Open in app

Sign in

Write

Sign in

Snowflake Schema Detection Feature for CSV Files

Data Engineering Simplified
8 min readAug 1, 2023

--

Check out my courses in Udemy. Popular courses include building utility using Snowpark Python API to ingest CSV data as well as JSON data automatically without writing any DDL/DML statement and save 95% of manual effort.

https://youtu.be/nb1iNEn5YjI

Introduction

Data loading processes are crucial for timely insights and decision-making. Snowflake, has add its support for CSV & JSON using “Infer Schema” table function that streamlines the process of loading data from CSV files into Snowflake tables. We will explore the importance of this feature and how it empowers data engineers and analysts to expedite data loading while ensuring accuracy and consistency.

Importance of Infer Schema in Snowflake

Traditionally, defining table structure (DDL) while loading data can be a manual and error-prone task, especially when dealing with large datasets. With Snowflake’s “Infer Schema” feature for CSV and JSON, this tedious process is simplified. The feature automatically reads a CSV file from a specified stage location(s) and intelligently infers column names and data types based on the CSV file’s contents. This automation not only saves valuable time and effort but also reduces the risk of schema mismatches, ensuring data integrity throughout the data pipeline.

How to Use Infer Schema in Snowflake

To utilize this powerful feature, data practitioners need to follow a few simple steps.

  1. Ensure that the CSV file is staged in Snowflake with appropriate permissions.
  2. Next, create a new table using the “CREATE TABLE” statement, and instead of specifying explicit column names and data types, use “INFER SCHEMA” as the column definition.
  3. Snowflake will automatically analyze the CSV data file(s) and create the table DDL accordingly.

Watch the video below to see the Infer Schema feature in action and discover how it can efficiency into your data loading workflows

Data Files

Download the CSV data files from this git location.

Part-1 : CSV Infer Schema SQL Construct

order_data_10.csv

order_id,order_date,order_time,product_name,order_price,tax_value,existing_customer,customer_name,mobile_number
1,2023-06-04,13:19:14,Moto G Power,956.27,95.63,False,Robert Sherman,9152218975
2,2023-06-19,06:44:23,Nokia 8.3,658.97,65.9,True,Raymond Phillips,9872334088
3,2023-04-08,05:11:22,Moto G Power,1487.21,148.72,False,Thomas Thompson,9513069603
4,2023-06-06,03:33:53,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,9620517344
5,2023-02-12,17:30:23,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,9045335526
6,2023-02-07,15:41:14,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,9377330463
7,2023-06-15,12:10:09,iPhone SE,593.82,59.38,False,Timothy Guerrero,9785992101
8,2023-01-01,00:33:04,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,9324423643
9,2023-01-26,16:47:11,iPhone SE,1051.47,105.15,False,Sara Williams,9854478422
10,2023-05-21,00:11:13,Galaxy S21,687.52,68.75,False,Donna Velez,9876003384
-- setting context for this worksheet.
use role sysadmin;
use database demo;
use schema sales;
use warehouse compute_wh;

select * from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10.csv',
      file_format=>'my_csv_format'
    )
  );

list @order_data;

create or replace file format my_file_format_01 
    type = 'csv' 
    compression = 'auto' 
    field_delimiter = ',' 
    record_delimiter = '\n'  
    field_optionally_enclosed_by = '\042' 
    skip_header = 1;

desc file format my_file_format_01;

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10.csv 
(file_format => my_file_format_01) t;

select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10.csv',
      file_format=>'my_file_format_01'
    )
  );

select    
    current_region() as SF_Hosted_Region, 
    current_version() as Snowflake_Version;

create table my_order_tbl
  using template (
    select array_agg(object_construct(*))
      from table(
        infer_schema(
            location=>'@order_data/csv/records=10/order_data_10.csv',
            file_format=>'my_csv_format'
    )
));

desc table my_order_tbl;
select * from my_order_tbl;

Part-02 SQL Scripts — Performance



create or replace file format my_file_format_02 
    type = 'csv' 
    compression = 'auto' 
    field_delimiter = ',' 
    record_delimiter = '\n'  
    field_optionally_enclosed_by = '\042' 
    parse_header = true;
    
list @order_data/csv/records;

alter session set use_cached_result = false;

use warehouse wh01;
-- csv file having only 10 orders
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10.csv',
      file_format=>'my_file_format_02'
    )
  );

