Window Functions¶

Uni supports Cypher window queries with OVER (...) clauses for both manual window functions (like ROW_NUMBER) and aggregate windows (like SUM OVER). These are useful for ranking, running totals, and partitioned analytics directly in graph queries.

What It Provides¶

Window specifications with PARTITION BY and ORDER BY.
Manual window functions: ROW_NUMBER, RANK, DENSE_RANK, LAG, LEAD, NTILE, FIRST_VALUE, LAST_VALUE, NTH_VALUE.
Aggregate window functions: SUM, AVG, MIN, MAX, COUNT with OVER.

Example: Manual Window Function¶

RustPython

use uni_db::Uni;

# async fn demo() -> Result<(), uni_db::UniError> {
let db = Uni::open("./my_db").build().await?;

let rows = db.query(
    "MATCH (p:Purchase) RETURN p.userId, p.amount, ROW_NUMBER() OVER (PARTITION BY p.userId ORDER BY p.timestamp) AS rn"
).await?;

println!("{:?}", rows);
# Ok(())
# }

import uni_db

db = uni_db.Database("./my_db")
rows = db.query(
    "MATCH (p:Purchase) RETURN p.userId, p.amount, ROW_NUMBER() OVER (PARTITION BY p.userId ORDER BY p.timestamp) AS rn"
)
print(rows)

Example: LAG/LEAD for Change Tracking¶

RustPython

use uni_db::Uni;

# async fn demo() -> Result<(), uni_db::UniError> {
let db = Uni::open("./my_db").build().await?;

let rows = db.query(
    "MATCH (p:Price) RETURN p.symbol, p.ts, p.value, LAG(p.value) OVER (PARTITION BY p.symbol ORDER BY p.ts) AS prev_value"
).await?;

println!("{:?}", rows);
# Ok(())
# }

import uni_db

db = uni_db.Database("./my_db")
rows = db.query(
    "MATCH (p:Price) RETURN p.symbol, p.ts, p.value, LAG(p.value) OVER (PARTITION BY p.symbol ORDER BY p.ts) AS prev_value"
)
print(rows)

Example: Aggregate Window Function¶

RustPython

use uni_db::Uni;

# async fn demo() -> Result<(), uni_db::UniError> {
let db = Uni::open("./my_db").build().await?;

let rows = db.query(
    "MATCH (p:Purchase) RETURN p.userId, p.amount, SUM(p.amount) OVER (PARTITION BY p.userId ORDER BY p.timestamp) AS running_total"
).await?;

println!("{:?}", rows);
# Ok(())
# }

import uni_db

db = uni_db.Database("./my_db")
rows = db.query(
    "MATCH (p:Purchase) RETURN p.userId, p.amount, SUM(p.amount) OVER (PARTITION BY p.userId ORDER BY p.timestamp) AS running_total"
)
print(rows)

Example: Partition Totals Without Ordering¶

RustPython

use uni_db::Uni;

# async fn demo() -> Result<(), uni_db::UniError> {
let db = Uni::open("./my_db").build().await?;

let rows = db.query(
    "MATCH (p:Purchase) RETURN p.userId, p.amount, COUNT(*) OVER (PARTITION BY p.userId) AS user_count"
).await?;

println!("{:?}", rows);
# Ok(())
# }

import uni_db

db = uni_db.Database("./my_db")
rows = db.query(
    "MATCH (p:Purchase) RETURN p.userId, p.amount, COUNT(*) OVER (PARTITION BY p.userId) AS user_count"
)
print(rows)

Use Cases¶

Per-user ranking or top-N within partitions.
Running totals and cumulative metrics.
Sessionization or time-ordered analytics.

When To Use¶

Use window functions when you need per-partition analytics within the same query result.

Current Limitation¶

Warning

Queries that mix manual window functions (like ROW_NUMBER) and aggregate window functions (like SUM OVER) in the same query are not yet supported.