ifctoneo4j 0.1.1

Convert IFC building models to a Linked Building Data graph in Neo4j — no RDF, no Turtle files.

ifctoneo4j

Convert IFC building models to a Linked Building Data graph in Neo4j — no RDF, no Turtle files.

---

Install

pip install ifctoneo4j

This will also install ifcopenshell and neo4j automatically.

For bounding box geometry and interface detection, also install:

pip install ifctoneo4j[geometry]

---

Python API

Basic usage

import ifctoneo4j

# Step 1 — parse the IFC file (no database needed)
result = ifctoneo4j.parse("model.ifc")
print(result)
# ParseResult(nodes=187, rels=186, elements=93)

# Step 2 — write to Neo4j
ifctoneo4j.write(
    result,
    neo4j_uri="bolt://localhost:7687",
    neo4j_user="neo4j",
    neo4j_password="secret",
)

parse() options

result = ifctoneo4j.parse(
    "model.ifc",
    properties_level=1,   # 1 = flat props on nodes (default)
                          # 2 = separate OPM PropertyNode per value
                          # 3 = OPM + versioned StateNode with timestamp
    include_units=False,  # attach QUDT unit URIs to numeric properties
    base_uri="https://linkedbuildingdata.org/building#",  # URI prefix
)

write() options

ifctoneo4j.write(
    result,
    neo4j_uri="bolt://localhost:7687",
    neo4j_user="neo4j",
    neo4j_password="secret",
    database="neo4j",   # target database name (important for AuraDB)
    clear_db=False,     # set True to wipe the database before importing
    batch_size=500,     # rows per Neo4j transaction
)
# returns {"nodes": 187, "rels": 186, "prop_nodes": 0}

AuraDB (cloud)

ifctoneo4j.write(
    result,
    neo4j_uri="neo4j+s://xxxxxxxx.databases.neo4j.io",
    neo4j_user="your-username",
    neo4j_password="your-password",
    database="your-database-name",
    clear_db=True,
)

---

CLI

ifctoneo4j model.ifc \
    --neo4j-uri bolt://localhost:7687 \
    --neo4j-user neo4j \
    --neo4j-password secret

Common flags

Flag	Description
--database	Target database name (default: neo4j)
--clear-db	Delete all data before importing
--properties-level 2	OPM property nodes instead of flat props
--units	Attach QUDT unit URIs to numeric properties
--geometry	Compute bounding boxes
--interfaces	Detect element interfaces (requires --geometry)
--geolocation	Extract site lat/lon as WKT POINT
--batch-size 1000	Rows per transaction (default: 500)
--no-elements	Spatial hierarchy only, skip elements
-v	Verbose / debug logging

# Full example with geometry and AuraDB
ifctoneo4j model.ifc \
    --neo4j-uri neo4j+s://xxxxxxxx.databases.neo4j.io \
    --neo4j-user myuser --neo4j-password mypassword \
    --database mydb \
    --clear-db --geometry --geolocation

---

What gets created in Neo4j

The graph follows the BOT spatial hierarchy:

Site -[HAS_BUILDING]-> Building -[HAS_STOREY]-> Storey -[HAS_SPACE]-> Space
Storey/Space -[CONTAINS_ELEMENT]-> Element
Wall -[HAS_SUB_ELEMENT]-> Door/Window

Element nodes get multi-label classification — a solid wall becomes :Element:Wall:Wall_SOLIDWALL, an axial fan becomes :Element:Fan:Fan_AXIAL.

Properties from IFC property sets are stored as camelCase_property_simple on each node:

MATCH (w:Wall)
WHERE w.isExternal_property_simple = true
RETURN w.name, w.globalId