Tasks for loading GA data into Snowflake (PART 1)

This is part 1 of the GA loading pipeline which DOES NOT depend on a
Luigi upgrade.

DE-1374 (PART 1)

edx/analytics/tasks/common/bigquery_load.py

@@ -53,9 +53,9 @@ def __init__(self, credentials_target, dataset_id, table, update_id):
         self.update_id = update_id
         with credentials_target.open('r') as credentials_file:
             json_creds = json.load(credentials_file)
-            self.project_id = json_creds['project_id']
-            credentials = service_account.Credentials.from_service_account_info(json_creds)
-            self.client = bigquery.Client(credentials=credentials, project=self.project_id)
+        self.project_id = json_creds['project_id']
+        credentials = service_account.Credentials.from_service_account_info(json_creds)
+        self.client = bigquery.Client(credentials=credentials, project=self.project_id)
 
     def touch(self):
         self.create_marker_table()

edx/analytics/tasks/common/snowflake_load.py

@@ -11,6 +11,7 @@
 from snowflake.connector import ProgrammingError
 
 from edx.analytics.tasks.util.overwrite import OverwriteOutputMixin
+from edx.analytics.tasks.util.s3_util import canonicalize_s3_url
 from edx.analytics.tasks.util.url import ExternalURL
 
 log = logging.getLogger(__name__)
@@ -90,9 +91,22 @@ def touch(self, connection):
         self.create_marker_table()
 
         connection.cursor().execute(
-            """INSERT INTO {database}.{schema}.{marker_table} (update_id, target_table)
-               VALUES (%s, %s)""".format(database=self.database, schema=self.schema, marker_table=self.marker_table),
-            (self.update_id, "{database}.{schema}.{table}".format(database=self.database, schema=self.schema, table=self.table))
+            """
+            INSERT INTO {database}.{schema}.{marker_table} (update_id, target_table)
+            VALUES (%s, %s)
+            """.format(
+                database=self.database,
+                schema=self.schema,
+                marker_table=self.marker_table,
+            ),
+            (
+                self.update_id,
+                "{database}.{schema}.{table}".format(
+                    database=self.database,
+                    schema=self.schema,
+                    table=self.table,
+                ),
+            )
         )
 
         # make sure update is properly marked
@@ -112,8 +126,17 @@ def exists(self, connection=None):
                 return False
 
             cursor = connection.cursor()
-            query = "SELECT 1 FROM {database}.{schema}.{marker_table} WHERE update_id='{update_id}' AND target_table='{database}.{schema}.{table}'".format(
-                database=self.database, schema=self.schema, marker_table=self.marker_table, update_id=self.update_id, table=self.table)
+            query = """
+            SELECT 1
+            FROM {database}.{schema}.{marker_table}
+            WHERE update_id='{update_id}' AND target_table='{database}.{schema}.{table}'
+            """.format(
+                database=self.database,
+                schema=self.schema,
+                marker_table=self.marker_table,
+                update_id=self.update_id,
+                table=self.table,
+            )
             log.debug(query)
             cursor.execute(query)
             row = cursor.fetchone()
@@ -135,8 +158,8 @@ def marker_table_exists(self, connection):
                 schema=self.schema,
             ))
             row = cursor.fetchone()
-        except ProgrammingError as e:
-            if "does not exist" in e.msg:
+        except ProgrammingError as err:
+            if "does not exist" in err.msg:
                 # If so then the query failed because the database or schema doesn't exist.
                 row = None
             else:
@@ -169,8 +192,14 @@ def clear_marker_table(self, connection):
         Delete all markers related to this table update.
         """
         if self.marker_table_exists(connection):
-            query = "DELETE FROM {database}.{schema}.{marker_table} where target_table='{database}.{schema}.{table}'".format(
-                database=self.database, schema=self.schema, marker_table=self.marker_table, table=self.table,
+            query = """
+            DELETE FROM {database}.{schema}.{marker_table}
+            WHERE target_table='{database}.{schema}.{table}'
+            """.format(
+                database=self.database,
+                schema=self.schema,
+                marker_table=self.marker_table,
+                table=self.table,
             )
             connection.cursor().execute(query)
 
@@ -233,20 +262,6 @@ def columns(self):
     def file_format_name(self):
         raise NotImplementedError
 
