prrao87 / kuzudb-study Goto Github PK
View Code? Open in Web Editor NEWBenchmark study on KùzuDB, an embedded OLAP graph database, on an artificial social network dataset
License: MIT License
Benchmark study on KùzuDB, an embedded OLAP graph database, on an artificial social network dataset
License: MIT License
I'm trying to run the following query that involves taking the result from one query, and passing it to another query in Kuzu, and it segfaults.
# In which city does the most-followed person in the network live?
query = """
MATCH (follower:Person)-[:Follows]->(person:Person)
WITH person, count(follower) as followers
ORDER BY followers DESC LIMIT 1
MATCH (person) -[:LivesIn]-> (city:City)
RETURN person.name AS name, followers AS numFollowers, city.city AS city
"""
Output:
[1] 15084 segmentation fault python query.py
The same query runs in Neo4j. Not sure about the best way to run this query in Kuzu to obtain a valid result.
Can you try to run Neo4j with indexes?
For example Country.country, Person.age, City.city
Create two properties Person.lower_gender and Interest.lower_interest with the property already in lowercase and index both... or just set the data to be the toLower version of it and index/use that directly.
Also what does the throughput of these queries look like? Not just the time.
Also comparing queries that take so little time is not the most useful.
0.0086 seconds vs 0.0046 seconds is kinda pointless.
Queries 9 and 10 run in Neo4j (relatively slowly), but they segfault in Kùzu. Is it something I'm doing wrong? It would help if there were more helpful error messages, but I can imagine there's something going on in the translation of Cypher logic into the C++ layer.
Query 9:
MATCH (:Person)-[r1:FOLLOWS]->(influencer:Person)-[r2:FOLLOWS]->(:Person)
WITH count(r1) AS numFollowers, influencer, r2
WHERE influencer.age <= $age_upper AND numFollowers > 3000
RETURN influencer.id AS influencerId, influencer.name AS name, count(r2) AS numFollows
ORDER BY numFollows DESC LIMIT 5;
Influencers below age 30 who follow the most people:
shape: (5, 3)
┌──────────────┬─────────────────┬────────────┐
│ influencerId ┆ name ┆ numFollows │
│ --- ┆ --- ┆ --- │
│ i64 ┆ str ┆ i64 │
╞══════════════╪═════════════════╪════════════╡
│ 89758 ┆ Joshua Williams ┆ 40 │
│ 85914 ┆ Micheal Holt ┆ 32 │
│ 8077 ┆ Ralph Floyd ┆ 32 │
│ 1348 ┆ Brett Wright ┆ 32 │
│ 70809 ┆ David Cooper ┆ 31 │
└──────────────┴─────────────────┴────────────┘
Query 10:
MATCH (:Person)-[r1:FOLLOWS]->(influencer:Person)-[r2:FOLLOWS]->(person:Person)
WITH count(r1) AS numFollowers, person, influencer, r2
WHERE influencer.age >= $age_lower AND influencer.age <= $age_upper AND numFollowers > 3000
RETURN person.id AS personId, person.name AS name, count(r2) AS numFollowers
ORDER BY numFollowers DESC LIMIT 5;
Influencers below the age of 18-25 who can be considered 'influencers' in the network:
shape: (5, 3)
┌──────────┬────────────────────┬──────────────┐
│ personId ┆ name ┆ numFollowers │
│ --- ┆ --- ┆ --- │
│ i64 ┆ str ┆ i64 │
╞══════════╪════════════════════╪══════════════╡
│ 88759 ┆ Kylie Chang ┆ 5 │
│ 39355 ┆ Elizabeth Hamilton ┆ 4 │
│ 12104 ┆ Cheryl Coleman ┆ 4 │
│ 27567 ┆ Michael Dominguez ┆ 4 │
│ 31072 ┆ Jeanette Nolan ┆ 4 │
└──────────┴────────────────────┴──────────────┘
Neo4j query script completed in 17.753386s
Query 9:
MATCH (:Person)-[r1:Follows]->(influencer:Person)-[r2:Follows]->(:Person)
WITH count(r1) AS numFollowers, influencer, r2
WHERE influencer.age <= $age_upper AND numFollowers > 3000
RETURN influencer.id AS influencerId, influencer.name AS name, count(r2) AS numFollows
ORDER BY numFollows DESC LIMIT 5;
[1] 15847 bus error python query.py
I'm passing a parameter to the following query in Kuzu and it doesn't seem to work as intended.
# Which are the top 5 cities in a particular country with the lowest average age in the network?
query = """
MATCH (country:Country)
WHERE co.country = $country
AND EXISTS { MATCH (p:Person)-[:LivesIn]->(ci:City)-[*..2]->(country) }
RETURN ci.city AS city, avg(p.age) AS averageAge
ORDER BY averageAge LIMIT 5
"""
Kuzu doesn't seem to like using the WHERE xx AND EXISTS {MATCH yy}
clause, despite this being valid Cypher in Kuzu as per the example notebook.
