snowpark-extensions 0.0.39

A set of helpers to extend snowpark functionality

snowpark-extensions-py

Snowpark by itself is a powerful library, but still some utility functions can always help.

BTW what about Java/ Scala/ SQL ? There is an additional repo where you will have also utility functions and extensions for those technologies.

NOTE: we have been working to integrate some of the snowpark extensions directly into the snowpark-python library. In most cases the APIs will be exactly the same, so there should no changes needed in your code. However there might be breaking changes, so consider that before updating. If any of these breaking changes are affecting you, please enter an issue so we can address it.

Installation

We recommended installing using PYPI

    $ pip install snowpark-extensions

note:: If you run this command on MacOS change pip by pip3

Usage

just import it at the top of your file and it will automatically extend your snowpark package. For example:

from snowflake.snowpark import Session
import snowpark_extensions
new_session = Session.builder.from_snowsql().appName("app1").getOrCreate()

Currently provided extensions:

Session Extensions

Session was extened to support IPython display. Using session as a value in a cell will display the session info

SessionBuilder extensions

Name	Description
SessionBuilder.from_snowsql	can read the information from the snowsql config file by default at ~/snowsql/config or at a given location
SessionBuilder.env	reads settings from SNOW_xxx or SNOWSQL_xxx variables
SessionBuilder.appName	Sets a query tag with the given appName Available as SessionBuilder.app_name since 1.13.0
SessionBuilder.append_tag	Appends a new tag to the existing query tag Use Session.append_query_tag or Session.update_query_tag instead available since 1.12.0
SessionBuilder.getOrCreate	Available in snowpark-python >= 1.3.0

You can the create your session like:

from snowflake.snowpark import Session
import snowpark_extensions
new_session = Session.builder.from_snowsql().appName("app1").create()

from snowflake.snowpark import Session
import snowpark_extensions
new_session = Session.builder.env().appName("app1").create()

NOTE: since 1.8.0 the python connector was updated and we provide support for an unified configuration storage for snowflake-python-connector and snowflake-snowpark-python with this approach.

You can use this connections leveraging Session.builder.getOrCreate() or Session.builder.create()

By default, we look for the connections.toml file in the location specified in the SNOWFLAKE_HOME environment variable (default: ~/.snowflake). If this folder does not exist, the Python connector looks for the file in the platformdirs location, as follows:

• On Linux: ~/.config/snowflake/, but follows XDG settings
• On Mac: ~/Library/Application Support/snowflake/
• On Windows: %USERPROFILE%\AppData\Local\snowflake\

The default connection by default is 'default' but it can be controlled with the environment variable: SNOWFLAKE_DEFAULT_CONNECTION_NAME.

If you dont want to use a file you can set the file contents thru the SNOWFLAKE_CONNECTIONS environment variable.

Connection file looks like:

[default]
account = "myaccount"
user = "user1"
password = 'xxxxx'
role = "user_role"
database = "demodb"
schema = "public"
warehouse = "load_wh"


[snowpark]
account = "myaccount"
user = "user2"
password = 'yyyyy'
role = "user_role"
database = "demodb"
schema = "public"
warehouse = "load_wh"

The Session.app_name or Session.append_query_tag can used to setup a query_tag like APPNAME=tag;execution_id=guid which can then be used to track job actions with a query like

You can then use a query like: To see all executions from an app or

select *
from table(information_schema.query_history())
whery query_tag like '%APPNAME=tag%'
order by start_time desc;

To see the executions for a particular execution:

select *
from table(information_schema.query_history())
whery query_tag like '%APPNAME=tag;execution_id=guid%'
order by start_time desc;

Column Extensions

Name	Description
Column.getItem	An expression that gets an item at position ordinal out of a list, or gets an item by key out of a dict.  Available in snowpark-python >= 1.3.0

DataFrame Extensions

Name	Description
DataFrame.dtypes	~~returns the list of datatypes in the DataFrame ~~Available in snowpark python >= 1.1.0
DataFrame.map	provides an equivalent for the map function for example df.map(func,input_types=[StringType(),StringType()],output_types=[StringType(),IntegerType()],to_row=True)
DataFrame.simple_map	if a simple lambda like lambda x: x.col1 + x.col2 is used this functions can be used like df.simple_map(lambda x: x.col1 + x.col2)
DataFrame.groupby.applyInPandas	Maps each group of the current DataFrame using a pandas udf and returns the result as a DataFrame. applyInPandas overload is kept to avoid breaking changes. But we recommend using the native apply_in_pandas Available in snowpark-python >= 1.8.0
DataFrame.replace	extends replace to allow using a regex
DataFrame.groupBy.pivot	extends the snowpark groupby to add a pivot operator
DataFrame.stack	This is an operator similar to the unpivot operator

