@@ -17,9 +17,8 @@ RUN mkdir /result
 RUN pyperformance run --debug-single-value --python=python2.7 -o /result/py2.7.json
 RUN pyperformance run --debug-single-value --python=python3.4 -o /result/py3.4.json
 RUN pyperformance run --debug-single-value --python=python3.5 -o /result/py3.5.json
-RUN pyperformance run --debug-single-value --python=python3.6 -o /result/py3.6.json
+RUN if [ -e "/opt/python3.6/bin/python3.6" ]; then pyperformance run --debug-single-value --python=python3.6 -o /result/py3.6.json; fi
 
 RUN pyperformance compare /result/py2.7.json /result/py3.4.json --output_style table
 RUN pyperformance compare /result/py3.4.json /result/py3.5.json --output_style table
-RUN pyperformance compare /result/py3.5.json /result/py3.6.json --output_style table
-
+RUN if [ -e "/result/py3.6.json" ]; then pyperformance compare /result/py3.5.json /result/py3.6.json --output_style table; fi
@@ -17,8 +17,8 @@ rm Dockerfile
 echo "Successfully built images"
 
 # Create folders to hold the files
-mkdir $TAG1
-mkdir $TAG2
+mkdir "$TAG1"
+mkdir "$TAG2"
 
 # Start running the containers and copy the benchmark result for python versions from container to host
 docker run -it --name benchmark_1 -h CONTAINER1 -v "${PWD}"/"$TAG1":/export benchmark_1 /bin/bash -c "cp /result/py*.json /export/"
@@ -30,7 +30,11 @@ echo "Start benchmarking the python interpreter performance between the two rele
 pyperformance compare "$TAG1"/py2.7.json "$TAG2"/py2.7.json --output_style table > py2.7_res
 pyperformance compare "$TAG1"/py3.4.json "$TAG2"/py3.4.json --output_style table > py3.4_res
 pyperformance compare "$TAG1"/py3.5.json "$TAG2"/py3.5.json --output_style table > py3.5_res
-pyperformance compare "$TAG1"/py3.6.json "$TAG2"/py3.6.json --output_style table > py3.6_res
+
+# Check if the python3.6 benchmark result exists
+if [[ ( -e '"$TAG1"/py3.6.json' ) && ( -e '"$TAG2"/py3.6.json' ) ]]; then
+    pyperformance compare "$TAG1"/py3.6.json "$TAG2"/py3.6.json --output_style table > py3.6_res;
+fi
 
 echo "Start extracting data and generating CSV file, then upload to Cloud Storage and insert to Big Query table"
 
@@ -43,22 +47,25 @@ for path_to_file in $TAG2/*.json; do
     python generate_csv.py --filename $path_to_file --tag $TAG2
 done
 
-# Get the list of existed release data on Cloud Storage and skip if the current TAG1 or TAG2 existed in the list
-gsutil ls gs://python_runtime_benchmark > existed_releases
+# Set the project that hold the cloud storage bucket and big query tables
+gcloud config set project cloud-python-runtime-qa
+
+# Get the list of existing release data on Cloud Storage and skip if the current TAG1 or TAG2 existing in the list
+gsutil ls gs://python-runtime-benchmark > existing_releases
 
 for container_tag in $TAG1 $TAG2; do
-    if grep --fixed-strings --quiet "$container_tag" existed_releases; then
-        echo "Performace data of $container_tag existed, so skip processing it."
+    if grep --fixed-strings --quiet "$container_tag" existing_releases; then
+        echo "Performance data of $container_tag existed, so skip processing it."
     else
         # Upload the CSV files to Cloud Storage
-        gsutil cp -r $container_tag gs://python_runtime_benchmark
+        gsutil cp -r $container_tag gs://python-runtime-benchmark
         # Load the CSV files from Cloud Storage to Big Query table
         # Load the performance data of each function
         for path_to_file in $container_tag/py2.7.csv $container_tag/py3.4.csv $container_tag/py3.5.csv; do
-             bq load benchmark_test.benchmark_functions gs://python_runtime_benchmark/"$path_to_file" container_tag:string,runtime_version:string,function_name:string,time_used:float,mem_usage:float
+             bq load benchmark.benchmark_functions gs://python-runtime-benchmark/"$path_to_file" container_tag:string,runtime_version:string,function_name:string,time_used:float,mem_usage:float
         done
         # Load the average performance data of each runtime version in a release
-        bq load benchmark_test.benchmark_statistics gs://python_runtime_benchmark/"$container_tag"/averages.csv container_tag:string,runtime_version:string,ave_time_used:float,ave_mem_usage:float
+        bq load benchmark.benchmark_statistics gs://python-runtime-benchmark/"$container_tag"/averages.csv container_tag:string,runtime_version:string,ave_time_used:float,ave_mem_usage:float
     fi
 done
 
 
@@ -40,16 +40,15 @@ def generate_csv(filename, tag):
                 # Get the function name
                 func_name = benchmark["metadata"]["name"]
                 # Get the time used for this function, convert to millisecond
-                time_used = benchmark["runs"][0]["values"][0] * 1000
+                time_used = float(benchmark["runs"][0]["values"][0]) * 1000
                 # Get the memory usage, convert to MB
-                mem_usage = benchmark["metadata"]["mem_max_rss"] / float(1<<20)
+                mem_usage = float(benchmark["metadata"]["mem_max_rss"]) / float(1<<20)
                 line = [tag, runtime_version, func_name, time_used, mem_usage]
                 # Write to CSV file
                 csv_writer.writerow(line)
             except KeyError:
                 # Skip the benchmark result if it does not contain the fields we want
                 pass
-    output.close()
 
 
 def get_averages(filename, tag):
@@ -67,7 +66,6 @@ def get_averages(filename, tag):
         col_sums = map(sum, zip(*rows_of_data))
         # Calculate the average of the two columns by using the sum divided by the total number of lines
         averages = [col_sum / len(lines) for col_sum in col_sums]
-    input.close()
 
     # Get the runtime version from filename
     runtime_version = os.path.basename(filename).split(".json")[0]
@@ -79,7 +77,6 @@ def get_averages(filename, tag):
             csv_writer.writerow([tag, runtime_version] + averages)
         except IOError:
             print "Could not write averages to file."
-    output.close()
 
 
 def parse_args(argv):