Merge 833518f into 3d9135c

CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## 0.0.11
+- `dtype` parameter removed from the DataFrame's constructor
+- `dtype` parameter added to `toMatrix` method
+
 ## 0.0.10
 - `DataFrame.sampleFromSeries` method's signature changed 
 

lib/ml_dataframe.dart

@@ -1,3 +1,3 @@
 export 'package:ml_dataframe/src/data_frame/data_frame.dart';
 export 'package:ml_dataframe/src/data_frame/factories/from_csv.dart';
-export 'package:ml_dataframe/src/data_frame/series.dart';
+export 'package:ml_dataframe/src/data_frame/series.dart';

lib/src/data_frame/data_frame.dart

@@ -43,7 +43,6 @@ abstract class DataFrame {
         String autoHeaderPrefix = defaultHeaderPrefix,
         Iterable<int> columns = const [],
         Iterable<String> columnNames = const [],
-        DType dtype = DType.float32,
       }
   ) => fromRawData(
     data,
@@ -52,19 +51,13 @@ abstract class DataFrame {
     autoHeaderPrefix: autoHeaderPrefix,
     columns: columns,
     columnNames: columnNames,
-    dtype: dtype,
   );
 
-  factory DataFrame.fromSeries(
-      Iterable<Series> series,
-      {
-        DType dtype = DType.float32,
-      }
-  ) => DataFrameImpl.fromSeries(
-      series,
-      NumericalConverterImpl(false),
-      dtype,
-  );
+  factory DataFrame.fromSeries(Iterable<Series> series) =>
+      DataFrameImpl.fromSeries(
+        series,
+        NumericalConverterImpl(false),
+      );
 
   factory DataFrame.fromMatrix(
       Matrix matrix,
@@ -84,9 +77,6 @@ abstract class DataFrame {
     discreteColumnNames: discreteColumnNames,
   );
 
-  /// A type used for converting the [DataFrame] into [Matrix]
-  DType get dtype;
-
   /// Returns a collection of names of all series (like a table header)
   Iterable<String> get header;
 
@@ -124,5 +114,5 @@ abstract class DataFrame {
   ///
   /// The method may throw an error if the [DataFrame] contains data, that
   /// cannot be converted to numerical representation
-  Matrix toMatrix();
+  Matrix toMatrix([DType dtype]);
 }

lib/src/data_frame/data_frame_impl.dart

@@ -8,28 +8,29 @@ import 'package:ml_linalg/matrix.dart';
 import 'package:quiver/iterables.dart';
 
 class DataFrameImpl implements DataFrame {
-  DataFrameImpl(this.rows, this.header, this._toNumber, this.dtype) :
+  DataFrameImpl(this.rows, this.header, this._toNumber) :
         series = convertRowsToSeries(header, rows);
 
-  DataFrameImpl.fromSeries(this.series, this._toNumber, this.dtype) :
+  DataFrameImpl.fromSeries(this.series, this._toNumber) :
         header = series.map((series) => series.name),
         rows = convertSeriesToRows(series);
 
   DataFrameImpl.fromMatrix(
-      this._cachedMatrix,
+      Matrix matrix,
       this.header,
       this._toNumber,
       Iterable<bool> areSeriesDiscrete,
   ) :
-        dtype = _cachedMatrix.dtype,
-        rows = _cachedMatrix.rows,
-        series = zip([header, _cachedMatrix.columns,
-          areSeriesDiscrete ?? List.filled(_cachedMatrix.columnsNum, false)])
+        rows = matrix.rows,
+        series = zip([header, matrix.columns,
+          areSeriesDiscrete ?? List.filled(matrix.columnsNum, false)])
             .map((seriesData) => Series(
               seriesData[0] as String,
               seriesData[1] as Iterable,
               isDiscrete: seriesData[2] as bool,
-            ));
+            )) {
+    _cachedMatrices[matrix.dtype] = matrix;
+  }
 
   @override
   final Iterable<String> header;
@@ -40,12 +41,9 @@ class DataFrameImpl implements DataFrame {
   @override
   final Iterable<Series> series;
 
-  @override
-  final DType dtype;
-
   final NumericalConverter _toNumber;
 
-  Matrix _cachedMatrix;
+  final Map<DType, Matrix> _cachedMatrices = {};
 
