Querying Neo4j Aura from R with neo2R
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.
1 Introduction
Graph databases excel at storing and traversing highly connected data used for recommendation engines, fraud detection, knowledge graphs, and social networks. Neo4j is one of the most widely used graph databases, and with Neo4j Aura, its managed cloud service now makes it easy to spin up a production-grade instance without any infrastructure overhead.
On the R side, the neo2R package has long been available for querying self-hosted Neo4j instances from R. Version 3.0.0 brings two important changes:
- Unified connection model — a single
startGraph()call handles both a self-hosted Neo4j instance (http://localhost:7474) and a cloud Neo4j Aura instance (https://<id>.databases.neo4j.io). - httr2 backend — the internal HTTP layer migrated from the deprecated
httrpackage tohttr2.
In this post, we’ll connect to the free Neo4j Aura demo database preloaded with the classic Movie Recommendations dataset, explore the graph with Cypher queries, and finish with an interactive network visualization built with visNetwork.
2 Prerequisites
install.packages(c("neo2R", "dplyr", "visNetwork"))
library(neo2R) library(dplyr)
Attaching package: 'dplyr'
The following objects are masked from 'package:stats':
filter, lag
The following objects are masked from 'package:base':
intersect, setdiff, setequal, union
library(visNetwork)
neo2R 3.0.0 requires R ≥ 4.1 and httr2 ≥ 1.0.0. Check your versions with packageVersion("neo2R") and packageVersion("httr2").
3 Connecting to Neo4j Aura
3.1 Create and Connect to an Aura Instance
Neo4j provides a free Aura Free tier (up to 200 k nodes / 400 k relationships).
Create a free instance at https://console.neo4j.io and get your connection details.
Connect to your instance with startGraph().
my_aura <- startGraph( url = "https://<INSTANCEID>.databases.neo4j.io", database = "INSTANCEID", username = "INSTANCEID", password = "INSTANCEPASSWORD" ## api = "v2" is set automatically for *.databases.neo4j.io URLs )
3.2 The Movie Recommendations Dataset
Neo4j provides example datasets, and most of them are available as a one-click templates in the Neo4j Aura console.
The Movie Recommendations dataset is a graph example using a dataset of movie reviews for generating personalized, real-time recommendations. This dataset is also available on a demo server that can be accessed as follows.
graph <- startGraph( url = "https://demo.neo4jlabs.com:7473", database = "recommendations", username = "recommendations", password = "recommendations" )
4 Exploring the schema
The Movie database contains the following node labels and relationship types:
| Node label | Key properties |
|---|---|
| Movie | title, released, imdbId |
| Genre | name |
| Actor | name, born, imdbId |
| Director | name, born, imdbId |
| User | name |
| Relationship type | Key properties |
|---|---|
| IN_GENRE | |
| ACTED_IN | role |
| DIRECTED | |
| RATED | rating || timestamp |
Let’s count the number of these different concepts:
## Node types and counts
cypher(
graph,
"
MATCH (n)
RETURN labels(n) AS label, count(n) AS n
ORDER BY n DESC
"
) |>
as_tibble() |>
## filtering out technical nodes
filter(label %notin% c("_Bloom_Perspective_", "_Bloom_Scene_", ""))
# A tibble: 6 × 2 label n <chr> <int> 1 Actor || Person 14956 2 Movie 9125 3 Director || Person 3604 4 User 671 5 Actor || Director || Person 487 6 Genre 20
## Relationship types and counts
cypher(
graph,
"
MATCH ()-[r]->()
RETURN type(r) AS type, count(r) AS n
ORDER BY n DESC
"
) |>
as_tibble() |>
## filtering out technical relationships
filter(type %notin% c("_Bloom_HAS_SCENE_"))
# A tibble: 4 × 2 type n <chr> <int> 1 RATED 100004 2 ACTED_IN 35910 3 IN_GENRE 20340 4 DIRECTED 10007
5 Querying with Cypher
5.1 Top prolific actors
cypher( graph, " MATCH (p:Person)-[:ACTED_IN]->(m:Movie) RETURN p.name AS actor, count(m) AS movies ORDER BY movies DESC LIMIT 10 " ) |> as_tibble()
# A tibble: 10 × 2 actor movies <chr> <int> 1 Robert De Niro 56 2 Bruce Willis 49 3 Samuel L. Jackson 45 4 Nicolas Cage 45 5 Michael Caine 40 6 Clint Eastwood 40 7 Tom Hanks 38 8 John Cusack 38 9 Morgan Freeman 38 10 Gene Hackman 38
5.2 Movies and their directors
