Neo4j GraphDay Seattle- Sept19- neo4j basic training

Welcome everyone

Please say hi to each other.
We'll spend the rest of the day together.

What are we going to learn today?
•  Querying graph pa;erns with Cypher,
•  Designing and implemen@ng a graph database model
•  Evolving an exis@ng graph to support new or changed requirements.

•  Intro to graphs and graph modeling
•  Evolving the graph model
•  Missing domain concept
•  Rela@onship bundling
•  Modeling guidelines
What are we going to do today?

Op:onal Modules
•  Further evolving the graph model
•  Speciﬁc rela@onship types
•  Refactoring large graphs
•  Mul@ple models
•  Your turn

Thank you for your @me

Any other ques@ons?

Feedback form: bit.ly/neo-survey gives a discount

•  Intro to the property graph model
•  The airport dataset
•  The modeling workﬂow
•  Load CSV data
What are we going to do?

The labelled property graph
•  Nodes
•  Rela@onships
•  Proper@es
•  Labels
CAR
name: “Dan”
born: May 29, 1970
twi;er: “@dan”
name: “Ann”
born: Dec 5, 1975
since:
Jan 10, 2011
brand: “Volvo”
model: “V70”
PERSON PERSON
LOVES
LOVES
LIVES WITH

CAR
Property Graph Model Components
Nodes
•  Represent the objects in the graph
•  Can be labeled
PERSON PERSON

CAR
Nodes
• Represent the objects in the
graph
• Can be labeled
Rela:onships
•  Relate nodes by type and direc-on
LOVES
LOVES
LIVES WITH
PERSON PERSON

CAR
name: “Dan”
born: May 29, 1970
twi;er: “@dan”
name: “Ann”
born: Dec 5, 1975
since:
Jan 10, 2011
brand: “Volvo”
model: “V70”
Nodes
• Represent the objects in the
graph
•  Can be labeled
Rela:onships
• Relate nodes by type and
direc-on
Proper:es
•  Name-value pairs that can go on
nodes and rela@onships.
LOVES
LOVES
LIVES WITH
PERSON PERSON

The modeling workﬂow
1. Derive the
question
2. Obtain the
data
3. Develop a
model
4. Ingest the
data
5. Query/Prove
our model

Find the most popular airports
As an air travel enthusiast I want to know how airports are connected
So that I can ﬁnd the busiest ones

Derive ques:ons
Is Airport A connected to Airport B?

Iden:ty en::es

airport

Iden:fy rela:onships between en::es

airport CONNECTED_TO airport

It’s a graph!

(airport)-[:CONNECTED_TO]->(airport)

Labels
Labels are for categorizing nodes.

Proper:es
Proper@es are for deﬁning a;ributes of nodes or rela@onships.
We need an airport code to uniquely iden@fy each airport.

Modeling diﬀerent connec:ons
Busy airports will have mul@ple connec@ons between them. We can
model this in diﬀerent ways:


Copy folders
• import
• plugins
from USB S:ck
to the default.graphdb folder
(or $NEO4J_HOME)

bit.ly/neo4j-modeling-data
Prepare your Neo4j graph.db directory

1.  Start the server.
2.  It should be running on: http://localhost:7474
3.  Log-in with default creden@als
user: neo4j password: neo4j
4.  Choose a new password
We’re good to go!

Make sure you’ve got Neo4j running

Find the most popular airports
Open your browser to http://localhost:7474 and execute the following
command:
:play https://guides.neo4j.com/modeling_airports

Play the guides in your browser un:l you see...

MERGE
The MERGE command is a combina@on of MATCH and CREATE. If the pa;ern
already exists then it will return it; if not it will create it.
en.wikipedia.org/wiki/Idempotence

MERGE (las:Airport {code: "LAS"})
RETURN las

We can MERGE nodes...