-- csv file having only 1000 orders
use warehouse wh02;
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=1000/order_data_1k.csv',
      file_format=>'my_file_format_02'
    )
  );

-- csv file having only 10,000 orders
use warehouse wh03;
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10000/order_data_10k.csv',
      file_format=>'my_file_format_02'
    )
  );

-- csv file having only 1,000,000 orders
use warehouse wh04;
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=1000000/order_data_1M.csv',
      file_format=>'my_file_format_02'
    )
  );

-- partition data and in infer-schema
list @order_data/partition-csv;

select *
  from table(
    infer_schema(
      location=>'@order_data/partition-csv',
      file_format=>'my_file_format_02'
    )
  );

Part-03 SQL Scripts Data Types

order_data_10_array.csv

order_id,order_date,order_time,product_name,order_price,tax_value,existing_customer,customer_name,mobile_number
1,2023-06-04,13:19:14,Moto G Power,956.27,[95.63,False,Robert Sherman,[9152218975,1234567890]
2,2023-06-19,06:44:23,Nokia 8.3,658.97,65.9,True,Raymond Phillips,[9872334088,1234567890]
3,2023-04-08,05:11:22,Moto G Power,1487.21,148.72,False,Thomas Thompson,[9513069603,1234567890]
4,2023-06-06,03:33:53,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,[9620517344,1234567890]
5,2023-02-12,17:30:23,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,[9045335526,1234567890]
6,2023-02-07,15:41:14,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,[9377330463,1234567890]
7,2023-06-15,12:10:09,iPhone SE,593.82,59.38,False,Timothy Guerrero,[9785992101,1234567890]
8,2023-01-01,00:33:04,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,[9324423643,1234567890]
9,2023-01-26,16:47:11,iPhone SE,1051.47,105.15,False,Sara Williams,[9854478422,1234567890]
10,2023-05-21,00:11:13,Galaxy S21,687.52,68.75,False,Donna Velez,[9876003384,1234567890]

order_data_10_json.csv

order_id,order_date,order_time,product_name,order_price,tax_value,existing_customer,customer_name,mobile_number
1,2023-06-04,13:19:14,Moto G Power,956.27,95.63,False,Robert Sherman,'{"mobile": "9152218975", "desk": "12345678"}'
2,2023-06-19,06:44:23,Nokia 8.3,658.97,65.9,True,Raymond Phillips,'{"mobile": "9152218975", "desk": "12345678"}'
3,2023-04-08,05:11:22,Moto G Power,1487.21,148.72,False,Thomas Thompson,'{"mobile": "9152218975", "desk": "12345678"}'
4,2023-06-06,03:33:53,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,'{"mobile": "9152218975", "desk": "12345678"}'
5,2023-02-12,17:30:23,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,'{"mobile": "9152218975", "desk": "12345678"}'
6,2023-02-07,15:41:14,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,'{"mobile": "9152218975", "desk": "12345678"}'
7,2023-06-15,12:10:09,iPhone SE,593.82,59.38,False,Timothy Guerrero,'{"mobile": "9152218975", "desk": "12345678"}'
8,2023-01-01,00:33:04,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,'{"mobile": "9152218975", "desk": "12345678"}'
9,2023-01-26,16:47:11,iPhone SE,1051.47,105.15,False,Sara Williams,'{"mobile": "9152218975", "desk": "12345678"}'
10,2023-05-21,00:11:13,Galaxy S21,687.52,68.75,False,Donna Velez,'{"mobile": "9152218975", "desk": "12345678"}'

order_data_10_timestamp.csv

order_id,order_date,product_name,order_price,tax_value,existing_customer,customer_name,mobile_number
1,2023-06-04 13:19:14.109,Moto G Power,956.27,95.63,False,Robert Sherman,9152218975
2,2023-06-19 06:44:23.208,Nokia 8.3,658.97,65.9,True,Raymond Phillips,9872334088
3,2023-04-08 05:11:22.307,Moto G Power,1487.21,148.72,False,Thomas Thompson,9513069603
4,2023-06-06 03:33:53.406,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,9620517344
5,2023-02-12 17:30:23.505,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,9045335526
6,2023-02-07 15:41:14.604,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,9377330463
7,2023-06-15 12:10:09.703,iPhone SE,593.82,59.38,False,Timothy Guerrero,9785992101
8,2023-01-01 00:33:04.802,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,9324423643
9,2023-01-26 16:47:11.901,iPhone SE,1051.47,105.15,False,Sara Williams,9854478422
10,2023-05-21 00:11:13.999,Galaxy S21,687.52,68.75,False,Donna Velez,9876003384

