ParquetReader.RowGroups property

src/Parquet.Test/ParquetReaderTest.cs

@@ -270,6 +270,18 @@ public async Task Metadata_file() {
             Assert.True(reader.CustomMetadata.ContainsKey("geo"));
             Assert.True(reader.CustomMetadata.ContainsKey("ARROW:schema"));
         }
+
+        [Theory]
+        [InlineData("multi.page.parquet")]
+        [InlineData("multi.page.v2.parquet")]
+        public async Task RowGroups_EnumeratesRowGroups(string parquetFile) {
+            using(ParquetReader reader = await ParquetReader.CreateAsync(OpenTestFile(parquetFile), leaveStreamOpen: false)) {
+
+                Assert.Single(reader.RowGroups);
+                IParquetRowGroupReader rowGroup = reader.RowGroups.Single();
+                Assert.Equal(927861, rowGroup.RowCount);
+            }
+        }
 
         class ReadableNonSeekableStream : DelegatedStream {
             public ReadableNonSeekableStream(Stream master) : base(master) {

src/Parquet/ParquetReader.cs

@@ -134,6 +134,11 @@ public ParquetRowGroupReader OpenRowGroupReader(int index) {
             return _groupReaders[index];
         }
 
+        /// <summary>
+        /// Collection of row group readers, fast random access and enumeration
+        /// </summary>
+        public IReadOnlyList<IParquetRowGroupReader> RowGroups => _groupReaders;
+
         /// <summary>
         /// Reads entire row group's data columns in one go.
         /// </summary>

src/Parquet/ParquetRowGroupReader.cs

@@ -11,10 +11,55 @@
 using Parquet.Schema;
 
 namespace Parquet {
+    /// <summary>
+    /// Operations available on a row group reader, omitting Dispose, which is 
+    /// exposed on the implementing class for backward compatibility only.
+    /// </summary>
+    public interface IParquetRowGroupReader {
+        /// <summary>
+        /// Exposes raw metadata about this row group
+        /// </summary>
+        RowGroup RowGroup { get; }
+
+        /// <summary>
+        /// Gets the number of rows in this row group
+        /// </summary>
+        long RowCount { get; }
+
+        /// <summary>
+        /// Checks if this field exists in source schema
+        /// </summary>
+        bool ColumnExists(DataField field);
+
+        /// <summary>
+        /// Reads a column from this row group. Unlike writing, columns can be read in any order.
+        /// If the column is missing, an exception will be thrown.
+        /// </summary>
+        Task<DataColumn> ReadColumnAsync(DataField field, CancellationToken cancellationToken = default);
+
+        /// <summary>
+        /// Gets raw column chunk metadata for this field
+        /// </summary>
+        ColumnChunk? GetMetadata(DataField field);
+
+        /// <summary>
+        /// Get custom key-value metadata for a data field
+        /// </summary>
+        Dictionary<string, string> GetCustomMetadata(DataField field);
+
+        /// <summary>
+        /// Returns data column statistics for a particular data field
+        /// </summary>
+        /// <param name="field"></param>
+        /// <returns></returns>
+        /// <exception cref="ParquetException"></exception>
+        DataColumnStatistics? GetStatistics(DataField field);
+    }
+
     /// <summary>
     /// Reader for Parquet row groups
     /// </summary>
-    public class ParquetRowGroupReader : IDisposable {
+    public class ParquetRowGroupReader : IDisposable, IParquetRowGroupReader {
         private readonly RowGroup _rowGroup;
         private readonly ThriftFooter _footer;
         private readonly Stream _stream;
@@ -118,12 +163,8 @@ public Dictionary<string, string> GetCustomMetadata(DataField field) {
         }
 
         /// <summary>
-        /// 
+        /// Dispose isn't required, retained for backward compatibility
         /// </summary>
-        public void Dispose() {
-            //don't need to dispose anything here, but for clarity we implement IDisposable and client must use it as we may add something
-            //important in it later
-        }
-
+        public void Dispose() { }
     }
 }