GraphForge vs. DuckDB
When should graph workloads use columnar SQL instead?
DuckDB is the in-process analytical SQL engine: columnar, vectorized, dependency-free, and able to query Parquet, CSV, and JSON in place. Graphs can be walked with recursive CTEs, and the DuckPGQ community extension adds SQL/PGQ, although DuckDB's docs say it is not available for 1.5.x. GraphForge shares the columnar, embedded philosophy but specializes in graphs: Cypher, 94 algorithms, and investigation history.
On this page
Should I choose DuckDB or GraphForge?
DuckDB for SQL analytics over files; GraphForge when graph patterns and algorithms are the core of the work.
Choose GraphForge when
- Graph patterns, variable-length paths, and algorithms are the core of the analysis.
- You need algorithms beyond PageRank, clustering coefficient, and weakly connected components.
- You want checkpoints, provenance, and recorded hypotheses kept with the data.
Choose DuckDB when
- Most of the work is SQL analytics over Parquet, CSV, or lakehouse files, and graph questions are a small part.
- You want an analyst's Swiss Army knife with bindings for nearly every language, R included.
- Graph steps can be expressed as joins or recursive CTEs with USING KEY.
What can each one do?
22 of 26 capabilities differ. For 7, DuckDB's docs did not confirm support, so they are marked Not established rather than No.
Each capability has a status and a source. “Not established” means we could not confirm the feature from DuckDB's primary documentation. It does not mean DuckDB lacks it.
| Capability | GraphForge | DuckDB |
|---|---|---|
| Storage | ||
| Property graph model | Shipped Labeled property graph with typed properties. 1 | Extension DuckPGQ community extension adds SQL/PGQ property graphs over tables; docs say unavailable on 1.5.x, use 1.4.4. 20 |
| Document model | Not offered No document-store model; properties accept nested, heterogeneous values. 1 | Partial JSON type and JSON reading via the json extension; no document-store model. 17 |
| Relational tables | Not offered Not a relational database; results arrive as Arrow tables. 1 | Yes Full SQL with a large function library and window functions. 17 |
| Key-value access | Not offered No key-value API; data is a property graph. 1 | Not established No key-value API documented. |
| Columnar storage | Shipped Durable projects store graph data as Arrow and Parquet. 4 | Yes Columnar-vectorized query execution engine. 17 |
| Vector search | Shipped Exact cosine retrieval over embeddings you supply; no ANN index. 5 | Extension Experimental vss extension adds HNSW indexes on fixed-size ARRAY columns. 21 |
| Full-text search | Shipped Tantivy-backed BM25; hybrid text and vector retrieval through find(). 5 | Extension fts extension with BM25 scoring; index does not update automatically when the table changes. 22 |
| Geospatial | Not offered No spatial types or indexes are documented. | Extension Core spatial extension for geospatial data processing. 28 |
| Execution | ||
| Transactional workloads (OLTP) | Partial Local analytical scope; durable atomic writes with queued and optimistic writer modes. 6 | Partial ACID with MVCC, but engine is optimized for analytics and one process holds write access. 1727 |
| Analytical workloads (OLAP) | Shipped Analytical verbs and Cypher over a local columnar project. 7 | Yes Built for analytical queries with vectorized columnar execution. 17 |
| Graph traversal | Shipped Cypher MATCH patterns plus BFS, DFS, Dijkstra, A*, and reachability. 8 | Partial WITH RECURSIVE traversal with manual cycle detection; USING KEY CTEs; pattern matching via DuckPGQ. 1920 |
| Variable-length paths | Shipped Bounded and unbounded patterns and shortestPath(). 8 | Extension DuckPGQ quantified paths such as ->{1,3} and ANY SHORTEST; core needs recursive CTEs. 20 |
| Graph algorithms | Shipped 94 native algorithms behind analysis verbs, not Cypher CALL procedures. 37 | Extension DuckPGQ table functions: pagerank, local_clustering_coefficient, weakly_connected_component. 20 |
| Incremental computation | Not offered Algorithms recompute on request; no incremental mode is documented. 7 | Not established No incremental view maintenance in core docs. |
| Deployment | ||
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Yes Runs fully embedded in the host process; no server software to install. 17 |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Extension Quack extension (beta since 1.5.3) turns an instance into an HTTP server for DuckDB clients. 24 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Yes DuckDB-Wasm runs in browsers; single-threaded by default, 4 GB memory ceiling. 23 |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Not established No distributed execution in core docs. |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Partial Via MotherDuck, a separate company's serverless warehouse built on DuckDB. 25 |