   @override
   Series operator [](Object key) {
@@ -74,7 +72,7 @@ class DataFrameImpl implements DataFrame {
         .from(names)
         .difference(Set.from(header));
     if (absentNames.isNotEmpty) {
-      throw Exception('Columns with the names $absentNames do not exist');
+      throw Exception('Columns with names $absentNames do not exist');
     };
     return _sampleFromSeries(names);
   }
@@ -92,34 +90,31 @@ class DataFrameImpl implements DataFrame {
   }
 
   @override
-  Matrix toMatrix() =>
-      _cachedMatrix ??= Matrix.fromList(
-        _toNumber
-            .convertRawData(rows)
-            .map((row) => row.toList())
-            .toList(),
-        dtype: dtype,
-      );
+  Matrix toMatrix([DType dtype = DType.float32]) =>
+    _cachedMatrices[dtype] ??= Matrix.fromList(
+      _toNumber
+          .convertRawData(rows)
+          .map((row) => row.toList())
+          .toList(),
+      dtype: dtype,
+    );
 
   DataFrame _sampleFromSeries(Iterable ids) =>
-      DataFrame.fromSeries(
-        ids.map((dynamic id) => this[id]),
-        dtype: dtype,
-      );
+      DataFrame.fromSeries(ids.map((dynamic id) => this[id]));
 
   DataFrame _dropByIndices(Iterable<int> indices, Iterable<Series> series) {
     final uniqueIndices = Set<int>.from(indices);
     final newSeries = enumerate(series)
         .where((indexedSeries) => !uniqueIndices.contains(indexedSeries.index))
         .map((indexedSeries) => indexedSeries.value);
-    return DataFrame.fromSeries(newSeries, dtype: dtype);
+    return DataFrame.fromSeries(newSeries);
   }
 
   DataFrame _dropByNames(Iterable<String> names, Iterable<Series> series) {
     final uniqueNames = Set<String>.from(names);
     final newSeries = series
         .where((series) => !uniqueNames.contains(series.name));
-    return DataFrame.fromSeries(newSeries, dtype: dtype);
+    return DataFrame.fromSeries(newSeries);
   }
 
   Map<String, Series> _getCachedOrCreateSeriesByName() =>

lib/src/data_frame/factories/from_csv.dart

@@ -2,7 +2,6 @@ import 'package:csv/csv.dart';
 import 'package:ml_dataframe/src/data_frame/data_frame.dart';
 import 'package:ml_dataframe/src/data_frame/factories/from_raw_data.dart';
 import 'package:ml_dataframe/src/data_reader/data_reader.dart';
-import 'package:ml_linalg/dtype.dart';
 
 Future<DataFrame> fromCsv(
     String fileName,
@@ -14,7 +13,6 @@ Future<DataFrame> fromCsv(
       String autoHeaderPrefix = defaultHeaderPrefix,
       Iterable<int> columns = const [],
       Iterable<String> columnNames = const [],
-      DType dtype = DType.float32,
     }
 ) async {
   final reader = DataReader.csv(fileName, columnDelimiter, eol);
@@ -27,6 +25,5 @@ Future<DataFrame> fromCsv(
     autoHeaderPrefix: autoHeaderPrefix,
     columns: columns,
     columnNames: columnNames,
-    dtype: dtype,
   );
 }

lib/src/data_frame/factories/from_raw_data.dart

@@ -12,7 +12,6 @@ DataFrame fromRawData(Iterable<Iterable<dynamic>> data, {
   String autoHeaderPrefix = defaultHeaderPrefix,
   Iterable<int> columns = const [],
   Iterable<String> columnNames = const [],
-  DType dtype = DType.float32,
 }) {
   final columnsNum = columns.isNotEmpty
       ? columns.length
@@ -55,5 +54,5 @@ DataFrame fromRawData(Iterable<Iterable<dynamic>> data, {
       .map((indexedName) => indexedName.value);
 
   return DataFrameImpl(selectedData, selectedHeader,
-      NumericalConverterImpl(false), dtype);
+      NumericalConverterImpl(false));
 }

pubspec.yaml

@@ -1,6 +1,6 @@
 name: ml_dataframe
 description: Dataframe - a way to store and manipulate data
-version: 0.0.10
+version: 0.0.11
 author: Ilia Gyrdymov <ilgyrd@gmail.com>
 homepage: https://github.com/gyrdym/ml_dataframe
 

test/data_frame/data_frame_from_matrix_test.dart

@@ -42,7 +42,6 @@ void main() {
       expect(dataFrame['col_3'].data, equals([4, 40, 400, 4000]));
 