Examples

map and simple_map

from snowflake.snowpark import Session
from snowflake.snowpark.types import *
import snowpark_extensions

session = Session.builder.from_snowsql().appName("app1").getOrCreate()
  
data = [('James','Smith','M',30),('Anna','Rose','F',41),('Robert','Williams','M',62)]
columns = ["firstname","lastname","gender","salary"]
df = session.createDataFrame(data=data, schema = columns)
df.show()

--------------------------------------------------
|"FIRSTNAME"  |"LASTNAME"  |"GENDER"  |"SALARY"  |
--------------------------------------------------
|James        |Smith       |M         |30        |
|Anna         |Rose        |F         |41        |
|Robert       |Williams    |M         |62        |
--------------------------------------------------

# using map with a lambda, the to_row indicates that the code will pass a row as x to the lambda
# if you have a lambda like lambda x,y,z you can use to_row=False
df2=df.map(lambda x: 
        (x[0]+","+x[1],x[2],x[3]*2),
        output_types=[StringType(),StringType(),IntegerType()],to_row=True)
df2.show()

-----------------------------------
|"C_1"            |"C_2"  |"C_3"  |
-----------------------------------
|James,Smith      |M      |60     |
|Anna,Rose        |F      |82     |
|Robert,Williams  |M      |124    |
-----------------------------------

# for simple lambda
# simple map will just pass the same dataframe to the function
# this approach is faster
df2 = df.simple_map(lambda x: (x[0]+","+x[1],x[2],x[3]*2))
df2.toDF(["name","gender","new_salary"]).show()

---------------------------------------------
|"NAME"           |"GENDER"  |"NEW_SALARY"  |
---------------------------------------------
|James,Smith      |M         |60            |
|Anna,Rose        |F         |82            |
|Robert,Williams  |M         |124           |
---------------------------------------------

replace with support for regex

df = session.createDataFrame([('bat',1,'abc'),('foo',2,'bar'),('bait',3,'xyz')],['A','C','B'])
# already supported replace
df.replace(to_replace=1, value=100).show()
# replace with regex
df.replace(to_replace=r'^ba.$', value='new',regex=True).show()

applyInPandas

from snowflake.snowpark import Session
import snowpark_extensions
session = Session.builder.from_snowsql().getOrCreate()
import pandas as pd  
df = session.createDataFrame(
    [(1, 1.0), (1, 2.0), (2, 3.0), (2, 5.0), (2, 10.0)],
    schema=["ID", "V"])
df1 = df.to_pandas()
def normalize(pdf):
    V = pdf.V
    return pdf.assign(V=(V - V.mean()) / V.std())
df2 = normalize(df1)
# schema can be an string or an StructType
df.group_by("ID").applyInPandas(
    normalize, schema="id long, v double").show()  

------------------------------
|"ID"  |"V"                  |
------------------------------
|2     |-0.8320502943378437  |
|2     |-0.2773500981126146  |
|2     |1.1094003924504583   |
|1     |-0.7071067811865475  |
|1     |0.7071067811865475   |
------------------------------

NOTE: since snowflake-snowpark-python==1.8.0 applyInPandas is available. This version is kept because:

1. It supports string schemas

2. It automatically wraps the column names. In snowpark applyInPandas you need to do:

   def func(pdf):
       pdf.columns = ['columnname1','columnname2']
       # rest of the code
   

   Before using your function, to guarantee the proper names are used. This implementation will just use the DF column names. Take in consideration that this still might imply changes as metadata in SF is upper case and lowercase references like df['v'] might fail.

