Creating multiple relationships and nodes from 1 file

[Kamalika17]

Hi,

Is it possible to create multiple relationships and nodes from 1 file or is it necessary to have separate files for each node type and edge type for a knowledge graph. eg: I have 1 data frame where all information is given:
Person, City, Country, Hobby
A, Bern, Switzerland, Painting

Each column is a node type and relationships - A "lives-in" Bern, Bern "capital-of" Switzerland, A "likes" Painting

Thanks for your help!

[andyfengHKU]

Hi @Kamalika17,

Yes if you want to load them into different relationship tables then they should be in different files.

However, if your dataset is not too large. It should also work fine if you make lives-in, capital-of, likes as a relationship property (of type string). So effectively you only have one node table and one relationship table. Everything else is modelled as a property on these two tables.

[andyfengHKU]

Yes I guess you could do something around

COPY <table_name> FROM ( LOAD FROM <data_frame> RETURN <columns> )

[prrao87]

Yup, the example snippet you're looking for is here. The link to this section was broken, it's been fixed now :).

Basically you'd do something along these lines:

# Assumes that you have a Kùzu connection object named `conn`
# Also assumes that you created a node table named `Person` with columns `name` and `age`
import pandas as pd

df = pd.DataFrame({
    "name": ["Adam", "Karissa", "Zhang", "Noura"],
    "age": [30, 40, 50, 25]
})

conn.execute("COPY Person FROM (LOAD FROM df WHERE age < 30 RETURN *)")

You can scan the Pandas DataFrame using the LOAD FROM df WHERE age < 30 RETURN * -- note that you can be selective and specify filter conditions, and whatever's found in Pandas will be returned by the RETURN * portion of the subquery (you can also return only specific columns of the Pandas DataFrame if desired). The results of the subquery are then passed to COPY FROM as shown above.

[kamalikaray]

Thanks for your response, I am getting an error while trying this:
conn.execute("COPY person FROM (LOAD FROM person_df RETURN *)")

RuntimeError: Parser exception: extraneous input 'LOAD' expecting {StringLiteral, SP} (line: 1, offset: 15)
"COPY person FROM (LOAD FROM person_df RETURN *)"
^^^^
Could you please help me resolve this? I do not need any filters and want to copy the entire dataframe.

[prrao87]

Hi @Kamalika17, that's odd. Could you please copy-paste this code and try running (after updating to the latest Kùzu version, pip install -U kuzu):

import shutil
import kuzu
import pandas as pd

shutil.rmtree("./test", ignore_errors=True)
db = kuzu.Database("./test")
conn = kuzu.Connection(db)
conn.execute("CREATE NODE TABLE Person(name STRING, age INT64, PRIMARY KEY(name))")

df = pd.DataFrame({
    "name": ["Adam", "Karissa", "Zhang", "Noura"],
    "age": [30, 40, 50, 25]
})

conn.execute("COPY Person FROM (LOAD FROM df RETURN *)")
print(conn.execute("MATCH (p:Person) RETURN p.name, p.age").get_as_df())

I tried it and got the following output:

    p.name  p.age
0     Adam     30
1  Karissa     40
2    Zhang     50
3    Noura     25

[kamalikaray]

Hi,

Thanks a lot for your response. It worked! :) I think updating kuzu did the job.

Regards,
Kamalika