beam: convert STEP assemblies via OCCT XCAF

converter/worker.py

@@ -1,20 +1,35 @@
 import hashlib
 import json
 import os
+import struct
 import sys
 import time
 from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any
 
-import cadquery as cq
+from OCP.BRepMesh import BRepMesh_IncrementalMesh
+from OCP.Message import Message_ProgressRange
+from OCP.RWGltf import RWGltf_CafWriter
+from OCP.STEPCAFControl import STEPCAFControl_Reader
+from OCP.TColStd import TColStd_IndexedDataMapOfStringString
+from OCP.TCollection import TCollection_AsciiString, TCollection_ExtendedString
+from OCP.TDF import TDF_LabelSequence
+from OCP.TDocStd import TDocStd_Document
+from OCP.XCAFApp import XCAFApp_Application
+from OCP.XCAFDoc import XCAFDoc_DocumentTool, XCAFDoc_ShapeTool
 
 
 CONVERTER_NAME = "NodeDcBimConverter"
-CONVERTER_VERSION = "0.2.0"
+CONVERTER_VERSION = "0.3.0"
 UPLOADS_DIR = Path(os.environ.get("NODEDC_BIM_CONVERTER_UPLOADS_DIR", "/beam/uploads")).resolve()
 POLL_INTERVAL_SECONDS = float(os.environ.get("NODEDC_BIM_CONVERTER_INTERVAL_SECONDS", "10"))
 PROCESS_ONCE = os.environ.get("NODEDC_BIM_CONVERTER_ONCE", "").lower() in {"1", "true", "yes"}
+REPROCESS_READY_ON_VERSION_CHANGE = os.environ.get(
+    "NODEDC_BIM_CONVERTER_REPROCESS_READY_ON_VERSION_CHANGE", ""
+).lower() in {"1", "true", "yes"}
+STEP_MESH_LINEAR_DEFLECTION = float(os.environ.get("NODEDC_BIM_CONVERTER_STEP_LINEAR_DEFLECTION", "0.1"))
+STEP_MESH_ANGULAR_DEFLECTION = float(os.environ.get("NODEDC_BIM_CONVERTER_STEP_ANGULAR_DEFLECTION", "0.1"))
 STEP_EXTENSIONS = {".step", ".stp"}
 
 
@@ -58,7 +73,7 @@ def json_safe(value: Any) -> Any:
     return str(value)
 
 
-def node_to_metadata(node: cq.Assembly, parent_path: str = "") -> dict[str, Any]:
+def node_to_metadata(node: Any, parent_path: str = "") -> dict[str, Any]:
     name = node.name or "root"
     node_path = f"{parent_path}/{name}" if parent_path else name
     children = [node_to_metadata(child, node_path) for child in node.children]
@@ -76,6 +91,11 @@ def count_nodes(node: dict[str, Any]) -> int:
     return 1 + sum(count_nodes(child) for child in node.get("children", []))
 
 
+def metadata_dict(value: Any) -> dict[str, Any]:
+    safe_value = json_safe(value or {})
+    return safe_value if isinstance(safe_value, dict) else {"value": safe_value}
+
+
 def get_shape_type(value: Any) -> str | None:
     try:
         shape_type = value.ShapeType()
@@ -84,7 +104,7 @@ def get_shape_type(value: Any) -> str | None:
         return None
 
 
-def collect_solids(workplane: cq.Workplane) -> list[Any]:
+def collect_solids(workplane: Any) -> list[Any]:
     solids: list[Any] = []
     seen: set[int] = set()
 
@@ -106,11 +126,61 @@ def collect_solids(workplane: cq.Workplane) -> list[Any]:
     return solids
 