In general it is recommended you use the the snowpark built-in. The extensions only overwrite applyInPandas, the apply_in_pandas refers to the official snowpark implementation

stack

Assuming you have a DataTable like:

  +-------+---------+-----+---------+----+
  |   Name|Analytics|   BI|Ingestion|  ML|
  +-------+---------+-----+---------+----+
  | Mickey|     null|12000|     null|8000|
  | Martin|     null| 5000|     null|null|
  |  Jerry|     null| null|     1000|null|
  |  Riley|     null| null|     null|9000|
  | Donald|     1000| null|     null|null|
  |   John|     null| null|     1000|null|
  |Patrick|     null| null|     null|1000|
  |  Emily|     8000| null|     3000|null|
  |   Arya|    10000| null|     2000|null|
  +-------+---------+-----+---------+----+  

df.select("NAME",df.stack(4,lit('Analytics'), "ANALYTICS", lit('BI'), "BI", lit('Ingestion'), "INGESTION", lit('ML'), "ML").alias("Project", "Cost_To_Project")).filter(col("Cost_To_Project").is_not_null()).orderBy("NAME","Project")

That will return:

'-------------------------------------------
|"NAME"   |"PROJECT"  |"COST_TO_PROJECT"  |
-------------------------------------------
|Arya     |Analytics  |10000              |
|Arya     |Ingestion  |2000               |
|Donald   |Analytics  |1000               |
|Emily    |Analytics  |8000               |
|Emily    |Ingestion  |3000               |
|Jerry    |Ingestion  |1000               |
|John     |Ingestion  |1000               |
|Martin   |BI         |5000               |
|Mickey   |BI         |12000              |
|Mickey   |ML         |8000               |
|Patrick  |ML         |1000               |
|Riley    |ML         |9000               |
-------------------------------------------

DataFrameReader Extensions

Name	Description
DataFrameReader.format	Specified the format of the file to load
DataFrameReader.load	Loads a dataframe from a file. It will upload the files to an stage if needed

Example

Functions Extensions

Name	Description
functions.array_sort	sorts the input array in ascending order or descending order. The elements of the input array must be orderable. Null elements will be placed at the end of the returned array. Available in snowpark-python >= 1.6.1
functions.unix_timestamp	returns the UNIX timestamp of current time. Available in snowpark-python >= 1.1.0
functions.from_unixtimestamp	can be used to convert UNIX time to Snowflake timestamp Available in snowpark-python >= 1.1.0
functions.to_utc_timestamp	converts a timezone-agnostic timestamp to a timezone-aware timestamp in the provided timezone before rendering that timestamp in UTC**Available in snowpark-python >= 1.11.1**
functions.to_utc_timestamp_ext	converts a timezone-agnostic timestamp to a timezone-aware timestamp in the provided timezone before rendering that timestamp in UTC it supports timezone names like
functions.format_number	formats numbers using the specified number of decimal places
functions.reverse	returns a reversed string Available in snowpark-python >= 1.2.0
functions.explode	returns a new row for each element in the given array Available in snowpark-python >= 1.4.0
functions.explode_outer	returns a new row for each element in the given array or map. Unlike explode, if the array/map is null or empty then null is producedThis Available in snowpark-python >= 1.4.0 There is a breaking change as the explode_outer does not need the map argument anymore.
functions.arrays_zip	returns a merged array of arrays
functions.array_sort	sorts the input array in ascending order. The elements of the input array must be orderable. Null elements will be placed at the end of the returned array. Available in snowpark-python >= 1.6.1
functions.array_max	returns the maximon value of the array. Available in snowpark-python >= 1.6.1
functions.array_min	returns the minimum value of the array. Available in snowpark-python >= 1.6.1
functions.array_distinct	removes duplicate values from the array. Available in snowpark-python >= 1.4.0
function.daydiff	this function returns the difference in days between two dates. This function will be direct equivalent of the spark datediff. You can simple replace spark datediff by daydiff Available in snowpark-python >= 1.4.0
functions.date_add	returns the date that is n days days after Available in snowpark-python >= 1.4.0
functions.date_sub	returns the date that is n days before Available in snowpark-python >= 1.4.0
functions.regexp_extract	extract a specific group matched by a regex, from the specified string column. Available in snowpark-python >= 1.4.0
functions.regexp_split	splits a specific group matched by a regex, it is an extension of split wich supports a limit parameter.
functions.asc	returns a sort expression based on the ascending order of the given column name. Available in snowpark-python >=1.1.0
functions.desc	returns a sort expression based on the descending order of the given column name. Available in snowpark-python >=1.1.0
functions.flatten	creates a single array from an array of arrays
functions.sort_array	sorts the input array in ascending or descending order according to the natural ordering of the array elements. Null elements will be placed at the beginning of the returned array in ascending order or at the end of the returned array in descending orderAvailable in snowpark-python >= 1.6.1
functions.map_values	Returns an unordered array containing the values of the map.
functions.struct	Returns an object built with the given columns Available in snowpark-python >= 1.4.0
functions.bround	This function receives a column with a number and rounds it to scale decimal places with HALF_EVEN round mode, often called as "Banker's rounding" . This means that if the number is at the same distance from an even or odd number, it will round to the even number. Available in snowpark-python >= 1.4.0

