Skip to content

CHANGELOG.md

Latest commit

 

History

History
160 lines (132 loc) · 8 KB

CHANGELOG.md

File metadata and controls

160 lines (132 loc) · 8 KB
Raw

CHANGELOG — ifc_to_ficr

Intentional deviations of the new single-tool converter relative to the old two-stage pipeline (IFCtoLBD 2.44.4 + lbd_to_ficr_converter.py) and its golden output golden/Duplex_A_20110907_ficr.ttl. Rule ids (MIG-/IMP-) match the buckets in the classified diff report produced by ifc-to-ficr verify.

1.2.2 — 2026-07-10

Changed

  • TBox updated to v1.1.1 by the maintainer (5 unused properties removed: 3 object, 2 datatype). All terms emitted by the converter remain present — no conversion behaviour change (Duplex output identical: 2,606 triples).
  • Fingerprint gate relaxed to a warning. The TBox is under active development, so a fingerprint mismatch now logs a WARNING instead of aborting; the pinned reference was refreshed to v1.1.1 (2098/241/62/40/96). Output safety continues to rely on the hard gates: startup mapping-target validation and the emitted-term check.
  • Output header now reports the TBox version read dynamically from owl:versionInfo instead of a hard-coded string.

Repository

  • FiCR_ontology/ is no longer tracked (added to .gitignore): the ontology is developed in the FiCR Ontology project and updated frequently; users must place ficr.ttl + bot.ttl there locally (see README).
  • GitHub repository renamed ifc2FiCR -> IFCtoFiCR; local remote URL updated.

1.2.1 — 2026-07-04

Fixed

  • IfcPlate no longer maps to ficr:FloorSlab. Exposed by Clinic_Architectural.ifc (first model with plates): all 172 IfcPlate instances are glazed curtain-wall panels (Decomposes IfcCurtainWall), which the verbatim-ported Plate -> FloorSlab rule (intended for structural plates) misclassified as floor slabs — semantically wrong for fire analysis. Plates now fall back to bot:Element, consistent with the treatment of curtain-wall mullions; they stay reachable from their ficr:Wall-typed curtain wall via bot:hasSubElement. The archived Duplex baseline contains no plates, so its output is unaffected.

1.2.0 — 2026-07-04

Repository simplified to "current tool + archive"

  • golden/ dissolved. The three baseline data files (Duplex_A_20110907 .ifc/.ttl/_ficr.ttl) moved into archive/legacy_pipeline/, joining the tools that produced them (kept for reproducibility). The development-time verification result they anchored is recorded in the 1.1.0 notes below.
  • Verification machinery removed. verify.py, the verify CLI subcommand (exit code 5) and the pytest suite (tests/) were development-time scaffolding for the golden diff; the tool is now convert-only. Runtime safety gates remain: schema gate, TBox fingerprint gate, and the emitted-term gate.
  • Ontology directory renamed by the maintainer: onto/ -> FiCR_ontology/ (contents unchanged: ficr.ttl frozen v1.1.0 + bot.ttl). CLI defaults, error messages, output headers and docs updated accordingly. The legacy ficr.properties artifact moved to the archive.
  • README rewritten for the simplified layout.
  • lbd_outputs/ working directory removed (nothing writes there; the --emit-intermediate debug dump defaults to <input>_lbd.ttl next to the input). Build artifact ifc_to_ficr.egg-info/ deleted (regenerates on install).

1.1.0 — 2026-07-04

Frozen TBox adopted (maintainer action, upstream of this tool)

  • onto/ficr.ttl replaced with the final frozen TBox v1.1.0, edited and saved in Protégé; consistency verified with HermiT. Edits: ficr:hasREI removed, ficr:NotRequired removed, ficr:hasActualREI kept with its ficr:FabricElement domain dropped. The file is read-only and authoritative.