| Investigation | ||
| Branching | Experimental Research versions, branches, slices, proposals, and forks ship in 0.6.0 release candidates. 14 | Not established No branching feature documented. |
| Snapshots / time travel | Shipped Named checkpoints you can diff and revert, plus transaction-time snapshots. 3 | Extension DuckLake tables support AT (VERSION => n) and AT (TIMESTAMP => t) queries. 26 |
| Immutable history | Shipped Knowledge ledger records immutable assertions and evidence lineage. 16 | Extension DuckLake keeps historic snapshots and changesets until compaction explicitly deletes them. 26 |
| Provenance | Shipped Provenance and knowledge ledger with evidence lineage. 16 | Not established No provenance feature documented. |
| Competing hypotheses | Shipped Immutable hypothesis groups with membership and selection events. 3 | Not established |
| Operations | ||
| ACID transactions | Partial Acknowledged-durable atomic publication; docs do not claim general ACID or serializable isolation. 6 | Yes ACID transactions via bulk-optimized MVCC; snapshot isolation. 17 |
| Replication | Not planned HA replication is out of scope. 9 | Not established No replication documented in core docs. |
Showing 26 of 26 capabilities.
How is each one built?
DuckDB: analytical relational (columnar). GraphForge: an embedded property graph library in Rust, persisted as Parquet.
GraphForge
GraphForge is a Rust library that runs inside your process: no daemon, server, or connection string. Projects persist to a local directory as Arrow and Parquet data plus JSON metadata, and every query and analysis verb returns an Apache Arrow table. Algorithms run in the same native engine as Cypher and are called as API verbs (rank, cluster, paths, analyze, similar), not as Cypher procedures.
DuckDB
DuckDB is a C++ library with a columnar-vectorized execution engine shipped as one header and one source file. Parquet, JSON, and S3 access are extensions that also work in DuckDB-Wasm. Graph traversal uses recursive CTEs with explicit cycle detection; DuckPGQ defines property graphs over existing tables.
DuckDB and GraphForge both put analytics in-process and speak Arrow. DuckDB is general SQL with graph as an add-on; GraphForge is graph-first, with algorithms as first-class verbs.
What the documentation says about DuckDB
- DuckDB runs fully embedded within a host process, with no DBMS server software to install, update or maintain.17
- DuckDB uses a columnar-vectorized query execution engine that processes batches of values per operation.17
- DuckDB has no external dependencies and ships as an amalgamation of one header and one implementation file.17
- Parquet, JSON, and HTTP(S)/S3 access are implemented as DuckDB extensions, which also work in DuckDB-Wasm.17
- The DuckDB docs show WITH RECURSIVE graph traversal over an LDBC Graphalytics graph, with explicit cycle detection.19
- A recursive CTE with USING KEY can replace earlier rows with the same key in each iteration, like a dictionary.19
- DuckPGQ implements SQL/PGQ syntax from SQL:2023, defining property graphs over existing tables and querying them with GRAPH_TABLE and MATCH.20
- The Quack extension, released May 12, 2026, lets other DuckDB instances connect to a DuckDB server over HTTP.24
- MotherDuck describes itself as the serverless cloud data warehouse built on DuckDB.25
Which one is faster?
Not measured yet. No head-to-head run has been published, so this page shows no performance chart.
Charts appear only after a reproducible run exists: pinned versions, disclosed hardware, generated data, raw results, and scripts.
No comparative results yet
The suites below apply to GraphForge and DuckDB. Each will be reported separately per benchmark class, never merged into one leaderboard.
Ingestion and writes
| Suite | Workloads | Metrics | Status |
|---|---|---|---|
| INGEST | CSV, JSONL, Parquet, graph edge lists | Records/s, MB/s, total elapsed | Not run |
Query, analytics, and system
| Suite | Workloads | Metrics | Status |
|---|---|---|---|
| LOOKUP | ID lookup, indexed property filter | p50/p95/p99 latency | Not run |
| TRAVERSE | BFS, k-hop, variable-length path | Latency, visited edges/s | Not run |
| JOIN | Multi-relation pattern matching | Latency, rows/s | Not run |
| ALGO | PageRank, WCC, SCC, Louvain, SSSP | Time, memory, correctness | Not run |
| EMBED | Startup, load, reopen, binary size | ms, MiB, disk footprint | Not run |
Which graph algorithms ship?