regexp_extract

session = Session.builder.from_snowsql().create()


df = session.createDataFrame([('100-200',)], ['str'])
res = df.select(F.regexp_extract('str',r'(\d+)-(\d+)',1).alias('d')).collect()
print(str(res))
# [Row(D='1')]

df = session.createDataFrame([['id_20_30', 10], ['id_40_50', 30]], ['id', 'age'])
df.show()
# --------------------
# |"ID"      |"AGE"  |
# --------------------
# |id_20_30  |10     |
# |id_40_50  |30     |
# --------------------


df.select(F.regexp_extract('id', r'(\d+)', 1)).show()
# ------------------------------------------------------
# |"COALESCE(REGEXP_SUBSTR(""ID"", '(\\D+)', 1, 1,...  |
# ------------------------------------------------------
# |20                                                  |
# |40                                                  |
# ------------------------------------------------------


df.select(F.regexp_extract('id', r'(\d+)_(\d+)', 2)).show()
# ------------------------------------------------------
# |"COALESCE(REGEXP_SUBSTR(""ID"", '(\\D+)_(\\D+)'...  |
# ------------------------------------------------------
# |30                                                  |
# |50                                                  |
# ------------------------------------------------------

regexp_split

session = Session.builder.from_snowsql().create()

df = session.createDataFrame([('oneAtwoBthreeC',)], ['s',])
res = df.select(regexp_split(df.s, '[ABC]', 2).alias('s')).collect()
print(str(res))
# [\n  "one",\n  "twoBthreeC"\n]

utilities

Name	Description
utils.map_to_python_type	maps from DataType to python type
utils.map_string_type_to_datatype	maps a type by name to a snowpark DataType
utils.schema_str_to_schema	maps an schema specified as an string to a StructType()

Jupyter Notebook support

A Jupyter extension has been created to allow integration in Jupyter notebooks. This extension implements a SQL magic, enabling users to run SQL commands within the Jupyter environment. This enhances the functionality of Jupyter notebooks and makes it easier for users to access and analyze their data using SQL. With this extension, data analysis becomes more streamlined, as users can execute SQL commands directly in the same environment where they are working on their notebooks.

To enable this extension just import the snowpark extensions module

import snowpark_extensions

After import a %%sql magic can be used to run queries. For example:

%%sql
select * from table1

Queries can use also use Jinja2 syntax. For example:

If a previous cell you had something like:

COL1=1

Then on following cells you can do:

%%sql
select * from tables where col={{COL1}}

You can give a name to the sql that you can use later for example:

%%sql tables
select * from information_schema.tables limit 5

and then use that as a normal dataframe:

if tables.count() > 5:
    print("There are more that 5 tables")

If you dont specify a name you can still access the last result using __df.

NOTE: By default only 50 rows are displays. You can customize this limit for example to 100 rows with:

DataFrame.__rows_count = 1000

You can configure Jupyter to run some imports and initialization code at the start of a notebook by creating a file called startup.ipy in the ~/.ipython/profile_default/startup directory.

Any code written in this file will be executed when you start a new Jupyter notebook.

An example startup.ipy is provided

Logging Extras

Snowpark Tags

Bart, an Snowflake Evangelist from EMEA created this amazing utility. We are just wrapping it here in the extensions. See his post for a great description https://medium.com/snowflake/simple-tags-in-snowflake-snowpark-for-python-c5910749273

You can use in your snowpark procedures:

def process(session,...):
    @Tag()
    def foo(...):
        ....
    # or inside your code:
    def goo(...):
        # only queries from this context will be tagged
        with Tag(session, f"drop_in_do_it_too"):
            session.sql(f'''DROP TABLE IF EXISTS {to_table}''').collect()