LDBC SNB Interactive — IC01–IC12
NamiDB targets the 12 in-scope Complex Read queries from the LDBC Social Network Benchmark — Interactive workload. All twelve parse, plan, and execute end-to-end on the v0.3 engine.
Quick reference
| # | Name | Pattern |
|---|---|---|
| IC01 | Friends with a given name | 1–3 hop KNOWS traversal |
| IC02 | Recent messages from friends | KNOWS + POST + ORDER BY |
| IC03 | Friends and friends-of-friends in countries X and Y | 2-hop + filter |
| IC04 | New topics | tag aggregation over a window |
| IC05 | New groups | forum membership window |
| IC06 | Tag co-occurrence | edge join + group by |
| IC07 | Recent likers | LIKES + recency |
| IC08 | Recent replies | REPLY_OF traversal |
| IC09 | Recent messages by friends and FoF | 1–2 hop KNOWS + posts |
| IC10 | Friend recommendation | 2-hop similarity |
| IC11 | Friend’s job referrals | KNOWS + WORKS_AT |
| IC12 | Expert search | KNOWS + tag intersection |
IC01 — Find friends with a given name
MATCH (p:Person {_id: $personId})-[:KNOWS*1..3]-(friend:Person {firstName: $name})WHERE friend._id <> $personIdRETURN friend._id AS friendId, friend.lastName AS friendLastName, length(path) AS distanceORDER BY distance ASC, friendLastName ASC, friendId ASCLIMIT 20IC02 — Recent messages from friends
MATCH (p:Person {_id: $personId})-[:KNOWS]-(friend)MATCH (friend)<-[:HAS_CREATOR]-(message)WHERE message.creationDate <= $maxDateRETURN friend._id AS friendId, friend.firstName AS firstName, message._id AS messageId, message.content AS content, message.creationDate AS creationDateORDER BY creationDate DESC, messageId ASCLIMIT 20(IC03–IC12 in the bench harness)
The full canonical query texts live in
crates/namidb-query/tests/fixtures/
in the engine repo. The bench harness under
bench/ runs all
twelve against a synthetic dataset and prints a comparison vs Kùzu.
Running the bench locally
git clone https://github.com/namidb/namidb.gitcd namidb/benchpython kuzu_runner.py # generate datasetcargo run --release -p namidb-benchSee bench/README.md
for the full reproduction recipe.