snowmobile 0.1.11

A simple set of modules for streamlined interaction with the Snowflake Database

snowmobile

snowmobile is a simple set of modules for streamlined interaction with the Snowflake Database for Data Scientists and Business Analysts.

As such the included codes are intended to be used for the execution of raw SQL or with pandas DataFrames and don't make use of an ORM to map Python objects to tabular Snowflake counterparts.

A quick overview of simplified usage is outlined below.

---

Basic usage

1. Install with pip install snowmobile

2. Create file called snowflake_credentials.json following the below structure with as many sets of credentials as desired and store anywhere on local file system

   {
   "Connection1": {
       "username":	"",
       "password":	"",
       "role": "",
       "account": "",
       "warehouse": "warehouse #1",
       "database":	"database #1",
       "schema": "schema #1"
     },"SANDBOX": {
       "username":	"",
       "password":	"",
       "role": "",
       "account": "",
       "warehouse": "warehouse #1",
       "database":	"database #1",
       "schema": "SANDBOX"
     }
   }
   

3. Import module for the use-case and execute simplified commands

   # <bundled authentication & statement-execution>  
   from snowmobile import snowquery
   
   # Instantiate an instance of a connection
   sandbox_conn = snowquery.Connector(conn_name='SANDBOX')
   
   # Execute statements on that connection 
   sample_df = sandbox_conn.execute_query('SELECT * FROM SAMPLE_TABLE')
   
   # Manipulate DataFrame
   transposed_df = sample_df.transpose() 
   
   # Instantiate a different connection
   user_conn = snowquery.Connector(conn_name='user_schema')
   
   # <flexible loading solution>
   from snowmobile import snowloader
   
   # Load to different location
   snowloader.df_to_snowflake(df=transposed_df, table_name='LATEST_SAMPLE', 
                              connector=user_conn, force_recreate=True)
   
   # Locate a bunch of sql files to execute
   import os
   paths_to_sql = [os.path.join(file, path) for file in os.listdir(path)]
   
   # <script parsing and execution>
   from snowmobile import snowscripter 
   
   # Creating connected script objects from paths on a specified connection
   script_objs = [snowscripter.Script(path, connector=user_conn) for path in paths_to_sql]
   
   # Run all scripts sequentially
   for script in script_objs:
      print(f"Running {script.source}")
      script.run()
   

Modules

All modules are included in the build for transparency & flexibility purposes, although the majority of use-cases will run on the front-end modules that make use of the others along the way.

Front-end / primary utilities

• snowquery instantiates a connection and provides an execute_query() method for executing statements against and querying data from the warehouse
• snowloader flexibly loads data from a local DataFrame into a table within the warehouse in the form of a df_to_snowflake() function, bundling a variety of utilities that standardize column names pre-loading, check DataFrame's structure compared to the table to be loaded into, and executes DDL in absence of a pre-existing table
• snowscripter parses and instantiates components of raw .sql scripts into Python objects for much easier interaction, particularly as it relates to single-statement execution and rendering statements as markdown when executed in IPython environments as well as easy execution of full .sql files

Back-end / supporting

• snowcreds locates the credentials file on a user's local file system
• snowconn establishes a connection to the database and returns a conn object

A more in-depth description of of each module and its usage outlined below.

---

snowquery

snowquery simplifies the execution of sql statements against the database via an execute_query() method, using pandas' pd.read_sql function to execute the SQL and returning results from the DataBase as a dataframe by default.

please note: snowquery is intended to streamline execution of sql that is typed within a Python script and is better-suited for ad-hoc statements whereas snowscripter imports an external .sql file & extracts its components into Python objects that come with cleaner methods for execution.

Usage

Its usage to query via set of credentials stored in snowflake_credentials.json labeled SANDBOX is as follows.

In[1]:
from snowmobile import snowquery

# Establishes snowquery.Connector() object with which statements can be excuted 
sf = snowquery.Connector(conn_name='SANDBOX')

# Executing a simple sql string
sample = sf.execute_query('select * from sample_sandbox_table')

# Verifying returned object
type(sample)

Out[1]
pandas.core.frame.DataFrame

---

snowscripter

snowscripter imports an external .sql file & transforms it into Python objects on which methods can be called to perform a variety of actions.

The Script object is instantiated with the following three arguments, of which only path is required for successful instantiation.

    def __init__(self, path: str, pattern: str = r"/\*-(\w.*)-\*/",
                 connector: snowquery.Connector = ''):
        """Instantiating an instance of 'script' by calling Script class on a
        full path to a sql script.

        Args:
            path: Full path to SQL script including .sql extension
            pattern: Regex pattern that SQL statement headers are wrapped in
            connector: Instantiated snowquery.Connector instance to use in the
            execution of Script or Statement objects
        """ 

The usage for this module is broken up into script-level and statement-level usage below, the latter of which contains the majority of application.

---

Usage (script-level)

In its simplest form, users can instantiate an instance of script by running the below where path is a full file path to a .sql file.

script = snowscripter.Script(path)

This will return a Script object on which the .run() method can be called to execute an entire script sequentially statement by statement.