GraphForge: 94 native algorithms. DuckDB: 4 named in its docs, delivered as extension.
Algorithms are listed by name, as documented. Counts of extension modules or third-party libraries are not presented as native coverage.
GraphForge
94 Rust algorithms run in the engine behind five analysis verbs (rank, cluster, paths, analyze, similar) and return Arrow tables.37
A selection of the catalog:
- PageRank
- ArticleRank
- Betweenness (Brandes)
- Closeness
- Harmonic closeness
- Eigenvector
- HITS
- CELF influence maximization
- Triangle count
- K-core decomposition
- Louvain
- Leiden
- Label propagation
- Infomap
- Walktrap
- Girvan-Newman
- Weakly connected components
- Strongly connected components
- Dijkstra
- A*
- Bellman-Ford
- Yen's k-shortest paths
- Max flow / min cut
- Steiner tree
- Minimum spanning tree
- Topological sort
- Node2Vec
- GraphSAGE
- Fast random projection
- Node similarity (Jaccard)
DuckDB
Core DuckDB has no graph algorithms; the DuckPGQ community extension adds a few as SQL table functions over a declared property graph.20
Delivered as: extension
- PageRank
- Local Clustering Coefficient
- Weakly Connected Components
- Shortest path (ANY SHORTEST path pattern)
Where can each one run?
GraphForge runs in-process only. DuckDB: embedded yes, server extension, cluster not established.
| Mode | GraphForge | DuckDB |
|---|---|---|
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Yes Runs fully embedded in the host process; no server software to install. 17 |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Extension Quack extension (beta since 1.5.3) turns an instance into an HTTP server for DuckDB clients. 24 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Yes DuckDB-Wasm runs in browsers; single-threaded by default, 4 GB memory ceiling. 23 |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Not established No distributed execution in core docs. |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Partial Via MotherDuck, a separate company's serverless warehouse built on DuckDB. 25 |
| Replication | Not planned HA replication is out of scope. 9 | Not established No replication documented in core docs. |
What will my team write?
GraphForge: openCypher from Python, Node.js, Rust. DuckDB: SQL (PostgreSQL-like dialect), SQL/PGQ (via DuckPGQ community extension).
DuckDB's SQL dialect is PostgreSQL-like, with bindings for Python, R, Java, Node.js, Go, Rust, C, C++, ODBC, and WebAssembly. GraphForge uses openCypher from Python, Node.js, or Rust. The two work well together: both exchange Arrow tables in the same Python process.
- Query languages
- GraphForge openCypher
- DuckDB SQL (PostgreSQL-like dialect), SQL/PGQ (via DuckPGQ community extension)
- SDKs and bindings
- GraphForge Python, Node.js, Rust
- DuckDB Python, R, Java (JDBC), Node.js (Neo), Go, Rust, C, C++, ODBC, WebAssembly (DuckDB-Wasm), CLI
- Written in
- GraphForge Rust
- DuckDB C++
- Surfaces
- GraphForge Python, Node.js, Rust crates, CLI, VS Code extension, Notebooks, Agent skills
What are the trade-offs?
DuckDB's graph extension is unavailable on 1.5.x. GraphForge has no SQL and cannot query files in place.
A fair assessment cuts both ways.
GraphForge
- No SQL and no direct querying of Parquet or CSV files in place; data is published into a project first.
- No WebAssembly build and fewer language bindings.
- Not a general analytical database for non-graph workloads.
DuckDB
- The DuckDB docs say DuckPGQ is not available in the latest release (1.5.x) and point graph users to DuckDB v1.4.4.20
- In read-write mode only one process can open the database; multiple processes are allowed only in read-only mode.27
- vss HNSW indexes are limited to in-memory databases unless experimental persistence is enabled, and are then rewritten wholesale at each checkpoint.21
- The FTS index does not update automatically when the input table changes.22
- DuckDB-Wasm is single-threaded by default, with experimental multithreading and a 4 GB WebAssembly memory limit.23
How would I migrate?
Load DuckDB data through Arrow, pandas, or Polars. SQL queries are rewritten as Cypher.