-- data type - object
-- array
-- json
list @order_data/csv/records=10/order_data_10_array.csv; -- array
list @order_data/csv/records=10/order_data_10_json.csv; -- json
list @order_data/csv/records=10/order_data_10_timestamp.csv; -- timestamp

alter session set use_cached_result = false;

use warehouse wh01;

-- object data type
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_array.csv',
      file_format=>'my_file_format_02'
    )
  );

  -- json data type
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_json.csv',
      file_format=>'my_file_format_02'
    )
  );

  -- timestamp column type 
  select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_timestamp.csv',
      file_format=>'my_file_format_02'
    )
  );

Part-04 SQL Scripts

order_data_10_col_mismatch-case-1.csv

order_id,order_date,order_time,product_name,order_price,tax_value,existing_customer,customer_name,mobile_number
1,2023-06-04,13:19:14,Moto G Power,956.27,95.63,False,Robert Sherman
2,2023-06-19,06:44:23,Nokia 8.3,658.97,65.9,True,Raymond Phillips
3,2023-04-08,05:11:22,Moto G Power,1487.21,148.72,False,Thomas Thompson
4,2023-06-06,03:33:53,iPhone 12,1310.52,131.05,True,Casey Mcdaniel
5,2023-02-12,17:30:23,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin
6,2023-02-07,15:41:14,iPhone 12,1351.61,135.16,False,Dr. Michele Huang
7,2023-06-15,12:10:09,iPhone SE,593.82,59.38,False,Timothy Guerrero
8,2023-01-01,00:33:04,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins
9,2023-01-26,16:47:11,iPhone SE,1051.47,105.15,False,Sara Williams
10,2023-05-21,00:11:13,Galaxy S21,687.52,68.75,False,Donna Velez

order_data_10_col_mismatch-case-2.csv

order_id,order_date,order_time,product_name,order_price,tax_value,existing_customer,customer_name
1,2023-06-04,13:19:14,Moto G Power,956.27,95.63,False,Robert Sherman,9152218975
2,2023-06-19,06:44:23,Nokia 8.3,658.97,65.9,True,Raymond Phillips,9872334088
3,2023-04-08,05:11:22,Moto G Power,1487.21,148.72,False,Thomas Thompson,9513069603
4,2023-06-06,03:33:53,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,9620517344
5,2023-02-12,17:30:23,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,9045335526
6,2023-02-07,15:41:14,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,9377330463
7,2023-06-15,12:10:09,iPhone SE,593.82,59.38,False,Timothy Guerrero,9785992101
8,2023-01-01,00:33:04,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,9324423643
9,2023-01-26,16:47:11,iPhone SE,1051.47,105.15,False,Sara Williams,9854478422
10,2023-05-21,00:11:13,Galaxy S21,687.52,68.75,False,Donna Velez,9876003384

list @order_data/csv/records=10/order_data_10_col_mismatch-case-1.csv; -- 9 header vs 8 data fields
list @order_data/csv/records=10/order_data_10_col_mismatch-case-2.csv; -- 8 header vs 9 data fields

alter session set use_cached_result = false;

use warehouse wh01;

-- 9 header vs 8 data fields
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_col_mismatch-case-1.csv',
      file_format=>'my_file_format_01'
    )
  );

desc file format my_file_format_01;
desc file format my_file_format_02;

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10_col_mismatch-case-1.csv
(file_format => my_file_format_01) t;

-- 8 header vs 9 data fields
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_col_mismatch-case-2.csv',
      file_format=>'my_file_format_01'
    )
  );

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10_col_mismatch-case-2.csv
(file_format => my_file_format_01) t;
-- dont get confused.. and make sure, use proper file formats

-- double quot
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_all_quote.csv',
      file_format=>'my_file_format_02'
    )
  );

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10_all_quote.csv
(file_format => my_file_format_01) t;


-- double quote and escape charaters...
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_quote.csv',
      file_format=>'my_file_format_02'
    )
  );