-    @property
-    def field_delimiter(self):
-        """
-        The delimiter in the data to be copied. Default is tab (\t).
-        """
-        return "\t"
-
-    @property
-    def null_marker(self):
-        """
-        The null sequence in the data to be copied. Default is Hive NULL (\\N).
-        """
-        return r'\\N'
-
     @property
     def pattern(self):
         """
@@ -260,7 +275,11 @@ def create_database(self, connection):
 
     def create_schema(self, connection):
         cursor = connection.cursor()
-        cursor.execute("CREATE SCHEMA IF NOT EXISTS {database}.{schema}".format(database=self.database, schema=self.schema))
+        cursor.execute(
+            "CREATE SCHEMA IF NOT EXISTS {database}.{schema}".format(
+                database=self.database, schema=self.schema,
+            )
+        )
 
     def create_table(self, connection):
         coldefs = ','.join(
@@ -273,29 +292,15 @@ def create_table(self, connection):
 
     def create_format(self, connection):
         """
-        Creates a named file format used for bulk loading data into Snowflake tables.
+        Invoke Snowflake's CREATE FILE FORMAT statement to create the named file format which
+        configures the loading.
+
+        The resulting file format name should be: {self.database}.{self.schema}.{self.file_format_name}
         """
-        query = """
-        CREATE OR REPLACE FILE FORMAT {database}.{schema}.{file_format_name}
-        TYPE = 'CSV' COMPRESSION = 'AUTO' FIELD_DELIMITER = '{field_delimiter}'
-        FIELD_OPTIONALLY_ENCLOSED_BY = 'NONE' ERROR_ON_COLUMN_COUNT_MISMATCH = TRUE
-        EMPTY_FIELD_AS_NULL = FALSE ESCAPE_UNENCLOSED_FIELD = 'NONE'
-        NULL_IF = ('{null_marker}')
-        """.format(
-            database=self.database,
-            schema=self.schema,
-            file_format_name=self.file_format_name,
-            field_delimiter=self.field_delimiter,
-            null_marker=self.null_marker,
-        )
-        log.debug(query)
-        connection.cursor().execute(query)
+        raise NotImplementedError
 
     def create_stage(self, connection):
-        """
-        Creates a named external stage to use for loading data into Snowflake.
-        """
-        stage_url = self.input()['insert_source_task'].path
+        stage_url = canonicalize_s3_url(self.input()['insert_source_task'].path)
         query = """
         CREATE OR REPLACE STAGE {database}.{schema}.{table}_stage
             URL = '{stage_url}'
@@ -377,3 +382,79 @@ def output(self):
 
     def update_id(self):
         return '{task_name}(date={key})'.format(task_name=self.task_family, key=self.date.isoformat())
+
+
+class SnowflakeLoadCSVTask(SnowflakeLoadTask):  # pylint: disable=abstract-method
+    """
+    Abstract Task for loading CSV data from s3 into a table in Snowflake.
+
+    Implementations should define the following properties:
+
+    - self.insert_source_task
+    - self.table
+    - self.columns
+    - self.file_format_name
+    """
+
+    @property
+    def field_delimiter(self):
+        """
+        The delimiter in the data to be copied. Default is tab (\t).
+        """
+        return "\t"
+
+    @property
+    def null_marker(self):
+        """
+        The null sequence in the data to be copied. Default is Hive NULL (\\N).
+        """
+        return r'\\N'
+
+    def create_format(self, connection):
+        query = """
+        CREATE OR REPLACE FILE FORMAT {database}.{schema}.{file_format_name}
+        TYPE = 'CSV' COMPRESSION = 'AUTO' FIELD_DELIMITER = '{field_delimiter}'
+        FIELD_OPTIONALLY_ENCLOSED_BY = 'NONE' ERROR_ON_COLUMN_COUNT_MISMATCH = TRUE
+        EMPTY_FIELD_AS_NULL = FALSE ESCAPE_UNENCLOSED_FIELD = 'NONE'
+        NULL_IF = ('{null_marker}')
+        """.format(
+            database=self.database,
+            schema=self.schema,
+            file_format_name=self.file_format_name,
+            field_delimiter=self.field_delimiter,
+            null_marker=self.null_marker,
+        )
+        log.debug(query)
+        connection.cursor().execute(query)
+
+
+class SnowflakeLoadJSONTask(SnowflakeLoadTask):  # pylint: disable=abstract-method
+    """
+    Abstract Task for loading JSON data from s3 into a table in Snowflake.  The resulting table will
+    contain a single VARIANT column called raw_json.
+
+    Implementations should define the following properties:
+
+    - self.insert_source_task
+    - self.table
+    - self.file_format_name
+    """
+
+    @property
+    def columns(self):
+        return [
+            ('raw_json', 'VARIANT'),
+        ]
+
+    def create_format(self, connection):
+        query = """
+        CREATE OR REPLACE FILE FORMAT {database}.{schema}.{file_format_name}
+        TYPE = 'JSON'
+        COMPRESSION = 'AUTO'
+        """.format(
+            database=self.database,
+            schema=self.schema,
+            file_format_name=self.file_format_name,
+        )
+        log.debug(query)
+        connection.cursor().execute(query)