...or rela:onships...
MATCH (las:Airport {code: "LAS"})
MATCH (lax:Airport {code: "LAX"})
MERGE (las)-[:CONNECTED_TO]->(lax)

...or both
MERGE (las:Airport {code: "LAS"})
MERGE (lax:Airport {code: "LAX"})
MERGE (las)-[:CONNECTED_TO]->(lax)

MERGE (las:Airport {code: "LAS"})-[:CONNECTED_TO]->(lax:Airport
{code: "LAX"})

Con:nue playing the guide
Con@nue the guide in your browser

As an air travel enthusiast
I want to know how airports are connected So that I can ﬁnd the busiest ones
The modeling workﬂow
MATCH (origin:Airport {code: "LAX"})
-[flight:CONNECTED_TO]->
(destination:Airport {code: "LAS"})
RETURN origin, destination, flight
LIMIT 10

CREATE (lax:Airport {code: "LAX"})
CREATE (las:Airport {code: "LAS"})
CREATE (las)-[connection:CONNECTED_TO {
airline: "WN",
flightNumber: "82",
date: "2008-1-3",
departure: 1715,
arrival: 1820}]->(lax)

|---------+------+------------|
| Origin | Dest | FlightNum |
|---------+------+------------|
| IAD | TPA | 335 |
| IAD | TPA | 3231 |
| IND | BWI | 448 |
| IND | BWI | 3920 |
|---------+------+------------|

(airport)-[:CONNECTED_TO]->(airport)

Modeling airports

Nodes

Labels

Rela:onships

Proper:es

Load CSV
A clause in the Cypher query language that lets us iterate over CSV ﬁles
and create graph structures based on the data contained in each row.

Load CSV
LOAD CSV // load csv data
WITH HEADERS // optionally use first header row as keys in "row" map
FROM "url" // file:// URL relative to $NEO4J_HOME/import or http://
AS row // return each row of the CSV as list of strings or map
// ... rest of the Cypher statement ...

Load CSV
[USING PERIODIC COMMIT] // optionally batch transactions
LOAD CSV // load csv data
WITH HEADERS // optionally use first header row as keys in "row" map
FROM "url" // file:// URL relative to $NEO4J_HOME/import or http://
AS row // return each row of the CSV as list of strings or map
[FIELDTERMINATOR ";"] // optionally alt. delimiter
// ... rest of the Cypher statement ...

Consider this CSV ﬁle
|---------+------+------------|
|---------+------+------------|
| IAD | TPA | 335 |
| IAD | TPA | 3231 |
| IND | BWI | 448 |
| IND | BWI | 3920 |
|---------+------+------------|

Create nodes
LOAD CSV WITH HEADERS FROM "file:///flights.csv" AS row
CREATE (:Airport {code: row.Origin})
CREATE (:Airport {code: row.Dest})

|---------+------+------------|
|---------+------+------------|
| IAD | TPA | 335 |
| IAD | TPA | 3231 |
| IND | BWI | 448 |
| IND | BWI | 3920 |
|---------+------+------------|

Create rela:onships
CREATE (origin:Airport {code: row.Origin})
CREATE (destination:Airport {code: row.Dest})
CREATE (origin)-[:CONNECTED_TO {flightNumber: row.FlightNum}]->(destination)

|---------+------+------------|
|---------+------+------------|
| IAD | TPA | 335 |
| IAD | TPA | 3231 |
| IND | BWI | 448 |
| IND | BWI | 3920 |
|---------+------+------------|

Find exis:ng nodes and rela:onships
MATCH (origin:Airport {code: row.Origin})
MATCH (destination:Airport {code: row.Dest})
MATCH (origin)-[:CONNECTED_TO {flightNumber: row.FlightNum}]->(destination)
...
|---------+------+------------|
|---------+------+------------|
| IAD | TPA | 335 |
| IAD | TPA | 3231 |
| IND | BWI | 448 |
| IND | BWI | 3920 |
|---------+------+------------|

Update exis:ng nodes and rela:onships
|---------+------+-----------+----------------|
| Origin | Dest | FlightNum | UniqueCarrier |
|---------+------+-----------+----------------|
| IAD | TPA | 335 | WN |
| IAD | TPA | 3231 | WN |
| IND | BWI | 448 | WN |
| IND | BWI | 3920 | WN |
|---------+------+-----------+----------------|

MATCH (origin:Airport {code: row.Origin})
MATCH (destination:Airport {code: row.Dest})
MATCH (origin)-[c:CONNECTED_TO {flightNumber:row.FlightNum}]->(destination)
SET c.airline = row.UniqueCarrier