Changed

  • REI retarget. _map_fire_ratings now emits ficr:hasActualREI (xsd:integer) instead of the removed ficr:hasREI. The golden dataset's fireRating values are all placeholders, so the REI channel is empty and this change produces zero verification diffs. The layer-2 normalizer gained the rename rule ficr:hasREI -> ficr:hasActualREI (counted in the report; expected 0 hits on the golden).
  • TBox fingerprint gate. tbox.py hard-codes the fingerprint of the frozen TBox (2110 total triples; in the ficr# namespace: 241 classes, 65 object properties, 42 datatype properties, 96 named individuals). Any mismatch at startup is a hard error: "onto/ficr.ttl is not the frozen v1.1.0 TBox".

Repository

  • Archive restructure. IFCtoLBD_CLI_2_44_4.jar and lbd_to_ficr_converter.py moved from golden/ to archive/legacy_pipeline/ (with a README explaining the superseded two-stage pipeline). golden/ now contains only the three baseline data files (.ifc, LBD .ttl, FiCR .ttl); their paths are unchanged.
  • README.md rewritten for the single-tool pipeline (new namespace, CLI, verification methodology).

1.0.0 — 2026-07-04

Expected migrations (mandated by the task spec)

  • MIG-01 — pure ABox. The old output embedded ~79 inline schema-stub triples (owl:Class / owl:DatatypeProperty / owl:ObjectProperty typings plus rdfs:range) so Protégé would type predicates without the TBox. The new output contains instance data only; the TBox is referenced via owl:imports <https://w3id.org/ficr>.
  • MIG-02 — ontology header retarget. owl:imports now points to https://w3id.org/ficr (frozen TBox v1.1.0) instead of https://w3id.org/bam/ficr, and the rdfs:comment text was updated accordingly. The ontology node IRI pattern (https://lbd.example.com/instances/<output-stem>) is unchanged.
  • Namespace migration. All ficr: terms moved from https://w3id.org/bam/ficr# to https://w3id.org/ficr#. Rename table applied during verification normalisation: ficr:hasElementUsage -> ficr:hasPhysicalObjectFireSafetyRole, ficr:islocatedIn -> ficr:isLocatedIn. Both renames matched 0 triples in the golden output — the old converter never emitted either term.

Expected improvements

  • IMP-01 — roof slab no longer double-typed as FloorSlab. The old converter mapped every LBD type triple independently, so the roof slab (IfcSlab with PredefinedType=ROOF, typed beo:Slab + beo:Slab-ROOF in LBD) received both ficr:FloorSlab (via beo:Slab) and ficr:RoofSlab (via beo:Slab-ROOF). The new class mapping is keyed by (IFC entity, PredefinedType) and resolves each element to exactly one most specific class: (IfcSlab, ROOF) -> ficr:RoofSlab. Asserting both ficr:FloorSlab and ficr:RoofSlab on one instance was a contradiction (floor vs roof), not a feature. Affects 1 triple in the golden diff.

Hardening (no diff impact on the golden dataset)

  • TBox term gate. Every emitted ficr:/bot: term is validated against the frozen TBox (onto/ficr.ttl + onto/bot.ttl); an unknown term is a hard error (exit 4). The old converter silently dropped unknown mapped properties and never validated classes.
  • Schema gate. Non-IFC2X3 input exits with code 3 and a message listing supported schemas, instead of undefined behaviour.

Documented reverse-engineering findings (behaviour replicated, not changed)

  • OmniClass classification data comes from Revit property-set strings ('OmniClass Table 13 Category', 'Category Description'); the IFC file contains no IfcClassificationReference/IfcRelAssociatesClassification entities, so the spec's stated channel is realised via these Psets.
  • bot:hasSubElement combines element-to-element IfcRelAggregates (stair/roof decomposition, 11 triples) with host->filler pairs derived from IfcRelVoidsElement+IfcRelFillsElement (wall->door, roof->window, 38 triples).
  • Elements without direct spatial containment inherit containment from the spatial node their ObjectPlacement is relative to (stair flights, stringers, railings -> storey); openings are excluded. Placement chain only, no coordinates.
  • Revit placeholder property values (string value equal to the property name, e.g. 'Area'='Area', 'FireRating'='FireRating') are skipped, matching the golden extraction; type-object property sets are not traversed.