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10_quote.csv
(file_format => my_file_format_01) t;


-- column names in ddl statement
create table my_order_tbl_01
  using template (
    select array_agg(object_construct(*))
      from table(
        infer_schema(
            location=>'@order_data/csv/records=10/order_data_10_all_quote.csv',
            file_format=>'my_file_format_02'
    )
));

select get_ddl('table','my_order_tbl_01')

Part-05 SQL Script

order_data_10_all_quote.csv

"order_id","order_date","order_time","product_name","order_price","tax_value","existing_customer","customer_name","mobile_number"
"1","2023-04-30","17:41:09","Nokia 8.3","291.09","29.11","False","Christina Garrett","9759437679"
"2","2023-01-13","11:22:40","OnePlus 9","678.22","67.82","True","Stephanie Flores","9119159206"
"3","2023-02-18","20:44:22","Pixel 5","1476.67","147.67","True","Christopher Johnson","9359736090"
"4","2023-07-11","09:53:46","iPhone 12","951.86","95.19","False","Sydney Wilson","9700131702"
"5","2023-04-18","22:11:08","iPhone SE","752.38","75.24","True","Chloe Jarvis","9645868562"
"6","2023-01-16","01:02:55","OnePlus 9","822.46","82.25","True","Michael Obrien V","9191187381"
"7","2023-04-17","21:44:38","Galaxy S21","1258.99","125.9","True","Nicholas Baker","9218283471"
"8","2023-06-02","20:08:29","Pixel 5","943.22","94.32","False","Jesse White","9541676867"
"9","2023-07-01","10:28:41","iPhone SE","1137.07","113.71","True","Jacob Ramirez","9141946204"
"10","2023-04-15","20:02:09","Moto G Power","1356.38","135.64","True","Joshua Carlson","9166630027"

list @order_data/csv/records=10/order_data_10_quote.csv; 

-- create a table using template
create or replace temporary table my_order_tbl_01
  using template (
    select array_agg(object_construct(*))
      from table(
        infer_schema(
             location=>'@order_data/csv/records=10/order_data_10_quote.csv',
              file_format=>'my_file_format_02'
    )
));

desc table my_order_tbl;

select get_ddl('table','my_order_tbl');

-- what happens if table already exist
-- can I create transient table
-- can I create temporary table

Part-06 SQL Scripts Input/Output

order_data_10_timestamp.csv

order_id,order_date,product_name,order_price,tax_value,existing_customer,customer_name,mobile_number
1,2023-06-04 13:19:14.109,Moto G Power,956.27,95.63,False,Robert Sherman,9152218975
2,2023-06-19 06:44:23.208,Nokia 8.3,658.97,65.9,True,Raymond Phillips,9872334088
3,2023-04-08 05:11:22.307,Moto G Power,1487.21,148.72,False,Thomas Thompson,9513069603
4,2023-06-06 03:33:53.406,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,9620517344
5,2023-02-12 17:30:23.505,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,9045335526
6,2023-02-07 15:41:14.604,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,9377330463
7,2023-06-15 12:10:09.703,iPhone SE,593.82,59.38,False,Timothy Guerrero,9785992101
8,2023-01-01 00:33:04.802,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,9324423643
9,2023-01-26 16:47:11.901,iPhone SE,1051.47,105.15,False,Sara Williams,9854478422
10,2023-05-21 00:11:13.999,Galaxy S21,687.52,68.75,False,Donna Velez,9876003384

order_data_10_case_sensitive.csv

ORDER_ID,ORDER_DATE,ORDER_TIME,PRODUCT_NAME,ORDER_PRICE,TAX_VALUE,EXISTING_CUSTOMER,customer_name,MOBILE_NUMBER
1,2023-06-04,13:19:14,Moto G Power,956.27,95.63,False,Robert Sherman,9152218975
2,2023-06-19,06:44:23,Nokia 8.3,658.97,65.9,True,Raymond Phillips,9872334088
3,2023-04-08,05:11:22,Moto G Power,1487.21,148.72,False,Thomas Thompson,9513069603
4,2023-06-06,03:33:53,iPhone 12,1310.52,131.05,True,Casey Mcdaniel,9620517344
5,2023-02-12,17:30:23,Galaxy Note 10,845.79,84.58,False,Jennifer Griffin,9045335526
6,2023-02-07,15:41:14,iPhone 12,1351.61,135.16,False,Dr. Michele Huang,9377330463
7,2023-06-15,12:10:09,iPhone SE,593.82,59.38,False,Timothy Guerrero,9785992101
8,2023-01-01,00:33:04,Galaxy Note 10,1334.09,133.41,False,Eric Hawkins,9324423643
9,2023-01-26,16:47:11,iPhone SE,1051.47,105.15,False,Sara Williams,9854478422
10,2023-05-21,00:11:13,Galaxy S21,687.52,68.75,False,Donna Velez,9876003384

