adding parquet test. the test creates several columns table using pyarrow. each column contains random numbers. the test verified s3select query results

Signed-off-by: gal salomon <gal.salomon@gmail.com>

s3tests_boto3/functional/test_s3select.py

@@ -218,6 +218,70 @@ def upload_csv_object(bucket_name,new_key,obj):
         response = c2.get_object(Bucket=bucket_name, Key=new_key)
         eq(response['Body'].read().decode('utf-8'), obj, 's3select error[ downloaded object not equal to uploaded objecy')
 
+def parquet_generator(bucket_name,parquet_obj_name,parquet_size):
+
+    #purpose: creating parquet object with several columns, each column contains random numbers
+    #the s3select query results(counts of where clause) is verified against arrays comparision results.
+
+    a=[]
+    for i in range(parquet_size):
+        a.append(int(random.randint(1,10000)))
+
+    b=[]
+    for i in range(parquet_size):
+        b.append(int(random.randint(1,10000)))
+
+    c=[]
+    for i in range(parquet_size):
+        c.append(int(random.randint(1,10000)))
+
+    d=[]
+    for i in range(parquet_size):
+        d.append(int(random.randint(1,10000)))
+
+    # count where a = b, validation against the s3select query result 
+    count_a_b_eq=0
+    for i in range(parquet_size):
+        if(a[i] == b[i]):
+            count_a_b_eq=count_a_b_eq+1
+
+    # count where c = d, validation against the s3select query result
+    count_c_d_eq=0
+    for i in range(parquet_size):
+        if(c[i] == d[i]):
+            count_c_d_eq=count_c_d_eq+1
+
+
+    #create dataframe. the actual columns to reside on parquet object
+    df3 = pd.DataFrame({'a': a,
+                   'b': b,
+                   'c': c,
+                   'd': d}
+                   )
+
+    # creating parquet object
+    table = pa.Table.from_pandas(df3,preserve_index=False)
+    obj = pa.BufferOutputStream()
+    pq.write_table(table, obj)##, compression="snappy")
+
+    ## load file-content into s3-storage
+    client = get_client()
+    client.create_bucket(Bucket=bucket_name)
+    client.put_object(Bucket=bucket_name, Key=parquet_obj_name, Body=obj.getvalue().to_pybytes())
+
+    # following queries verify query result against the count on arrays that reside on parquet object
+    res_s3select = remove_xml_tags_from_result(  run_s3select(bucket_name,parquet_obj_name,"select count(0) from s3object;")  ).replace(",","")
+
+    s3select_assert_result( parquet_size , int( res_s3select ))
+
+    res_s3select = remove_xml_tags_from_result(  run_s3select(bucket_name,parquet_obj_name,"select count(0) from s3object where a = b;")  ).replace(",","")
+
+    s3select_assert_result( count_a_b_eq  , int( res_s3select ))
+
+    res_s3select = remove_xml_tags_from_result(  run_s3select(bucket_name,parquet_obj_name,"select count(0) from s3object where c = d;")  ).replace(",","")
+
+    s3select_assert_result( count_c_d_eq  , int( res_s3select ))
+
 def run_s3select(bucket,key,query,column_delim=",",row_delim="\n",quot_char='"',esc_char='\\',csv_header_info="NONE", progress = False):
 
     s3 = get_client()
@@ -1274,3 +1338,11 @@ def test_output_serial_expressions():
     
     s3select_assert_result( res_s3select_quot, res_s3select_final )
 
+@attr('s3select_parquet')
+def test_parquet():
+
+    parquet_obj_name = "4col.parquet"
+    size = 1000000
+
+    parquet_generator("test",parquet_obj_name,size)
+