Into GraphForge
- GraphForge 0.5.x loads data through Arrow, pandas, and Polars bulk calls or Cypher scripts; it has no native CSV loader.15
- Algorithms are API verbs (rank, cluster, paths, analyze, similar), not Cypher CALL procedures, so procedure calls become verb calls.7
- GraphForge passes the full openCypher 2024.3 TCK, so standard openCypher queries are the portable part of a migration.10
Moving from DuckDB
- DuckPGQ property graphs are views over ordinary vertex and edge tables, so the same tables can be exported to CSV or Parquet for graph bulk import.20
- SQL/PGQ MATCH patterns resemble Cypher ASCII-art patterns but are embedded in SQL GRAPH_TABLE calls, so queries need syntactic rewriting.20
- Graph logic written as recursive CTEs, including USING KEY variants, must be rewritten by hand as Cypher path patterns.19
What else do people ask about DuckDB?
Can DuckDB do graph queries? Through recursive CTEs; the DuckPGQ extension is unavailable on 1.5.x, per DuckDB's docs.
Can DuckDB do graph queries?
Core DuckDB uses recursive CTEs for traversal. The DuckPGQ community extension adds SQL/PGQ and a few algorithms, but DuckDB's docs say it is not available in 1.5.x and point graph users to v1.4.4.
Can I use DuckDB and GraphForge together?
Yes. Both exchange Apache Arrow tables, so you can prepare data in DuckDB, publish it into a GraphForge project, and bring algorithm results back into DuckDB.
Is GraphForge a DuckDB for graphs?
In spirit: both are embedded, columnar, and built for analysts. GraphForge stores a property graph, speaks Cypher, and ships graph algorithms rather than general SQL.
How was this researched?
From 28 primary sources, accessed 2026-10-19, each cited where it is used.
Capabilities were checked against primary documentation, repositories, and release notes, accessed 2026-10-19. Each cell records two things: the capability’s state and the evidence behind it. Every GraphForge cell carries a release status. Shipped means stable 0.5.2; experimental means 0.6.0 release candidates only. GraphForge is pre-1.0 and makes no backward-compatibility guarantee yet.
- Yes documented natively
- Partial limited or caveated
- Extension official extension or module
- Paid edition paid or enterprise edition only
- Removed removed, with the version cited
- Not established not confirmed by our research; not a “no”
- Shipped in stable 0.5.2
- Partial shipped with documented limits
- Experimental in 0.6.0 release candidates only
- Planned on the roadmap
- Not offered outside the product's model
- Not planned explicitly out of scope
Benchmark classes
- Equivalent workloads. The same logical operation with the same correctness requirement on every engine.
- Native optimized workloads. Each engine's best supported implementation, including projections, indexes, extensions, and native algorithms.
- End-to-end application workloads. Time from raw data to a usable result: import, indexing, projection, computation, and output retrieval.
Sources · accessed 2026-10-19
- GraphForge repository README — GitHub
- GraphForge releases — GitHub
- GraphForge 0.5.0 changelog — GitHub
- Storage architecture — GraphForge Book
- Analytics integration: search and retrieval — GraphForge Guide
- ADR 0018: acknowledged durability and isolation — GitHub
- Algorithm verbs — GraphForge Book
- Cypher guide — GraphForge Guide
- openCypher compatibility — GraphForge Reference
- openCypher TCK compliance — GraphForge Reference
- Remote authority peer extension (#215) — GitHub
- ADR 0027: native runtime boundary — GitHub
- Release scope and roadmap — GraphForge Docs
- Keep research history — GraphForge Guide
- Graph construction — GraphForge Guide
- Knowledge public API v1 — GraphForge Book
- Why DuckDB — duckdb.org
- Releases · duckdb/duckdb — GitHub
- WITH Clause (recursive CTEs, graph traversal, USING KEY) — DuckDB docs
- Graph Queries (DuckPGQ) — DuckDB docs
- Vector Similarity Search Extension (vss) — DuckDB docs
- Full-Text Search Extension — DuckDB docs
- DuckDB-Wasm Overview — DuckDB docs
- Quack Remote Protocol — DuckDB docs
- Getting Started — MotherDuck docs
- Time Travel — DuckLake docs
- Concurrency — DuckDB docs
- Spatial Extension — DuckDB docs
Try GraphForge on your own graph.
Install the package, open a project, and run the same notebook we measure with.