edx/analytics/tasks/util/s3_util.py

@@ -5,7 +5,7 @@
 import os
 import time
 from fnmatch import fnmatch
-from urlparse import urlparse
+from urlparse import urlparse, urlunparse
 
 from luigi.contrib.hdfs.format import Plain
 from luigi.contrib.hdfs.target import HdfsTarget
@@ -122,22 +122,19 @@ def func(name):
     return (n for n in names if func(n))
 
 
+# TODO: Once we upgrade to boto3 (luigi>=2.7.6), delete this class! In boto3/luigi>=2.7.6, we must
+# NOT pass `host` to the s3 client or else it will throw a KeyError.  boto3 will already default to
+# s3.amazonaws.com.
 class ScalableS3Client(S3Client):
     """
-    S3 client that adds support for defaulting host name.
+    S3 client that adds support for defaulting host name to s3.amazonaws.com.
     """
-    # TODO: Make this behavior configurable and submit this change upstream.
 
-    def __init__(self, aws_access_key_id=None, aws_secret_access_key=None, **kwargs):
-
-        if not aws_access_key_id:
-            aws_access_key_id = self._get_s3_config('aws_access_key_id')
-        if not aws_secret_access_key:
-            aws_secret_access_key = self._get_s3_config('aws_secret_access_key')
+    def __init__(self, **kwargs):
         if 'host' not in kwargs:
             kwargs['host'] = self._get_s3_config('host') or 's3.amazonaws.com'
 
-        super(ScalableS3Client, self).__init__(aws_access_key_id=aws_access_key_id, aws_secret_access_key=aws_secret_access_key, **kwargs)
+        super(ScalableS3Client, self).__init__(**kwargs)
 
 
 class S3HdfsTarget(HdfsTarget):
@@ -162,3 +159,29 @@ def open(self, mode='r'):
             if not hasattr(self, 's3_client'):
                 self.s3_client = ScalableS3Client()
             return AtomicS3File(safe_path, self.s3_client, policy=DEFAULT_KEY_ACCESS_POLICY)
+
+
+def canonicalize_s3_url(url):
+    """
+    Convert the given s3 URL into a form which is safe to use with external tools.
+
+    Specifically, URL Schemes such as "s3+https" are unrecognized by gsutil and Snowflake, and must
+    be converted to "s3".
+
+    Args:
+        url (str): An s3 URL.
+
+    Raises:
+        ValueError: if the scheme of the input url is unrecognized as S3 at all.
+    """
+    parsed_url = urlparse(url)
+    if parsed_url.scheme == 's3':
+        canonical_url = url  # Simple passthrough, no change needed.
+    if parsed_url.scheme == 's3+https':
+        new_url_parts = parsed_url._replace(scheme='s3')
+        canonical_url = urlunparse(new_url_parts)
+    else:
+        raise ValueError(
+            'The URL scheme "{}" does not appear to be an S3 URL scheme.'.format(parsed_url.scheme)
+        )
+    return canonical_url

edx/analytics/tasks/util/tests/test_s3_util.py

@@ -1,4 +1,7 @@
-"""Tests for S3-related utility functionality."""
+"""
+Tests for S3-related utility functionality.
+"""
+from __future__ import print_function
 
 from unittest import TestCase
 
@@ -25,7 +28,7 @@ def _make_s3_generator(self, bucket_name, root, path_info, patterns):
         target_list = [self._make_key("{root}/{path}".format(root=root, path=path), size)
                        for path, size in path_info.iteritems()]
         s3_bucket.list = MagicMock(return_value=target_list)
-        print [(k.key, k.size) for k in target_list]
+        print([(k.key, k.size) for k in target_list])
 
         s3_bucket.name = bucket_name
         source = "s3://{bucket}/{root}".format(bucket=bucket_name, root=root)