From SQL to NoSQL -- Changing Your Mindset

Lauren Schaefer, Developer Advocate, MongoDB
From SQL to NoSQL—Changing Your Mindset
Lauren_Schaefer

Parks and Recreation, Season 6, Episode 14

@Lauren_Schaefer
Developer Advocate, MongoDB

From SQL to NoSQL
Changing Your Mindset

The mental journey from SQL to MongoDB
1. Map terms & concepts from SQL to MongoDB
2. Discover the 4 humongous advantages of MongoDB
3. Change your mindset in 3 key ways

MongoDB stores data in documents

MongoDB stores data in documents
{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
profession: ["banking", "finance", "trader"],
cars: [
{
model: "Bentley",
year: 1973
},
{
model: "Rolls Royce",
year: 1965
}
]
}

Modeling data in MongoDB vs SQL
{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}

{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}
ID first_name surname cell city location_x location_y
1 Paul Miller 447557505611 London 45.123 47.232
Users

{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}
Users
ID user_id profession
10 1 banking
11 1 finance
12 1 trader
Professions

{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}
Users
10 1 banking
11 1 finance
12 1 trader
Professions
ID user_id model year
20 1 Bentley 1973
21 1 Rolls Royce 1965
Cars

Collections vs Tables
{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}
{
first_name: ”Lauren",
surname: ”Schaefer",
cell: ”1235552222",
city: ”Lancaster",
profession: [”software engineer", ”developer advocate"],
}
{
first_name: ”Sydney",
school: ”Daisy’s Daycare”
}
2 Lauren Schaefer 1235552222 Lancaster NULL NULL
3 Sydney Schaefer NULL Lancaster NULL NULL
UsersUsers

Collections vs Tables
{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}
{
cell: ”1235552222",
}
{
}
UsersUsers
2 Lauren Schaefer 1235552222 Lancaster NULL NULL
3 Sydney Schaefer NULL Lancaster NULL NULL

Don’t panic!
Use schema validation.
Schemaless
database

RowDocument
{
...
a: “b”
...
}
ID a ...
1 b ...
2 ... ...
3 ... ...

Row(s)Document
{
...
a: “b”
...
}
ID a ...
1 b ...
2 ... ...
3 ... ...
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...

ColumnField
ID a ...
1 b ...
2 c ...
3 ... ...
{
...
a: “b”
...
}
{
...
a: “c”
...
}

TableCollection
{
...
}
... ... ...
... ... ...
... ... ...
... ... ...
{
...
}
{
...
}

DatabaseDatabase
... ... ...
... ... ...
... ... ...
... ... ...
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...

IndexIndex
{
...
}
{
...
}
{
...
}
{
...
}
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...

ViewView
{
...
}
... ... ...
... ... ...
... ... ...
... ... ...
{
...
}
{
...
}

JoinEmbedding
{
...
a: “b”,
...
c: {
d: “e”
...
},
...
}
ID a ...
1 b ...
2 ... ...
3 ... ...
... d ...
1 e ...
... ... ...

JoinDatabase References
ID ... ...
1 ... ...
2 ... ...
3 ... ...
... ... ...
1 ... ...
... ... ...
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}

Left Outer Join$lookup
(Aggregation Pipeline)
ID ... ...
1 ... ...
2 ... ...
3 ... ...
... ... ...
1 ... ...
4 ... ...
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}

Recursive Common Table
Expressions
$graphLookup
(Aggregation Pipeline)
{
...
}
... ... ...
... ... ...
... ... ...
... ... ...
{
...
}
{
...
}

Multi-Record ACID
Transaction
Multi-Document ACID
Transaction
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...

Term mapping summary
Row Column Table Database Index Join Join Left Outer
Join
Recursive
Common Table
Expressions
View Transaction
Document Field Collection Database Index Embedding
Database
References $lookup $graphLookup View Transaction

Scale cheaper!
As the size of your database grows, scale horizontally.