       expect(dataFrame.toMatrix(), same(matrix));
-      expect(dataFrame.dtype, DType.float32);
     });
 
     test('should initialize from matrix with predefined header', () {
@@ -83,7 +82,6 @@ void main() {
       expect(dataFrame['little'].data, equals([4, 40, 400, 4000]));
 
       expect(dataFrame.toMatrix(), same(matrix));
-      expect(dataFrame.dtype, DType.float32);
     });
 
     test('should ignore predefined header list elements that are out of '
@@ -121,7 +119,6 @@ void main() {
 
       expect(dataFrame[3].name, 'little');
       expect(dataFrame['little'].data, equals([4, 40, 400, 4000]));
-      expect(dataFrame.dtype, DType.float32);
     });
 
     test('should ignore predefined header list elements that are out of '
@@ -186,7 +183,6 @@ void main() {
       expect(dataFrame[1].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['doth'].data, equals([4, 40, 400, 4000]));
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), equals([
         [1,    4   ],
         [10,   40  ],
@@ -234,7 +230,6 @@ void main() {
       expect(dataFrame[3].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['super_3'].data, equals([4, 40, 400, 4000]));
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), same(matrix));
     });
 
@@ -266,7 +261,6 @@ void main() {
       expect(dataFrame[1].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['col_1'].data, equals([4, 40, 400, 4000]));
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), equals([
         [1,    4   ],
         [10,   40  ],
@@ -314,7 +308,6 @@ void main() {
       expect(dataFrame['col_3'].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['col_3'].isDiscrete, isFalse);
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), same(matrix));
     });
 
@@ -359,7 +352,6 @@ void main() {
       expect(dataFrame['fourth'].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['fourth'].isDiscrete, isTrue);
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), same(matrix));
     });
 
@@ -403,7 +395,6 @@ void main() {
       expect(dataFrame['col_3'].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['col_3'].isDiscrete, isTrue);
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), same(matrix));
     });
 
@@ -447,7 +438,6 @@ void main() {
       expect(dataFrame['col_3'].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['col_3'].isDiscrete, isFalse);
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), same(matrix));
     });
 
@@ -490,7 +480,6 @@ void main() {
       expect(dataFrame[3].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['col_3'].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['col_3'].isDiscrete, isFalse);
-      expect(dataFrame.dtype, DType.float32);
 
       expect(dataFrame.toMatrix(), same(matrix));
     });
@@ -536,7 +525,6 @@ void main() {
       expect(dataFrame['fourth'].data, equals([4, 40, 400, 4000]));
       expect(dataFrame['fourth'].isDiscrete, isTrue);
 
-      expect(dataFrame.dtype, DType.float32);
       expect(dataFrame.toMatrix(), same(matrix));
     });
   });

test/data_frame/data_frame_from_raw_data_test.dart

@@ -1,5 +1,4 @@
 import 'package:ml_dataframe/src/data_frame/data_frame.dart';
-import 'package:ml_linalg/dtype.dart';
 import 'package:test/test.dart';
 
 void main() {
@@ -14,7 +13,6 @@ void main() {
       final frame = DataFrame(
         data,
         headerExists: false,
-        dtype: DType.float64,
       );
 
       expect(frame.header,
@@ -33,7 +31,6 @@ void main() {
         [true, false, true],
         ['32', '1132', 'abs'],
       ]));
-      expect(frame.dtype, DType.float64);
     });
 
     test('should initialize from dynamic-typed data with header row', () {

test/data_frame/data_frame_from_series_test.dart

@@ -1,6 +1,5 @@
 import 'package:ml_dataframe/src/data_frame/data_frame.dart';
 import 'package:ml_dataframe/src/data_frame/series.dart';
-import 'package:ml_linalg/dtype.dart';
 import 'package:test/test.dart';
 
 void main() {
@@ -11,7 +10,7 @@ void main() {
         Series('second', <dynamic>[10, 12, 323, false, '1132']),
         Series('third', <dynamic>[-10, 202, null, true, 'abs']),
       ];
-      final frame = DataFrame.fromSeries(series, dtype: DType.float64);
+      final frame = DataFrame.fromSeries(series);
 
       expect(frame.header, equals(['first', 'second', 'third']));
       expect(frame.rows, equals([
@@ -28,7 +27,6 @@ void main() {
         <dynamic>[10, 12, 323, false, '1132'],
         <dynamic>[-10, 202, null, true, 'abs'],
       ]));
-      expect(frame.dtype, DType.float64);
     });
   });
 }