Patching 0.7.1 release (#639)

* making sure that bigquery reader is returning meaningfull defaults for null values (#626)

* making sure that bigquery reader is returning meaningfull defaults for null values

* linter fixes

* Use balanced sharding strategy in bigquery read session. (#601)

We use sharding to provide a way of shuffling the read order. With liquid sharding, a single stream can read all data which wouldn't be shuffled.

* bumping release version 0.7.1->0.7.2

* properly merging tensorflow_io/bigquery/kernels/bigquery_lib.h

Author: vlasenkoalexey

setup.py

@@ -95,7 +95,7 @@ def has_ext_modules(self):
 """
 
 package = 'tensorflow>=1.14.0,<1.15.0'
-version = '0.7.1'
+version = '0.7.2'
 project = 'tensorflow-io'
 if '--package-version' in sys.argv:
   print(package)

tensorflow_io/bigquery/kernels/bigquery_kernels.cc

@@ -149,6 +149,7 @@ class BigQueryReadSessionOp : public OpKernel {
     createReadSessionRequest.mutable_read_options()->set_row_restriction(
         row_restriction_);                                        
     createReadSessionRequest.set_requested_streams(requested_streams_);
+    createReadSessionRequest.set_sharding_strategy(apiv1beta1::ShardingStrategy::BALANCED);
     createReadSessionRequest.set_format(apiv1beta1::DataFormat::AVRO);
     VLOG(3) << "createReadSessionRequest: "
             << createReadSessionRequest.DebugString();

tensorflow_io/bigquery/kernels/bigquery_lib.h

@@ -253,7 +253,31 @@ class BigQueryReaderDatasetIterator : public DatasetIterator<Dataset> {
           ((*out_tensors)[i]).scalar<string>()() =
               field.value<avro::GenericEnum>().symbol();
           break;
-        case avro::AVRO_NULL:  // Fallthrough;
+        case avro::AVRO_NULL:
+          switch(output_types[i]) {
+            case DT_BOOL:
+              ((*out_tensors)[i]).scalar<bool>()() = false;
+              break;
+            case DT_INT32:
+              ((*out_tensors)[i]).scalar<int32>()() = 0;
+              break;
+            case DT_INT64:
+              ((*out_tensors)[i]).scalar<int64>()() = 0l;
+              break;
+            case DT_FLOAT:
+              ((*out_tensors)[i]).scalar<float>()() = 0.0f;
+              break;
+            case DT_DOUBLE:
+              ((*out_tensors)[i]).scalar<double>()() = 0.0;
+              break;
+            case DT_STRING:
+              ((*out_tensors)[i]).scalar<string>()() = "";
+              break;
+            default:
+              return errors::InvalidArgument(
+                "unsupported data type against AVRO_NULL: ",
+                output_types[i]);
+          }
           break;
         default:
           return errors::InvalidArgument("unsupported data type: ",