Program faster!
Documents map to data structures in most popular
languages.
Update Your Profile

Program faster!
{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
}
Users
10 1 banking
11 1 finance
12 1 trader
Professions

import pymongo
from pymongo import MongoClient
# CONNECT TO THE DB
client = MongoClient()
client =
pymongo.MongoClient("mongodb+srv://
username:password@cluster0nsdia.
mongodb.net/test?retryWrites=true")
db = client.fabapp
# THE ID OF THE USER WHOSE PROFILE WE WILL
BE RETRIEVING AND UPDATING
userId = 1
Program faster!
import mysql.connector
# CONNECT TO THE DB
mydb = mysql.connector.connect(
host="localhost",
user=”username",
passwd=”password",
database=”fabapp”
)
mycursor = mydb.cursor()
# THE ID OF THE USER WHOSE PROFILE WE WILL
BE RETRIEVING AND UPDATING
userId = 1

# GET THE USER'S PROFILE INFORMATION
### Pull the info from the Users table & put
it in the user dictionary
sql = "Select * FROM Users WHERE Users.ID=%s”
values = (userId,)
mycursor.execute(sql, values)
result = mycursor.fetchone()
user = {
"first_name": result[1],
"surname": result[2],
"cell": result[3],
"city": result[4],
"location_x": result[5],
"location_y": result[6]
}
# GET THE USER'S PROFILE INFORMATION
## We can pull all of the info from the same
document since we used embedding
user = db['Users'].find_one({"_id": userId})
Program faster!
### Pull the info from the Professions table
& put it in the user dictionary
sql = "Select * FROM Professions WHERE
Professions.user_id=%s”
values = (userId,)
results = mycursor.fetchall()
professions = []
for result in results:
professions.append(result[2])
user["professions"] = professions

# UPDATE THE USER DICTIONARY BASED ON USER
INPUT IN THE APP
### We'll update the user dictionary
manually for simplicity
user = {
"first_name": "NewFirst",
"surname": "NewSurname",
"cell": "123-456-7890",
"city": "NewCity",
"location": [40.762, -73.979],
"professions": [”Manager", "Engineer"]
}
Program faster!
# UPDATE THE USER DICTIONARY BASED ON USER
INPUT IN THE APP
### We'll update the user dictionary
manually for simplicity
user = {
"first_name": "NewFirst",
"surname": "NewSurname",
"cell": "123-456-7890",
"city": "NewCity",
"location_x": 40.762,
"location_y": 73.979,
"professions": [”Manager", "Engineer"]
}

# UPDATE THE USER'S PROFILE IN THE DATABASE
### First update what is stored in the Users
table
sql = "UPDATE Users SET first_name=%s,
surname=%s, cell=%s, city=%s, location_x=%s,
location_y=%s WHERE (ID=%s)"
values = (
user["first_name"],
user["surname"],
user["cell"],
user["city"],
user["location_x"],
user["location_y"],
userId)
mydb.commit()
### Since the user's data is stored in a
single document, we only have to make one
update
result = db['Users'].update_one(
{"_id": userId}, {"$set": user})
Program faster!
### Delete existing records in Professions
table and add new ones
sql = "DELETE FROM Professions WHERE
user_id=%s”
values = (userId,)
mydb.commit()
if(len(user["professions"]) > 0):
sql = "INSERT INTO Professions
(user_id,profession) VALUES (%s, %s)”
values = []
for profession in user["professions"]:
values.append((userId, profession))
mycursor.executemany(sql,values)
mydb.commit()

30 lines of code
Program faster!
73 lines of code

SQLMongoDB
Query faster!
Stop doing expensive joins to get your data.
ID a ...
1 b ...
2 ... ...
3 ... ...
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...
... ... ...
{
...
a: “b”,
...
c: {
d: “e”
...
},
...
}

Pivot easier!
Easily change the shape of your data as your app evolves.
SQLMongoDB
{
a: “b”,
c: “one”,
e: “f”
}
{
a: “b”,
c: 1,
new: “no biggee”
}
ALTER TABLE `mydb`.`letters_table`
DROP COLUMN `e`,
ADD COLUMN `New` VARCHAR(45) NULL AFTER `C`,
CHANGE COLUMN `C` `C` INT NULL DEFAULT NULL ;

