@ -21,6 +21,7 @@
                        matlab.mixin.Scalar
 
     properties (Dependent, SetAccess=private, GetAccess=public)
+        NumRows
         NumColumns
         ColumnNames
         Schema
@@ -39,6 +40,10 @@
             obj.Proxy = proxy;
         end
 
+        function numRows = get.NumRows(obj)
+            numRows = obj.Proxy.getNumRows();
+        end
+
         function numColumns = get.NumColumns(obj)
             numColumns = obj.Proxy.getNumColumns();
         end

diff --git a/matlab/test/arrow/tabular/tRecordBatch.m b/matlab/test/arrow/tabular/tRecordBatch.m
@@ -62,6 +62,19 @@ function ColumnNames(tc)
             tc.verifyEqual(arrowRecordBatch.ColumnNames, columnNames);
         end
 
+        function NumRows(testCase)
+            % Verify that the NumRows property of arrow.tabular.RecordBatch
+            % returns the expected number of rows.
+            numRows = int64([1, 5, 100]);
+
+            for expectedNumRows = numRows
+                matlabTable = array2table(ones(expectedNumRows, 1));
+                arrowRecordBatch = arrow.recordBatch(matlabTable);
+                testCase.verifyEqual(arrowRecordBatch.NumRows, expectedNumRows);
+            end
+
+        end
+
         function NumColumns(tc)
             numColumns = int32([1, 5, 100]);
 
@@ -84,11 +97,27 @@ function UnicodeColumnNames(tc)
             tc.verifyEqual(TOriginal, TConverted);
         end
 
-        function EmptyTable(tc)
-            TOriginal = table();
-            arrowRecordBatch = arrow.recordBatch(TOriginal);
-            TConverted = arrowRecordBatch.toMATLAB();
-            tc.verifyEqual(TOriginal, TConverted);
+        function EmptyRecordBatch(testCase)
+            % Verify that an arrow.tabular.RecordBatch can be created from
+            % an empty MATLAB table.
+            matlabTable = table.empty(0, 0);
+            arrowRecordBatch = arrow.recordBatch(matlabTable);
+            testCase.verifyEqual(arrowRecordBatch.NumRows, int64(0));
+            testCase.verifyEqual(arrowRecordBatch.NumColumns, int32(0));
+            testCase.verifyEqual(arrowRecordBatch.ColumnNames, string.empty(1, 0));
+            testCase.verifyEqual(toMATLAB(arrowRecordBatch), matlabTable);
+
+            matlabTable = table.empty(1, 0);
+            arrowRecordBatch = arrow.recordBatch(matlabTable);
+            testCase.verifyEqual(arrowRecordBatch.NumRows, int64(0));
+            testCase.verifyEqual(arrowRecordBatch.NumColumns, int32(0));
+            testCase.verifyEqual(arrowRecordBatch.ColumnNames, string.empty(1, 0));
+
+            matlabTable = table.empty(0, 1);
+            arrowRecordBatch = arrow.recordBatch(matlabTable);
+            testCase.verifyEqual(arrowRecordBatch.NumRows, int64(0));
+            testCase.verifyEqual(arrowRecordBatch.NumColumns, int32(1));
+            testCase.verifyEqual(arrowRecordBatch.ColumnNames, "Var1");
         end
 
         function EmptyRecordBatchColumnIndexError(tc)
@@ -196,6 +225,15 @@ function FromArraysColumnNamesHasMissingString(tc)
             tc.verifyError(fcn, "MATLAB:validators:mustBeNonmissing");
         end
 
+        function FromArraysNoInputs(testCase)
+            % Verify that an empty RecordBatch is returned when calling
+            % fromArrays with no input arguments.
+            arrowRecordBatch = arrow.tabular.RecordBatch.fromArrays();
+            testCase.verifyEqual(arrowRecordBatch.NumRows, int64(0));
+            testCase.verifyEqual(arrowRecordBatch.NumColumns, int32(0));
+            testCase.verifyEqual(arrowRecordBatch.ColumnNames, string.empty(1, 0));
+        end
+
         function Schema(tc)
         % Verify that the public Schema property returns an approprate
         % instance of arrow.tabular.Schema.