---

Usage (statement-level)

Background

The real benefit of snowscripter comes from the use of 'header' tags within the sql script that form a link between a specified name and an individual statement - these are identified in the parser by the regex pattern contained in the pattern argument of the object's instantiation.

The pattern's default is '/\*-(\w.*)-\*/' and will return all text between a standard sql block-comment whose contents are wrapped in an additional '-', such as:

/*-sample_statement_header-*/
select 
   a.*
from...

To make this more clear, a usage example is outlined in the below in which snowscripter is used to transform a few simple sql statements and execute them against a sample table loaded in the warehouse.

---

Snowmobile.snowscripter Example Usage

Setup / creating data

The following Python snippet creates a dummy DataFrame and loads it to the warehouse for use during the exercise

# Up-front setup for snowscripter usage
import numpy as np
import pandas as pd
from snowmobile import snowloader, snowquery

# Creating dummy df 
df = pd.DataFrame({f"col{i}": 
                   np.random.normal(0, 1, 1000) for i in range(0, 10)}).reset_index()

# Instantiating instance of a specified connection to run on for demo 
demo_conn = snowquery.Connector('demo')

# Loading into a table called SAMPLE_TABLE
snowloader.df_to_snowflake(df, table_name='SAMPLE_TABLE', force_recreate=True, connector=demo_conn)

1. Local sql script (snowscripter_sample.sql)

Now that we have a sample table to query against, consider two statements stored in the .sql file below.

--snowscripter_sample.sql

/*-contrived_example_aggregation-*/
create or replace temp table index_agg as with
contrived_1 as (
  select
    index
    ,sum(col1) as col1
  from SAMPLE_TABLE
  group by 1
),
contrived_2 as (
  select
    index
    ,sum(col1) as col2
  from SAMPLE_TABLE
  group by 1
)
  select
    a.*
    ,b.col2
  from contrived_1 a
  inner join contrived_2 b
    on a.index = b.index;

/*-verify_contrived_join-*/
SELECT
  index
	,count(*)
		as cnt_all
FROM index_agg
group by 1
having count(*) <> 1;

2. Instantiating parsed script object

In Python, we can instantiate a snowscripter.Script object from this file with:

from snowmobile import snowscripter

# path_to_script = full path to .sql file
script = snowscripter.Script(path_to_script, snowflake=demo_conn)

3. Accessing & executing statements

Now that we have a script object in memory, we can work with different parts of our script either through the script object or extracting individual Statement objects & their associated methods.

Note: It's technically possible to access each statement as a raw string by pulling it out of the script object's namespace with:

script.statement.get('contrived_example_aggregation')

However, this isn't recommended as manually executing all those strings is significantly less convenient than the following two options outlined below.

---

Option 1: Access as a Statement object via the script's .fetch() method

sample_statement_obj = script.fetch('contrived_example_aggregation')
type(sample_statement_obj)  # snowmobile.snowscripter.Statement

This method is preferred because the snowmobile.snowscripter.Statement object comes with the following three methods:

• .execute() to execute
• .render() to render the syntactically-highlighted code as a markdown in IPython environments
• .raw() to return the raw sql as a string

Option 2: Access Statement objects for all statements via the script's .get_statements() method

This will return an iterable containing individual Statement objects for all statements in the script

iterable_statements = script.get_statements()
for statement_header, statement in iterable_statements.items():
  # 'statement_header' will iterate through [contrived_example_aggregation, verify_contrived_join]
  # 'statement' will be the Statement objects associated with each of the headers

---

4. Executing and rendering statements simultaneously

Lastly, it's often helpful to execute a statement as well as render the sql behind it or see descriptions of results, particularly when troubleshooting a broken pipeline or trying to determine at what point in a lengthy sql script the data stops representing what we think it does.

To avoid the need to call multiple methods in these instances, the .execute() method is defined with the following arguments allowing for this flexibility.

    def execute(self, results: bool = True, render: bool = False,
                describe: bool = False) -> object:
        """Executes sql with option to return results / render sql as Markdown.

        Args:
            results: Boolean indicating whether or not to
            return results
            render: Boolean indicating whether or not to render the
            raw sql as markdown
            describe: Boolean indicating whether or not to
            print output of a pandas df.describe() on returned results (mostly
            useful for QA queries that are expected to return null-sets)
        """

Resulting in the following functionality within IPython/Jupyter environments

End Snowmobile.snowscripter Example Usage

---

snowloader

snowloader streamlines the bulk-loading protocol outlined in the Snowflake documentation in the form of a df_to_snowflake() function and is intended to be a one-stop solution for the quick loading of data.

Its main features are:

• Standardizing of DataFrame's columns prior to loading into the warehouse
• DDL creation & execution if a pre-defined table to load data into doesn't exist
• Parameter-based flexibility to append DataFrame's contents or replace pre-existing contents
• Returns a boolean indicating whether or not a load was successful for exception-handling when iteratively loading/appending multiple files into a single table.