Not all documents in a collection need to have
the same fields.

the same fields.
The Polymorphic Pattern
{
first_name: "Paul",
surname: "Miller",
cell: "447557505611",
city: "London",
location: [45.123,47.232],
cars: [
{
model: "Bentley",
year: 1973
},
{
year: 1965
}
]
}
{
cell: ”1235552222",
}
{
}

the same fields.
The Outlier Pattern
{
_id: ”Lauren_Schaefer",
displayName: ”Lauren Schaefer",
numFollowers: 1310
followers: [
“naomi_pen”,
“kenwalger”,
“mylynn”
...
]
}
{
_id: ”Nick_Offerman",
displayName: ”Nick Offerman",
numFollowers: 1730332
followers: [
“c_hotaling”,
“IAmJerdog”,
“ChloeCondon”
...
],
has_extras: true
}
{
_id: ”Nick_Offerman_1",
twitter_id: “Nick_Offerman”,
is_overflow: true,
followers: [
“StephenAtHome”,
“TheEllenShow”,
“hulu”
...
]
}

Data that is accessed together should be stored
together.

together.
0
20000
40000
60000
80000
100000
120000
1985 2017
Storage vs Developer Costs
Storage Cost per GB Developer Salary

together.
Don’t normalize your data for the sake of normalizing it.
{
a: “b”,
c: {
d: “e”
...
},
f: [“g”, “h”, “i”],
j: [
{
k: “l”
},
{
m: “n”
}
]
}

Stop using transactions regularly.
Relying on transactions is a bad design smell.
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}
{
...
}

Row Column Table Database Index Join Join Left Outer
Join
Recursive
Common Table
Expressions
View Transaction
Document Field Collection Database Index Embedding
Database
References $lookup $graphLookup View Transaction

1. Scale cheaper

1. Scale cheaper
2. Program faster
### First update what is stored in the Users table
sql = "UPDATE Users SET first_name=%s, surname=%s, cell=%s,
city=%s, location_x=%s, location_y=%s WHERE (ID=%s)"
values = (
user["first_name"],
user["surname"],
user["cell"],
user["city"],
user["location_x"],
user["location_y"],
userId)
mydb.commit()
### Since the user's data is stored in a single document, we only
have to make one update
result = db['Users'].update_one(
{"_id": userId}, {"$set": user})

1. Scale cheaper
2. Program faster
3. Query faster

1. Scale cheaper
2. Program faster
3. Query faster
4. Pivot easier

1. Not all documents in a collection need to have the same
fields.

fields.
2. Data that is accessed together should be stored together.

fields.
2. Data that is accessed together should be stored together.
3. Stop using transactions regularly.

Don’t be Ron Swanson
(in this particular case)

Change your mindset &
get the full value of MongoDB

Suggested Sessions
Today @11a Does Remote Work Really Work? by Me
Today @2p
A Complete Methodology to Data Modeling for MongoDB by
Daniel Coupal
Today @3:15p
Look, Ma, No Servers! Serverless Application Development
with MongoDB Stitch by me
Wednesday @10a-3p Schema Design Patterns Pod @ Builder’s Fest

Additional resources on data modeling patterns
• Advanced Schema Design Patterns (webinar)
• Building with Patterns: A Summary (blog series)
• M320: Data Modeling (MongoDB University Course -- coming
soon!)

Additional resources
• The MongoDB Docs
• JSON Schema Validation – Locking down your model the
smart way
• JSON Schema Validation - Checking Your Arrays
• M121: The MongoDB Aggregation Framework

Change your mindset &
get the full value of MongoDB
Get the slides on my
Twitter page:
@Lauren_Schaefer

From SQL to NoSQL -- Changing Your Mindset

More Related Content

Similar to From SQL to NoSQL -- Changing Your Mindset (20)

More from Lauren Hayward Schaefer (20)

Recently uploaded (20)

From SQL to NoSQL -- Changing Your Mindset