Idempotently create nodes and rela:onships
MERGE (origin:Airport {code: row.Origin})
MERGE (destination:Airport {code: row.Dest})
MERGE (origin)-[:CONNECTED_TO {flightNumber: row.FlightNum}]-
>(destination)

|---------+------+------------|
|---------+------+------------|
| IAD | TPA | 335 |
| IAD | TPA | 3231 |
| IND | BWI | 448 |
| IND | BWI | 3920 |
|---------+------+------------|

Let’s give it a try
Con@nue the guide in your browser

End of Module
Intro to Graph Modeling
Ques:ons ?

•  Proﬁle queries
•  Introduce a missing concept in the domain
•  Constraints and indexes
•  Write our ﬁrst refactoring query

Start playing the next guide….
...if you aren’t playing it already

:play http://guides.neo4j.com/modeling_airports/02_flight.html

How do I proﬁle a query?
By preﬁxing the query with:
EXPLAIN
shows the execu@on plan without actually execu@ng it or returning any results.

By preﬁxing the query with:
EXPLAIN
shows the execu@on plan without actually execu@ng it or returning any results.
PROFILE
executes the statement and returns the results along with proﬁling informa@on.

EXPLAIN
MATCH (origin:Airport)-[c:CONNECTED_TO]
->(destination:Airport)
WHERE destination.code = "LAS"
RETURN origin, destination, c
LIMIT 10

PROFILE
LIMIT 10

PROFILE
LIMIT 10

Anatomy of an execu:on plan

Opera:ons
Rows come in
Do some work
Rows go out

What’s our goal?
At a high level, the goal is
simple: get the number of db
hits down.

What is a database hit?
an abstract unit of storage
engine work.
“
”

Let’s proﬁle some queries
Con@nue playing the guide in your browser

Introducing a missing
domain concept

Introducing a missing domain concept
Some@mes we design a model and amer working with it for a bit we
realise that we’ve missed an important domain concept.

Some@mes we design a model and amer working with it for a bit we
realise that we’ve missed an important domain concept.

In our case we’re missing nodes to represent Flights

Flight as a ﬁrst class ci:zen
???

Flight as a ﬁrst class ci:zen
MATCH (o:Airport)-[:CONNECTED_TO]->(d:Airport)
MERGE...

Unique Constraints
We create unique constraints to:
•  ensure uniqueness
•  allow fast lookup of nodes which match label-property pairs.

Unique Constraints
We create unique constraints to:
•  ensure uniqueness

CREATE CONSTRAINT ON (identifier:Label)
ASSERT identifier.property IS UNIQUE

This is the ﬁrst of the refactoring pa;erns that we’ll encounter today.
We have this pa;ern:

(origin)-[:CONNECTED_TO]->(destination)
And we’ll create a new node using the proper@es of the CONNECTED_TO
rela@onship:

(origin)<-[:ORIGIN]-(flight)-[:DESTINATION]->(destination)

Refactoring: Derive node from rela:onship

Indexes
We create indexes to:

Indexes
We create indexes to:

CREATE INDEX ON :Label(property)

What gets indexed?
The following are index backed:
•  Equality
•  STARTS WITH
•  CONTAINS
•  ENDS WITH
•  Range searches
•  (Non-) existence checks

How are indexes used in neo4j?
Indexes are only used to ﬁnd the star@ng point for queries.
Use index scans to look up rows in
tables and join them with rows
from other tables
Use indexes to ﬁnd the star@ng
points for a query.
Relational
Graph

Let’s get on with the refactoring
Con@nue playing the guide in your browser

End of Module
Missing Concept
Ques:ons?

Rela:onship Bundling
So far we’ve been wri@ng queries from the perspec@ve of an air travel
enthusiast.
In this sec@on we’re going to evolve the model based on the needs of a
frequent traveller who wants to find flights on a specific date.