Usage

Continuing on the above example, the below will convert all columns in the sample_table DataFrame to floats and load it into the warehouse, executing new-DDL to overwrite an existing table or create one in its absence.

import numpy as np
import pandas as pd
from snowmobile import snowloader, snowquery

# Instantiating instance of a specified connection to run on for demo 
demo_conn = snowquery.Connector('demo')

# Selecting all value from dummy table created in 'snowscripter' usage
df = demo_conn.execute_query('select * from sample_table')

# Converting all numeric values to floats (index dropping/adding to handle
# the index along the way/not change the structure of the final table)
df = df.drop(columns=[col for col in df.columns if 'index' in 
                      col]).reset_index()
df = df.applymap(float).reset_index()

# Option 1 
snowloader.df_to_snowflake(df=df, table_name='SAMPLE_TABLE', force_recreate=True,
                            snowflake=demo_conn)

# Option 2 
snowloader.df_to_snowflake(df=df, table_name='SAMPLE_TABLE', force_recreate=True)

In the above,

• Option 1 will load the data back into Snowflake on the same connection that was established in the sf = snowquery.Snowflake(conn_name='SANDBOX') statement by use of the snowflake=sf parameter
• In Option 2 this argument is omitted and the function will instantiate a new connection based on the first set of credentials in snowflake_credentials.json

In general and particularly when iteratively loaded multiple files into the database, it will be faster to instantiate a single instance of snowquery that's passed into the df_to_snowflake() function so that it does not need to find, read-in and parse the credentials file each time its called.

---

snowcreds

Description

snowcreds is a single class intentionally extracted for easier evolving along with security standards, its instantiation of Credentials() accepts the below two arguments and associated defaults

def __init__(self, config_file: str = 'snowflake_credentials.json',
                 conn_name: str = '') -> None:
        """Instantiates an instance of credentials file.

        Args:
            config_file: Name of .json configuration file following the
            format of connection_credentials_SAMPLE.json.
            conn_name: Name of connection within json file to use, will
            use first set of credentials in the file if no argument is passed.
        """
        self.config_file = config_file
        self.conn_name = conn_name

It contains a single .get() method that will traverse a user's file system from the bottom-up until it finds a filename that matches the config_file parameter and unless specified otherwise via the conn_name parameter will return the first set of credentials stored in the .json file.

The .json file itself is assumed to store its credentials following this format

Usage

Note: The instantiation of snowcreds is somewhat verbose and has been left us such for the time being as it is a back-end utility and not intended to be called by the users explicitly

In[1]:
from snowmobile import snowcreds
creds = snowcreds.Credentials(config_file='sample_credentials.json').get()

Out[1]:
Locating & importing credentials..
	<1 of 4> Searching for sample_credentials.json in local file system..
	<2 of 4> Located & loaded sample_credentials.json from:
		C:\User\Desktop\sample_credentials.json
	<3 of 4> No explicit connection passed, fetching 'Connection1' credentials by default
	<4 of 4> Successfully imported credentials for conn_name='Connection1'

---

snowconn

Description

snowconn is also comprised of a single class, Connection(), that inherits Credentials() to retrieve a set of credentials with which to establish a connection to the database.

Its instantiation and usage is very similar to snowcreds as it inherits the config_file and conn_name attributes and includes a .get_conn() method will authenticate using the credentials returned by snowcreds.get()

Usage

The below codes instantiate an instance of Connection used in higher-level modules.

In[1]:
from snowmobile import snowconn
creds = snowconn.Connection().get_conn()
type(creds)

Out[1]:
snowflake.connector.connection.SnowflakeConnection

---

---

Appendix

snowscripter

The below outputs all the attributes and methods associated with an instantiated script object

In[1]:
attrs = {k: v for k, v in script.__dict__.items()}

print("---------/Attributes/---------")
for i, (k, v) in enumerate(attrs.items(), start=1):
    print(f"<a{i}> {k}:\n\t{type(v)}")


print("\n----------/Methods/---------")
for i2, k in enumerate(script.__dir__(), start=1):
    if str(k) not in list(attrs.keys()) and '__' not in str(k):
        print(f"<m{i2-18}> .{k}()")
    else:
        i2 -= 1

Out[1]:
---------/Attributes/---------
<a1> sql:
	<class 'snowmobile.snowscripter.Script'>
<a2> snowflake:
	<class 'snowmobile.snowquery.Connector'>
<a3> pattern:
	<class 're.Pattern'>
<a4> source:
	<class 'str'>
<a5> name:
	<class 'str'>
<a6> script_txt:
	<class 'str'>
<a7> list_of_statements:
	<class 'list'>
<a8> statement:
	<class 'str'>
<a9> statement_names:
	<class 'list'>
<a10> statements:
	<class 'dict'>
<a11> spans:
	<class 'dict'>
<a12> ordered_statements:
	<class 'list'>
<a13> header_statements:
	<class 'list'>
<a14> full_sql:
	<class 'str'>
<a15> returned:
	<class 'dict'>

----------/Methods/---------
<m1> .reload_source()
<m2> .run()
<m3> .get_statements()
<m4> .fetch()
<m5> .render()
<m6> .raw()
<m7> .execute()