[DATA] Fix indices calculation in rebatching.

hybridbackend/cpp/tensorflow/ops/parquet_dataset_ops.cc

@@ -127,7 +127,7 @@ class ParquetTabularDatasetOp::Dataset : public DatasetBase {
   }
 
   Status Open() {
-    VLOG(0) << "Starting to read " << filename_ << " ...";
+    VLOG(1) << "Starting to read " << filename_ << " ...";
     return reader_->Open();
   }
 

hybridbackend/cpp/tensorflow/ops/rebatch_dataset_ops.cc

@@ -294,7 +294,7 @@ class RebatchTabularDatasetOp::Dataset::Iterator
       return 0;
     }
     if (field_ranks_[0] > 0) {
-      return input_tensors[1].dim_size(0);
+      return input_tensors[1].dim_size(0) - 1;
     }
     return input_tensors[0].dim_size(0);
   }

tests/tensorflow/data/parquet_dataset_rebatch_test.py

@@ -22,7 +22,10 @@
 
 import numpy as np
 import pandas as pd
+import pyarrow as pa
+import pyarrow.parquet as pq
 import os
+import random
 from six.moves import xrange  # pylint: disable=redefined-builtin
 import tempfile
 import tensorflow as tf
@@ -234,6 +237,39 @@ def test_thread_pool(self):
           actual.nested_row_splits, expected.nested_row_splits)
 
 
+class ParquetDatasetSequenceRebatchTest(unittest.TestCase):
+  def setUp(self):  # pylint: disable=invalid-name
+    self._workspace = tempfile.mkdtemp()
+    self._filename = os.path.join(self._workspace, 'seqtest.parquet')
+    self._nrows = 1000
+    self._ncols = 10
+    self._data = {
+      'clicks': [
+        [random.randint(0, 100) for col in range(self._ncols)]
+        for row in range(self._nrows)]}
+    pq.write_table(pa.Table.from_pydict(self._data), self._filename)
+
+  def tearDown(self):  # pylint: disable=invalid-name
+    os.remove(self._filename)
+
+  def test_ragged(self):
+    batch_size = 8
+    with tf.Graph().as_default() as graph:
+      ds = ParquetDataset(self._filename, batch_size=batch_size)
+      ds = ds.apply(rebatch(batch_size))
+      batch = make_one_shot_iterator(ds).get_next()
+
+    clicks = self._data['clicks']
+    with tf.Session(graph=graph) as sess:
+      for i in xrange(3):
+        actual = sess.run(batch['clicks'])
+        start_row = i * batch_size
+        end_row = (i + 1) * batch_size
+        expected = clicks[start_row:end_row]
+        expected_values = [v for sublist in expected for v in sublist]
+        np.testing.assert_equal(actual.values, expected_values)
+
+
 if __name__ == '__main__':
   register(['cpu', 'data'])
   os.environ['CUDA_VISIBLE_DEVICES'] = ''