•  Write a query to book a flight on a specific day
•  Profile our flight booking query
•  Op@mise our model to deal with this query

Find a flight to book
As a frequent traveller I want to find flights from <origin> to
<des@na@on> on <date> So that I can book my business flight

Start playing the next guide….
...if you aren’t playing it already

:play http://guides.neo4j.com/modeling_airports/03_flight_booking.html

Bundling flight rela:onships by airport day
Searching for flights by day is a bit problema@c. We have to scan every
rela@onship between origin and des@na@on airports and analyse
proper@es on each flight to work out which ones we’re interested in.

Searching for flights by day is a bit problema@c. We have to scan every
rela@onship between origin and des@na@on airports and analyse
proper@es on each flight to work out which ones we’re interested in.
We can reduce this problem by bundling flights together by
AirportDay

Refactoring to Airport Day
???

Refactoring: Derive node from property
This refactoring creates a new node using the property of an exis@ng node.
We currently have this pa;ern:

And we’ll create a new node using the date property of the flight node:

(origin)-[:HAS_DAY]->(originAirportDay)<-[:ORIGIN]-(flight),
(destination)-[:HAS_DAY]->(destAirportDay)<-[:DESTINATION]-(flight)

Let’s get on with the refactoring
Over to you!

End of Module
Bundling Rela:onships
Ques:ons?

•  Recap of the Property Graph Model
•  Modeling choices
•  Refactorings

Nodes, rela:onships,
labels, proper:es

We only need to pull out a node if we’re going to query through it,
otherwise a property will suﬃce.

But if we pull out every single property then we end up with an RDF
model and lose the beneﬁt of the property graph
Proper:es vs Rela:onships

No need to have the rela:onship in both direc:ons

Use single rela:onship and ignore direc:on in queries
MATCH (:Person {name:'Eric'})-[:MARRIED_TO]-(p2)
RETURN p2

General vs. Specific
Relationships

General: Find ﬂights on a speciﬁc date
MATCH (origin:Airport {code: "LAS"})-[:HAS_DAY]->(originDay:AirportDay),
(originDay)<-[:ORIGIN]-(flight:Flight),
(flight)-[:DESTINATION]->(destinationDay),
(destinationDay:AirportDay)<-[:HAS_DAY]-(destination:Airport {code: "MDW"})

WHERE originDay.date = "2008-1-3" AND destinationDay.date = "2008-1-3"

RETURN flight.date, flight.number, flight.airline, flight.departure, flight.arrival

ORDER BY flight.date, flight.departure

Specific: Find flights on a specific date
MATCH (origin:Airport {code: "LAS"})-[:`2008-1-3`]->(originDay:AirportDay),
(originDay)<-[:ORIGIN]-(flight:Flight),
(flight)-[:DESTINATION]->(destinationDay),
(destinationDay:AirportDay)<-[:`2008-1-3`]-(destination:Airport {code: "MDW"})


ORDER BY flight.date, flight.departure;

General: Find ﬂights by year and month
MATCH (origin:Airport {code: "LAS"})-[:HAS_DAY]->(originDay:AirportDay),
(originDay)<-[:ORIGIN]-(flight:Flight)

WHERE originDay.date STARTS WITH "2008-1"



Speciﬁc: Find ﬂights by year and month
MATCH (origin:Airport {code: "LAS"})
-[:`2008-1-3`|:`2008-1-4`|:`2008-1-5`|:`2008-1-6`]->(originDay:AirportDay),
(originDay)<-[:ORIGIN]-(flight:Flight)



Derive node from rela:onship
(origin)-[:CONNECTED_TO]->(destination)


Derive node from property

(origin)-[:HAS_DAY]->(originAirportDay)<-[:ORIGIN]-(flight),
(destination)-[:HAS_DAY]->(destAirportDay)<-[:DESTINATION]-(flight)

Derive rela:onship from property
(airport)-[:HAS_DAY]->(airportDay {date: "2008-1-3"})

(airport)-[:`2008-1-3`]->(airportDay {date: "2008-1-3"})

End of Module
Modeling Guidelines
Ques:ons ?

Neo4j GraphDay Seattle- Sept19- neo4j basic training

More Related Content

What's hot (20)

Similar to Neo4j GraphDay Seattle- Sept19- neo4j basic training (20)

More from Neo4j (20)

Recently uploaded (20)

Neo4j GraphDay Seattle- Sept19- neo4j basic training