 
-def import_step_assembly(source_path: Path) -> tuple[cq.Assembly, str]:
+def unique_assembly_name(parent: Any, requested_name: Any) -> str:
+    base_name = str(requested_name or "component")
+    if base_name not in parent.objects:
+        return base_name
+
+    index = 2
+    while True:
+        candidate = f"{base_name}__{index:03d}"
+        if candidate not in parent.objects:
+            return candidate
+        index += 1
+
+
+def import_step_with_unique_component_names(source_path: Path, cq_module: Any) -> Any:
+    original_add = cq_module.Assembly.add
+
+    def add_with_unique_names(self: Any, arg: Any, *args: Any, **kwargs: Any) -> Any:
+        if isinstance(arg, cq_module.Assembly):
+            positional_name = args[0] if args else None
+            requested_name = kwargs.get("name", positional_name) or arg.name
+            unique_name = unique_assembly_name(self, requested_name)
+
+            if unique_name != requested_name:
+                arg.metadata = {
+                    **metadata_dict(arg.metadata),
+                    "originalName": str(requested_name),
+                }
+                if args:
+                    args = (unique_name, *args[1:])
+                else:
+                    kwargs = {**kwargs, "name": unique_name}
+
+        return original_add(self, arg, *args, **kwargs)
+
+    cq_module.Assembly.add = add_with_unique_names
     try:
-        return cq.Assembly.importStep(str(source_path)), "assembly"
+        return cq_module.Assembly.importStep(str(source_path))
+    finally:
+        cq_module.Assembly.add = original_add
+
+
+def import_step_assembly(source_path: Path) -> tuple[Any, str, str]:
+    import cadquery as cq
+
+    try:
+        return cq.Assembly.importStep(str(source_path)), "assembly", getattr(cq, "__version__", "unknown")
     except ValueError as exc:
-        if "does not contain an assembly" not in str(exc):
+        error_message = str(exc)
+        if "Unique name is required" in error_message and "already in the assembly" in error_message:
+            return (
+                import_step_with_unique_component_names(source_path, cq),
+                "assembly-unique-names",
+                getattr(cq, "__version__", "unknown"),
+            )
+        if "does not contain an assembly" not in error_message:
             raise
 
     step_model = cq.importers.importStep(str(source_path))
@@ -120,17 +190,213 @@ def import_step_assembly(source_path: Path) -> tuple[cq.Assembly, str]:
     if len(solids) > 1:
         for index, solid in enumerate(solids, start=1):
             assy.add(solid, name=f"{source_path.stem}_solid_{index:03d}")
-        return assy, "split-solids"
+        return assy, "split-solids", getattr(cq, "__version__", "unknown")
 
     assy.add(step_model, name=source_path.stem)
-    return assy, "single-shape"
+    return assy, "single-shape", getattr(cq, "__version__", "unknown")
 