cypher( graph, " MATCH (d:Person)-[:DIRECTED]->(m:Movie) RETURN m.title AS movie, m.released as released, d.name AS director ORDER BY m.released IS NOT NULL DESC, m.released DESC LIMIT 10 " ) |> as_tibble()
# A tibble: 10 × 3 movie released director <chr> <chr> <chr> 1 Solace 2016-09-02 "Afonso Poyart" 2 Ben-hur 2016-08-12 "Timur Bekmambetov" 3 Rustom 2016-08-12 "Tinu Suresh Desai" 4 Mohenjo Daro 2016-08-12 "Ashutosh Gowariker" 5 Suicide Squad 2016-08-05 "David Ayer" 6 Shin Godzilla 2016-07-29 "Hideaki Anno" 7 Shin Godzilla 2016-07-29 " Shinji Higuchi" 8 Jason Bourne 2016-07-29 "Paul Greengrass" 9 Star Trek 3 2016-07-22 "Justin Lin" 10 Ghostbusters 2016-07-15 "Paul Feig"
5.3 Parameterised queries
neo2R supports named parameters, keeping queries safe from injection and easy to reuse:
## Find all co-stars of a given actor
cypher(
graph,
"
MATCH (a:Person {name: $actor})-[:ACTED_IN]->(m:Movie)<-[:ACTED_IN]-(co:Person)
RETURN DISTINCT co.name AS co_star, m.title AS shared_movie
ORDER BY co_star
",
parameters = list(actor = "Tom Hanks")
) |>
as_tibble()
# A tibble: 114 × 2 co_star shared_movie <chr> <chr> 1 Adrian Zmed Bachelor Party 2 Alexander Godunov Money Pit, The 3 Amy Adams Charlie Wilson's War 4 Annie Rose Buckley Saving Mr. Banks 5 Audrey Tautou Da Vinci Code, The 6 Ayelet Zurer Angels & Demons 7 Barkhad Abdi Captain Phillips 8 Barkhad Abdirahman Captain Phillips 9 Barry Pepper Saving Private Ryan 10 Bill Paxton Apollo 13 # ℹ 104 more rows
6 Network visualisation with visNetwork
The real power of a graph database is visible when you draw the graph. Let’s pull Tom Hanks’s ego network, everyone he has acted alongside, and render it with visNetwork.
6.1 Step 1 — Fetch nodes and edges
## Tom Hanks, his movies, and his co-stars
hub <- "Tom Hanks"
nodes_raw <- cypher(
graph,
"
MATCH (hub:Person {name: $hub})-[hr:ACTED_IN]->(m:Movie)
<-[cr:ACTED_IN]-(co:Person)
RETURN hub.name AS hub, hr.role AS hub_role,
m.title AS movie, m.year AS year,
co.name AS co, cr.role AS co_role
",
parameters = list(hub = hub)
) |>
as_tibble()
6.2 Step 2 — Shape data for visNetwork
visNetwork expects two data frames: nodes (with columns id, label, group, …) and edges (with columns from, to, …).
nodes <- bind_rows(
nodes_raw |>
distinct(
id = hub,
group = "Hub"
),
nodes_raw |>
distinct(
id = co,
group = "Co-star"
),
nodes_raw |>
distinct(
id = movie,
group = "Movie",
year
)
) |>
distinct() |>
mutate(
title = sprintf(
'<b>%s</b>: %s%s',
group,
id,
ifelse(!is.na(year), sprintf("(%s)", year), "")
),
shape = ifelse(group == "Movie", "dot", "star"),
size = ifelse(group == "Hub", 30, 18)
) |>
arrange(id)
edges <- bind_rows(
nodes_raw |>
distinct(
from = hub,
to = movie,
role = hub_role
),
nodes_raw |>
distinct(
from = co,
to = movie,
role = co_role
)
) |>
mutate(
title = sprintf('<b>Role</b>: %s', role),
arrows = "to"
)
6.3 Step 3 — Draw the network
visNetwork(nodes, edges) |>
visGroups(
groupname = "hub",
color = list(
background = "#3B82F6",
border = "#1D4ED8",
highlight = "#93C5FD"
)
) |>
visGroups(
groupname = "movie",
color = list(
background = "#F97316",
border = "#C2410C",
highlight = "#FED7AA"
),
shape = "square"
) |>
visGroups(
groupname = "costar",
color = list(
background = "#6B7280",
border = "#374151",
highlight = "#D1D5DB"
)
) |>
visEdges(
color = list(color = "#CBD5E1", highlight = "#3B82F6"),
width = 1.5
) |>
visOptions(
highlightNearest = list(enabled = TRUE, degree = 1, hover = TRUE),
nodesIdSelection = TRUE
) |>
visLayout(randomSeed = 42) |>
visPhysics(
solver = "forceAtlas2Based",
forceAtlas2Based = list(
gravitationalConstant = -60,
springLength = 120,
springConstant = 0.04
)
) |>
visLegend(position = "right", main = "Node type")
Hover over any node to see its label. Use the Select by id dropdown or a node to highlight movies shared with Tom Hanks.
7 Conclusion
neo2R 3.0.0 removes the last friction point for R users who want to work with Neo4j Aura: a single startGraph() call now handles cloud and local instances uniformly, the httr2 backend gives reliable retries and clean error handling, and the Cypher query interface remains exactly as it was.
7.1 Further reading
- neo2R on CRAN — package documentation
- neo2R GitHub — source, changelog, and issues
- Neo4j Aura console — create your free instance
- Neo4j Cypher reference — query language docs
- visNetwork documentation — all chart options
R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.