Extend get_transformer to support non-Python built-in transform types

get_transformer now handles jq, xslt, jsonld-frame, and node in addition
to python. Non-Python types are dispatched via a one-shot sub-subprocess
running under the main venv interpreter (_MAIN_PYTHON), which has all
ogc.bblocks dependencies (lxml, jq, pyld, rdflib, etc.) available
regardless of any sandboxed venv the outer Python transform may be using.

The callable signature gains an optional source_mime_type parameter so
callers can tell transforms like jsonld-frame how to parse their input.

For node transforms, npm dependencies declared in the transform's metadata
are installed into <sandbox_base>/get_transformer/<bblock_id>/<transform_id>/node
on the first call, mirroring the layout used by ordinary node sandboxes.

Implementation changes:
- TransformMetadata gains _transforms_registry and _sandbox_base_dir
  (underscore-prefixed: internal, not part of the transform author API)
- _build_transforms_registry includes all _GET_TRANSFORMER_TYPES (python,
  node, jq, xslt, jsonld-frame), storing the transform type alongside code
- apply_transforms passes _sandbox_base_dir=sandbox_dir to TransformMetadata
- python.py restructured: harness generation split into _NON_PYTHON_HARNESS_CODE,
  _HARNESS_IMPORTS, and _HARNESS_BODY constants assembled by
  _build_persistent_harness, eliminating all brace-escaping from the harness
  template; _MAIN_PYTHON and _SANDBOX_BASE embedded as harness constants

Author: avillar

ogc/bblocks/models.py

@@ -846,7 +846,8 @@ class TransformMetadata:
     sandbox_dir: Path | None = None
     id: str | None = None
     ctx: TransformContext | None = None
-    transforms_registry: dict | None = None
+    _transforms_registry: dict | None = None
+    _sandbox_base_dir: Path | None = None
 
 
 @dataclasses.dataclass

ogc/bblocks/transform.py

@@ -296,42 +296,45 @@ def cleanup_sandbox(sandbox_dir: Path, bblocks: list[BuildingBlock]) -> None:
             # temporarily renamed) — leave sandboxes intact to avoid a full rebuild on rename-back
 
 
+_GET_TRANSFORMER_TYPES = frozenset({'python', 'node', 'jq', 'xslt', 'jsonld-frame'})
+
+
 def _build_transforms_registry(bblocks_register: BuildingBlockRegister) -> dict:
     registry = {}
 
     for bblock_id, bblock in bblocks_register.bblocks.items():
         transforms = bblock.transforms
         if not transforms:
             continue
-        python_transforms = {
-            t['id']: {'code': t['code'], 'metadata': t.get('metadata') or {}}
-            for t in transforms if t.get('type') == 'python' and t.get('code')
+        supported_transforms = {
+            t['id']: {'type': t['type'], 'code': t['code'], 'metadata': t.get('metadata') or {}}
+            for t in transforms if t.get('type') in _GET_TRANSFORMER_TYPES and t.get('code')
         }
-        if not python_transforms:
+        if not supported_transforms:
             continue
         bblock_metadata = json.loads(json.dumps(bblock.metadata, default=str))
         registry[bblock_id] = {
             'name': bblock.metadata.get('name', bblock_id),
             'version': bblock.metadata.get('version'),
             'bblock_metadata': bblock_metadata,
-            'transforms': python_transforms,
+            'transforms': supported_transforms,
         }
 
     for bblock_id, raw in bblocks_register.imported_bblocks.items():
         transforms = raw.get('transforms') or []
-        python_transforms = {
-            t['id']: {'code': t['code'], 'metadata': t.get('metadata') or {}}
-            for t in transforms if t.get('type') == 'python' and t.get('code')
+        supported_transforms = {
+            t['id']: {'type': t['type'], 'code': t['code'], 'metadata': t.get('metadata') or {}}
+            for t in transforms if t.get('type') in _GET_TRANSFORMER_TYPES and t.get('code')
         }
-        if not python_transforms:
+        if not supported_transforms:
             continue
         raw_copy = {k: v for k, v in raw.items() if k != 'register'}
         bblock_metadata = json.loads(json.dumps(raw_copy, default=str))
         registry[bblock_id] = {
             'name': raw.get('name', bblock_id),
             'version': raw.get('version'),
             'bblock_metadata': bblock_metadata,
-            'transforms': python_transforms,
+            'transforms': supported_transforms,
         }
 
     return registry
@@ -490,8 +493,9 @@ def apply_transforms(bblock: BuildingBlock,
                                                        input_data=snippet['code'],
                                                        sandbox_dir=transform_sandboxes.get(
                                                            transform['id'], sandbox_dir),
+                                                       _sandbox_base_dir=sandbox_dir,
                                                        ctx=ctx,
-                                                       transforms_registry=transforms_registry)
+                                                       _transforms_registry=transforms_registry)
 
                 try:
                     result = transformer.transform(transform_metadata)