 
-def convert_step_to_glb(source_path: Path, glb_path: Path, metadata_path: Path) -> dict[str, Any]:
+def read_glb_json(glb_path: Path) -> tuple[dict[str, Any], list[tuple[bytes, bytes]]]:
-    assy, import_strategy = import_step_assembly(source_path)
+    data = glb_path.read_bytes()
+    if len(data) < 20:
+        raise ValueError("GLB is too small.")
+
+    magic, version, declared_length = struct.unpack("<4sII", data[:12])
+    if magic != b"glTF" or version != 2:
+        raise ValueError("Only binary glTF v2 files are supported.")
+    if declared_length != len(data):
+        raise ValueError("GLB declared length does not match file size.")
+
+    gltf_json: dict[str, Any] | None = None
+    chunks: list[tuple[bytes, bytes]] = []
+    offset = 12
+    while offset < len(data):
+        if offset + 8 > len(data):
+            raise ValueError("Invalid GLB chunk header.")
+        chunk_length, chunk_type = struct.unpack("<I4s", data[offset:offset + 8])
+        offset += 8
+        chunk_data = data[offset:offset + chunk_length]
+        offset += chunk_length
+        if chunk_type == b"JSON":
+            gltf_json = json.loads(chunk_data.decode("utf-8"))
+        else:
+            chunks.append((chunk_type, chunk_data))
+
+    if gltf_json is None:
+        raise ValueError("GLB JSON chunk is missing.")
+    return gltf_json, chunks
+
+
+def write_glb_json(glb_path: Path, gltf_json: dict[str, Any], chunks: list[tuple[bytes, bytes]]) -> None:
+    json_bytes = json.dumps(gltf_json, ensure_ascii=False, separators=(",", ":")).encode("utf-8")
+    json_padding = (4 - (len(json_bytes) % 4)) % 4
+    json_chunk = json_bytes + (b" " * json_padding)
+
+    body = struct.pack("<I4s", len(json_chunk), b"JSON") + json_chunk
+    for chunk_type, chunk_data in chunks:
+        chunk_padding = (4 - (len(chunk_data) % 4)) % 4
+        padded_chunk = chunk_data + (b"\x00" * chunk_padding)
+        body += struct.pack("<I4s", len(padded_chunk), chunk_type) + padded_chunk
+
+    glb_path.write_bytes(struct.pack("<4sII", b"glTF", 2, 12 + len(body)) + body)
+
+
+def normalize_glb_node_names(glb_path: Path) -> tuple[dict[str, Any], int]:
+    gltf_json, chunks = read_glb_json(glb_path)
+    nodes = gltf_json.get("nodes", [])
+    meshes = gltf_json.get("meshes", [])
+    renamed = 0
+
+    for node in nodes:
+        mesh_index = node.get("mesh")
+        if not isinstance(mesh_index, int) or mesh_index < 0 or mesh_index >= len(meshes):
+            continue
+        mesh_name = meshes[mesh_index].get("name")
+        node_name = node.get("name")
+        if mesh_name and (not node_name or str(node_name).startswith("NAUO")):
+            node["name"] = mesh_name
+            renamed += 1
+
+    if renamed:
+        write_glb_json(glb_path, gltf_json, chunks)
+    return gltf_json, renamed
+
+
+def gltf_node_name(gltf_json: dict[str, Any], node: dict[str, Any], index: int) -> str:
+    name = node.get("name")
+    if name:
+        return str(name)
+    meshes = gltf_json.get("meshes", [])
+    mesh_index = node.get("mesh")
+    if isinstance(mesh_index, int) and 0 <= mesh_index < len(meshes):
+        mesh_name = meshes[mesh_index].get("name")
+        if mesh_name:
+            return str(mesh_name)
+    return f"node_{index}"
+
+
+def gltf_node_to_metadata(gltf_json: dict[str, Any], node_index: int, parent_path: str = "") -> dict[str, Any]:
+    nodes = gltf_json.get("nodes", [])
+    node = nodes[node_index]
+    name = gltf_node_name(gltf_json, node, node_index)
+    node_path = f"{parent_path}/{name}#{node_index}" if parent_path else f"{name}#{node_index}"
+    children = [
+        gltf_node_to_metadata(gltf_json, child_index, node_path)
+        for child_index in node.get("children", [])
+        if isinstance(child_index, int) and 0 <= child_index < len(nodes)
+    ]
+    return {
+        "id": hash_id(node_path),
+        "name": name,
+        "path": node_path,
+        "hasShape": isinstance(node.get("mesh"), int),
+        "metadata": {},
+        "children": children,
+    }
+
+
+def gltf_component_tree(gltf_json: dict[str, Any], fallback_name: str) -> dict[str, Any]:
+    nodes = gltf_json.get("nodes", [])
+    scenes = gltf_json.get("scenes", [])
+    scene_index = gltf_json.get("scene", 0)
+    scene = scenes[scene_index] if isinstance(scene_index, int) and 0 <= scene_index < len(scenes) else {}
+    root_indexes = [idx for idx in scene.get("nodes", []) if isinstance(idx, int) and 0 <= idx < len(nodes)]
+
+    if len(root_indexes) == 1:
+        return gltf_node_to_metadata(gltf_json, root_indexes[0])
+
+    children = [gltf_node_to_metadata(gltf_json, idx, fallback_name) for idx in root_indexes]
+    return {
+        "id": hash_id(fallback_name),
+        "name": fallback_name,
+        "path": fallback_name,
+        "hasShape": False,