-- input and outputs of infer-schema function


-- create a table using template
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10.csv',
      file_format=>'my_file_format_02'
    )
  );
select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_timestamp.csv',
      file_format=>'my_file_format_02'
    )
  );

--order_data_10_timestamp.csv

-- case sensitivity of the column headers

select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10_case_sensitive.csv',
      file_format=>'my_file_format_02'
      ,ignore_case => false
    )
  );
-- ,ignore_case => true


-- file names 
select *
  from table(
    infer_schema(
      location=>'@order_data/partition-csv',
      files => '/part_1.csv',
      file_format=>'my_file_format_02'
      ,ignore_case => true
    )
  );

Part-07 SQL Scripts

create or replace file format file_format_skip_header
    type = 'csv' 
    compression = 'auto' 
    field_delimiter = ',' 
    record_delimiter = '\n'  
    field_optionally_enclosed_by = '\042' 
    skip_header = 1;

create or replace file format file_format_parse_header
    type = 'csv' 
    compression = 'auto' 
    field_delimiter = ',' 
    record_delimiter = '\n'  
    field_optionally_enclosed_by = '\042' 
    parse_header = true;


-- file_format_parse_header and standard behaviour

select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10.csv',
      file_format=>'file_format_parse_header'
    )
  );

select *
  from table(
    infer_schema(
      location=>'@order_data/csv/records=10/order_data_10.csv',
      file_format=>'file_format_skip_header'
    )
  );

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10.csv 
(file_format => file_format_skip_header) t;

select t.$1,t.$2,t.$3,t.$4,t.$5,t.$6,t.$7,t.$8,t.$9
from @order_data/csv/records=10/order_data_10.csv 
(file_format => file_format_parse_header) t;

-- the copy command 
-- create a table and then run copy command using parse-header 
create or replace table my_order_tbl
  using template (
    select array_agg(object_construct(*))
      from table(
        infer_schema(
            location=>'@order_data/csv/records=10/order_data_10.csv',
            file_format=>'file_format_parse_header'
    )
));

copy into my_order_tbl from @order_data/csv/records=10/order_data_10.csv
FILE_FORMAT = (FORMAT_NAME = 'file_format_skip_header');

select * from my_order_tbl;

Part-08 SQL Scripts

-- we have already seen how to use using-teamplet to create tables 
-- lets explore this sql construc more

select array_agg(object_construct(*))
      from table(
        infer_schema(
            location=>'@order_data/csv/records=10/order_data_10.csv',
            file_format=>'my_csv_format'
        )
);
--array_agg(object_construct(*))
 
create table order_data_10
  using template (
    select array_agg(object_construct(*))
      from table(
        infer_schema(
            location=>'@order_data/csv/records=10/order_data_10.csv',
            file_format=>'my_csv_format'
    )
));

-- 1M records
create table order_data_1M
  using template (
    select array_agg(object_construct(*))
      from table(
        infer_schema(
            location=>'@order_data/csv/records=1000000/order_data_1M.csv',
            file_format=>'my_file_format_02'
    )
));


list @my_stage

Part-09 — Infer Schema CSV Limitations

Part-10- Snowpark Python API Automation

Click on this link to access the Snowpark Python API automation code

Snowpark Python
Detect Schema Snowflake
Infer Schema Snowflake
Snowpark Automation
Csv File Infer Schema

--

--

Data Engineering Simplified
Data Engineering Simplified

Written by Data Engineering Simplified

1.1K Followers
5 Following

Passionate About Data Engineering, Snowflake Cloud Data Warehouse & Cloud Native Data Services

Responses (1)

Help

Status

About

Careers

Press

Blog

Privacy

Terms

Text to speech

Teams