Output:
Traceback (most recent call last):
File "/Users/prrao/code/kuzudb-study/kuzudb/query.py", line 85, in <module>
main(CONNECTION)
File "/Users/prrao/code/kuzudb-study/kuzudb/query.py", line 76, in main
run_query3(conn, params=[("country", "Canada")])
File "/Users/prrao/code/kuzudb-study/kuzudb/query.py", line 51, in run_query3
response = conn.execute(query, parameters=params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/prrao/.pyenv/versions/3.11.2/lib/python3.11/site-packages/kuzu/connection.py", line 88, in execute
self._connection.execute(
RuntimeError: Parser exception: Invalid input <MATCH (country:Country)
WHERE co.country = $country
AND EXISTS { MATCH (p:Person)-[:LivesIn]->(ci:City)-[*..>: expected rule oC_SingleQuery (line: 4, offset: 62)
" RETURN ci.city AS city, avg(p.age) AS averageAge"
The source of the error isn't clear, which needs a deeper dive. Query 4 also suffers from the same issue.
CSV files are useful for debugging during initial data inspection, but to scale things up, it makes sense to just output all the node/edge generation data to parquet prior to ingesting into a graph.
neo4j
and kuzudb
directories -- the description of all 10 queries is already in the main pageRelated to kuzudb/kuzu#2190.
Kùzu has recently switched from the arrow parquet reader to their own parquet reader to achieve zero-copy during copying parquet files to a database. This affects existing logic in this repo that relies on Polars upstream, leading to the following decompression error when ingesting from parquet:
Traceback (most recent call last):
File "/code/kuzudb-study/kuzudb/build_graph.py", line 133, in <module>
main(CONNECTION)
File "/code/kuzudb-study/kuzudb/build_graph.py", line 106, in main
conn.execute(f"COPY Person FROM '{NODES_PATH}/persons.parquet';")
File "/code/kuzudb-study/.venv/lib/python3.11/site-packages/kuzu/connection.py", line 90, in execute
self._connection.execute(
RuntimeError: ColumnReader::decompressInternal
By default, Polars uses zstd
compression, but Kùzu's custom parquet reader only supports snappy
compression for now. As a result, any parquet files exported from Polars have to be updated to use snappy
compression until Kùzu supports other compression formats for parquet.
Queries 7 and 8, which count the edges between persons and locations do not show consistent results between the two DBs. The other queries showed identical results, so it doesn't make sense why these two queries specifically don't add up.
Query 7:
MATCH (p:Person)-[:LIVES_IN]->(:City)-[:CITY_IN]->(s:State)
WHERE p.age >= $age_lower AND p.age <= $age_upper AND s.country = $country
WITH p, s
MATCH (p)-[:HAS_INTEREST]->(i:Interest)
WHERE tolower(i.interest) = tolower($interest)
RETURN count(p) AS numPersons, s.state AS state, s.country AS country
ORDER BY numPersons DESC LIMIT 1
State in United States with the most users between ages 23-30 who have an interest in photography:
shape: (1, 3)
┌────────────┬────────────┬───────────────┐
│ numPersons ┆ state ┆ country │
│ --- ┆ --- ┆ --- │
│ i64 ┆ str ┆ str │
╞════════════╪════════════╪═══════════════╡
│ 172 ┆ California ┆ United States │
└────────────┴────────────┴───────────────┘
Query 7 completed in 0.163613s
Query 8:
MATCH (p1:Person)-[f:FOLLOWS]->(p2:Person)
WHERE p1.personID > p2.personID
RETURN count(f) as numFollowers
Number of second degree connections reachable in the graph:
shape: (1, 1)
┌──────────────┐
│ numFollowers │
│ --- │
│ i64 │
╞══════════════╡
│ 1219517 │
└──────────────┘
Query 8 completed in 0.990533s
Query 7:
MATCH (p:Person)-[:LivesIn]->(:City)-[:CityIn]->(s:State)
WHERE p.age >= $age_lower AND p.age <= $age_upper AND s.country = $country
WITH p, s
MATCH (p)-[:HasInterest]->(i:Interest)
WHERE lower(i.interest) = lower($interest)
RETURN count(p.id) AS numPersons, s.state AS state, s.country AS country
ORDER BY numPersons DESC LIMIT 1
State in United States with the most users between ages 23-30 who have an interest in photography:
shape: (1, 3)
┌────────────┬────────────┬───────────────┐
│ numPersons ┆ state ┆ country │
│ --- ┆ --- ┆ --- │
│ i64 ┆ str ┆ str │
╞════════════╪════════════╪═══════════════╡
│ 169 ┆ California ┆ United States │
└────────────┴────────────┴───────────────┘
Query 7 completed in 0.010869s
Query 8:
MATCH (p1:Person)-[f:Follows]->(p2:Person)
WHERE p1.id > p2.id
RETURN count(f) as numFollowers
Number of second degree connections reachable in the graph:
shape: (1, 1)
┌──────────────┐
│ numFollowers │
│ --- │
│ i64 │
╞══════════════╡
│ 1214477 │
└──────────────┘
Query 8 completed in 0.027979s
Query | Neo4j (count) | Kùzu (count) |
---|---|---|
7 | 172 | 169 |
8 | 1219517 | 1214477 |
What could be going wrong??
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.