+        "metadata": {},
+        "children": children,
+    }
+
+
+def convert_step_xcaf_to_glb(source_path: Path, glb_path: Path) -> tuple[dict[str, Any], dict[str, Any]]:
+    app = XCAFApp_Application.GetApplication_s()
+    doc = TDocStd_Document(TCollection_ExtendedString("MDTV-XCAF"))
+    app.NewDocument(TCollection_ExtendedString("MDTV-XCAF"), doc)
+
+    print(f"[{CONVERTER_NAME}] STEPCAF read {source_path.name}", flush=True)
+    reader = STEPCAFControl_Reader()
+    reader.SetNameMode(True)
+    reader.SetColorMode(True)
+    reader.SetLayerMode(True)
+    read_status = reader.ReadFile(str(source_path))
+    if "RetDone" not in str(read_status):
+        raise ValueError(f"STEPCAF read failed: {read_status}")
+    if not reader.Transfer(doc):
+        raise ValueError("STEPCAF transfer failed.")
+
+    shape_tool = XCAFDoc_DocumentTool.ShapeTool_s(doc.Main())
+    labels = TDF_LabelSequence()
+    shape_tool.GetFreeShapes(labels)
+    if labels.Length() == 0:
+        raise ValueError("STEPCAF document has no free shapes.")
+
+    print(
+        f"[{CONVERTER_NAME}] meshing {source_path.name}: "
+        f"{labels.Length()} free shape(s), deflection={STEP_MESH_LINEAR_DEFLECTION}",
+        flush=True,
+    )
+    for index in range(1, labels.Length() + 1):
+        shape = XCAFDoc_ShapeTool.GetShape_s(labels.Value(index))
+        if not shape.IsNull():
+            BRepMesh_IncrementalMesh(
+                shape,
+                STEP_MESH_LINEAR_DEFLECTION,
+                False,
+                STEP_MESH_ANGULAR_DEFLECTION,
+                True,
+            )
+
+    glb_path.parent.mkdir(parents=True, exist_ok=True)
+    print(f"[{CONVERTER_NAME}] GLB export {source_path.name}", flush=True)
+    writer = RWGltf_CafWriter(TCollection_AsciiString(str(glb_path)), True)
+    file_info = TColStd_IndexedDataMapOfStringString()
+    if not writer.Perform(doc, file_info, Message_ProgressRange()):
+        raise ValueError("GLB export failed.")
+    if not glb_path.exists() or glb_path.stat().st_size == 0:
+        raise ValueError("GLB export produced no output.")
+
+    gltf_json, renamed_nodes = normalize_glb_node_names(glb_path)
+    tree = gltf_component_tree(gltf_json, source_path.stem)
+    stats = {
+        "freeShapeCount": labels.Length(),
+        "gltfNodeCount": len(gltf_json.get("nodes", [])),
+        "gltfMeshCount": len(gltf_json.get("meshes", [])),
+        "renamedNodeCount": renamed_nodes,
+        "linearDeflection": STEP_MESH_LINEAR_DEFLECTION,
+        "angularDeflection": STEP_MESH_ANGULAR_DEFLECTION,
+    }
+    return tree, stats
+
+
+def convert_step_cadquery_to_glb(source_path: Path, glb_path: Path) -> tuple[dict[str, Any], dict[str, Any]]:
+    assy, import_strategy, cadquery_version = import_step_assembly(source_path)
     tree = node_to_metadata(assy)
     glb_path.parent.mkdir(parents=True, exist_ok=True)
     assy.export(str(glb_path), tolerance=0.1, angularTolerance=0.1)
+    stats = {
+        "fallback": True,
+        "importStrategy": import_strategy,
+        "cadqueryVersion": cadquery_version,
+    }
+    return tree, stats
+
+
+def convert_step_to_glb(source_path: Path, glb_path: Path, metadata_path: Path) -> dict[str, Any]:
+    import_strategy = "occt-xcaf-gltf"
+    try:
+        tree, stats = convert_step_xcaf_to_glb(source_path, glb_path)
+    except Exception as exc:
+        print(f"[{CONVERTER_NAME}] OCCT/XCAF path failed for {source_path}: {exc}", file=sys.stderr, flush=True)
+        tree, stats = convert_step_cadquery_to_glb(source_path, glb_path)
+        import_strategy = stats.get("importStrategy") or "cadquery"
 
     metadata = {
         "source": src_from_path(source_path),
@@ -144,9 +410,10 @@ def convert_step_to_glb(source_path: Path, glb_path: Path, metadata_path: Path)
         "converter": {
             "name": CONVERTER_NAME,
             "version": CONVERTER_VERSION,
-            "engine": "cadquery",
+            "engine": "occt-xcaf" if import_strategy == "occt-xcaf-gltf" else "cadquery",
-            "cadqueryVersion": getattr(cq, "__version__", "unknown"),
+            "cadqueryVersion": stats.get("cadqueryVersion"),
         },
+        "stats": stats,
     }
     write_json_atomic(metadata_path, metadata)
     return metadata
@@ -155,7 +422,7 @@ def convert_step_to_glb(source_path: Path, glb_path: Path, metadata_path: Path)
 def should_process(source_path: Path, manifest: dict[str, Any], glb_path: Path) -> bool:
     status = manifest.get("status")
     if status == "ready" and glb_path.exists():
-        return manifest.get("converterVersion") != CONVERTER_VERSION
+        return REPROCESS_READY_ON_VERSION_CHANGE and manifest.get("converterVersion") != CONVERTER_VERSION
     if status == "processing":
         updated_at = manifest.get("updatedAt")
         if updated_at: