ummaya 0.2.3 → 0.2.5

package/src/ummaya/tools/documents/diff.py

@@ -0,0 +1,275 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Structured derivative diff generation for document patches."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+
+from ummaya.tools.documents.models import (
+    DocumentChange,
+    DocumentChangeType,
+    DocumentDiff,
+    DocumentExtraction,
+    DocumentPatch,
+    DocumentViewportCamera,
+    MetadataValue,
+    OperationType,
+    RenderArtifactRecord,
+    TableCell,
+)
+
+__all__ = [
+    "DocumentChange",
+    "DocumentChangeType",
+    "DocumentDiff",
+    "DocumentViewportCamera",
+    "RenderArtifactRecord",
+    "diff_from_patch",
+]
+
+
+def diff_from_patch(
+    patch: DocumentPatch,
+    *,
+    source_artifact_id: str,
+    derivative_artifact_id: str,
+    render_artifacts: tuple[RenderArtifactRecord, ...] = (),
+    before_extraction: DocumentExtraction | None = None,
+    after_extraction: DocumentExtraction | None = None,
+) -> DocumentDiff:
+    """Build a structured diff from an ordered patch request."""
+    changes = tuple(
+        DocumentChange(
+            change_id=f"change-{index:03d}",
+            operation_id=operation.operation_id,
+            change_type=_change_type(operation.operation_type),
+            target_path=operation.target_path,
+            display_label=_display_label_for_target(
+                before_extraction or after_extraction,
+                operation.target_path,
+            ),
+            before_value=_value_for_target(before_extraction, operation.target_path),
+            after_value=_after_value(
+                operation.value,
+                after_extraction,
+                operation.target_path,
+            ),
+        )
+        for index, operation in enumerate(patch.operations, start=1)
+    )
+    diff_sha256 = _diff_sha256(
+        source_artifact_id=source_artifact_id,
+        derivative_artifact_id=derivative_artifact_id,
+        changes=changes,
+        render_artifacts=render_artifacts,
+    )
+    diff_id = f"diff-{diff_sha256[:16]}"
+    return DocumentDiff(
+        diff_id=diff_id,
+        diff_sha256=diff_sha256,
+        resource_ref=f"document-diff://{diff_id}",
+        source_artifact_id=source_artifact_id,
+        derivative_artifact_id=derivative_artifact_id,
+        changes=changes,
+        render_artifacts=render_artifacts,
+    )
+
+
+def _diff_sha256(
+    *,
+    source_artifact_id: str,
+    derivative_artifact_id: str,
+    changes: tuple[DocumentChange, ...],
+    render_artifacts: tuple[RenderArtifactRecord, ...],
+) -> str:
+    payload = {
+        "changes": [change.model_dump(mode="json") for change in changes],
+        "derivative_artifact_id": derivative_artifact_id,
+        "render_artifacts": [record.model_dump(mode="json") for record in render_artifacts],
+        "source_artifact_id": source_artifact_id,
+    }
+    encoded = json.dumps(
+        payload,
+        ensure_ascii=False,
+        sort_keys=True,
+        separators=(",", ":"),
+    ).encode("utf-8")
+    return hashlib.sha256(encoded).hexdigest()
+
+
+def _change_type(operation_type: OperationType) -> DocumentChangeType:
+    if operation_type is OperationType.set_field_value:
+        return "field"
+    if operation_type is OperationType.set_table_cell:
+        return "table_cell"
+    if operation_type is OperationType.replace_text:
+        return "text"
+    if operation_type in {
+        OperationType.set_paragraph_style,
+        OperationType.set_run_style,
+        OperationType.set_cell_style,
+    }:
+        return "style"
+    if operation_type is OperationType.set_document_metadata:
+        return "metadata"
+    return "copy"
+
+
+def _value_for_target(
+    extraction: DocumentExtraction | None,
+    target_path: str,
+) -> str | None:
+    if extraction is None:
+        return None
+    value = _field_value_for_target(extraction, target_path)
+    if value is not None:
+        return value
+    value = _paragraph_value_for_target(extraction, target_path)
+    if value is not None:
+        return value
+    value = _table_value_for_target(extraction, target_path)
+    if value is not None:
+        return value
+    return _metadata_value_for_target(extraction, target_path)
+
+
+def _display_label_for_target(
+    extraction: DocumentExtraction | None,
+    target_path: str,
+) -> str | None:
+    if extraction is None:
+        return None
+    for field in extraction.fields:
+        if field.path == target_path:
+            return field.label
+    for table in extraction.tables:
+        matching_cell = _table_cell_for_target(table.cells, target_path)
+        if matching_cell is None:
+            continue
+        return _left_neighbor_label(table.cells, matching_cell)
+    return None
+
+
+def _after_value(
+    operation_value: MetadataValue,
+    after_extraction: DocumentExtraction | None,
+    target_path: str,
+) -> str | None:
+    if operation_value is not None:
+        return str(operation_value)
+    return _value_for_target(after_extraction, target_path)
+
+
+def _metadata_value_to_str(value: MetadataValue) -> str | None:
+    if value is None:
+        return None
+    return str(value)
+
+
+def _field_value_for_target(
+    extraction: DocumentExtraction,
+    target_path: str,
+) -> str | None:
+    for field in extraction.fields:
+        if field.path == target_path:
+            return _metadata_value_to_str(field.current_value)
+    return None
+
+
+def _paragraph_value_for_target(
+    extraction: DocumentExtraction,
+    target_path: str,
+) -> str | None:
+    if target_path in {"/text/body", "/data/body", "/code/body"}:
+        return "\n".join(paragraph.text for paragraph in extraction.paragraphs)
+    for paragraph in extraction.paragraphs:
+        if (
+            _target_path_matches(paragraph.source_path, target_path)
+            or paragraph.block_id == target_path
+        ):
+            return paragraph.text
+    return None
+
+
+def _table_value_for_target(
+    extraction: DocumentExtraction,
+    target_path: str,
+) -> str | None:
+    for table in extraction.tables:
+        if _target_path_matches(table.source_path, target_path) or table.block_id == target_path:
+            return _table_text(table.cells)
+        for cell in table.cells:
+            if _target_path_matches(cell.source_path, target_path):
+                return cell.text
+    return None
+
+
+def _table_cell_for_target(
+    cells: list[TableCell],
+    target_path: str,
+) -> TableCell | None:
+    for cell in cells:
+        if _target_path_matches(cell.source_path, target_path):
+            return cell
+    return None
+
+
+def _left_neighbor_label(
+    cells: list[TableCell],
+    target_cell: TableCell,
+) -> str | None:
+    candidates = [
+        cell
+        for cell in cells
+        if cell.row_index == target_cell.row_index
+        and cell.column_index < target_cell.column_index
+        and cell.text.strip()
+    ]
+    if not candidates:
+        return None
+    label_cell = max(candidates, key=lambda cell: cell.column_index)
+    return label_cell.text.strip()
+
+
+def _target_path_matches(extracted_path: str, target_path: str) -> bool:
+    if extracted_path == target_path:
+        return True
+    extracted_tail = _engine_path_tail(extracted_path)
+    target_tail = _engine_path_tail(target_path)
+    if extracted_tail is not None and extracted_tail == target_tail:
+        return True
+    if extracted_tail is not None:
+        return extracted_tail == target_path.lstrip("/")
+    return False
+
+
+def _engine_path_tail(path: str) -> str | None:
+    prefix = "engine://"
+    if not path.startswith(prefix):
+        return None
+    parts = path[len(prefix) :].split("/")
+    if len(parts) < 3:
+        return None
+    return "/".join(parts[2:])
+
+
+def _metadata_value_for_target(
+    extraction: DocumentExtraction,
+    target_path: str,
+) -> str | None:
+    metadata_key = _metadata_key_from_target(target_path)
+    if metadata_key is None or metadata_key not in extraction.metadata:
+        return None
+    return _metadata_value_to_str(extraction.metadata[metadata_key])
+
+
+def _metadata_key_from_target(target_path: str) -> str | None:
+    prefix = "/metadata/"
+    if target_path.startswith(prefix) and len(target_path) > len(prefix):
+        return target_path[len(prefix) :]
+    return None
+
+
+def _table_text(cells: list[TableCell]) -> str:
+    return "\n".join(cell.text for cell in cells if cell.text)

package/src/ummaya/tools/documents/engines.py

@@ -0,0 +1,163 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Engine registry for the Public AX document harness."""
+
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Protocol, runtime_checkable
+
+from ummaya.tools.documents.models import (
+    BlockedReason,
+    DocumentExtraction,
+    DocumentFormat,
+    DocumentPatch,
+)
+
+
+class DocumentMutationBlockedError(ValueError):
+    """Raised when a mutation engine blocks with a typed document reason."""
+
+    def __init__(self, reason: BlockedReason, message: str) -> None:
+        super().__init__(message)
+        self.reason = reason
+
+
+class DocumentInspectionEngine(Protocol):
+    """Promoted engine that can inspect one document format."""
+
+    document_format: DocumentFormat
+    engine_id: str
+
+    def inspect(self, path: Path, *, artifact_id: str) -> DocumentExtraction:
+        """Return LLM-readable document IR for a local artifact."""
+
+
+@runtime_checkable
+class DocumentMutationEngine(DocumentInspectionEngine, Protocol):
+    """Promoted engine that can apply an ordered patch to one document format."""
+
+    def apply_patch(self, path: Path, patch: DocumentPatch) -> bytes:
+        """Return derivative bytes after applying the patch."""
+
+
+class UnsupportedDocumentEngineError(LookupError):
+    """Raised when no promoted engine is registered for a format."""
+
+    def __init__(self, document_format: DocumentFormat) -> None:
+        super().__init__(f"No promoted document engine registered for {document_format.value}")
+        self.document_format = document_format
+
+
+class DocumentEngineRegistry:
+    """Session-local registry of promoted document engines."""
+
+    def __init__(self) -> None:
+        self._engines: dict[DocumentFormat, DocumentInspectionEngine] = {}
+
+    def register(self, engine: DocumentInspectionEngine) -> None:
+        """Register one promoted engine for its document format."""
+        if engine.document_format in self._engines:
+            raise ValueError(
+                f"document engine already registered for {engine.document_format.value}"
+            )
+        self._engines[engine.document_format] = engine
+
+    def get(self, document_format: DocumentFormat) -> DocumentInspectionEngine | None:
+        """Return the promoted engine for a format, if present."""
+        return self._engines.get(document_format)
+
+    def require(self, document_format: DocumentFormat) -> DocumentInspectionEngine:
+        """Return the promoted engine or fail closed."""
+        engine = self.get(document_format)
+        if engine is None:
+            raise UnsupportedDocumentEngineError(document_format)
+        return engine
+
+    def require_mutation(self, document_format: DocumentFormat) -> DocumentMutationEngine:
+        """Return a mutation-capable engine or fail closed."""
+        engine = self.require(document_format)
+        if not isinstance(engine, DocumentMutationEngine):
+            raise UnsupportedDocumentEngineError(document_format)
+        return engine
+
+
+def build_default_document_engine_registry() -> DocumentEngineRegistry:
+    """Build the default promoted-engine registry for local document tools."""
+    from ummaya.tools.documents.formats.archive import (  # noqa: PLC0415
+        ArchiveContainerDocumentEngine,
+    )
+    from ummaya.tools.documents.formats.code_file import (  # noqa: PLC0415
+        PythonSourceDocumentEngine,
+    )
+    from ummaya.tools.documents.formats.data_file import (  # noqa: PLC0415
+        DataFileDocumentEngine,
+    )
+    from ummaya.tools.documents.formats.hwp import (  # noqa: PLC0415
+        UnhwpReadOnlyInspectionEngine,
+    )
+    from ummaya.tools.documents.formats.hwpx import (  # noqa: PLC0415
+        HwpXPackageTextEngine,
+        OwpmlPackageTextEngine,
+    )
+    from ummaya.tools.documents.formats.odf import (  # noqa: PLC0415
+        OdfdoPresentationDocumentEngine,
+        OdfdoSpreadsheetDocumentEngine,
+        OdfdoTextDocumentEngine,
+    )
+    from ummaya.tools.documents.formats.ooxml import (  # noqa: PLC0415
+        OpenPyxlDocumentEngine,
+        PythonDocxDocumentEngine,
+        PythonPptxDocumentEngine,
+    )
+    from ummaya.tools.documents.formats.pdf import PypdfAcroFormEngine  # noqa: PLC0415
+    from ummaya.tools.documents.formats.text_web import (  # noqa: PLC0415
+        TextWebDocumentEngine,
+    )
+
+    registry = DocumentEngineRegistry()
+    registry.register(HwpXPackageTextEngine())
+    registry.register(OwpmlPackageTextEngine())
+    registry.register(UnhwpReadOnlyInspectionEngine())
+    registry.register(PythonDocxDocumentEngine())
+    registry.register(OpenPyxlDocumentEngine())
+    registry.register(PythonPptxDocumentEngine())
+    registry.register(PypdfAcroFormEngine())
+    registry.register(OdfdoTextDocumentEngine())
+    registry.register(OdfdoSpreadsheetDocumentEngine())
+    registry.register(OdfdoPresentationDocumentEngine())
+    registry.register(TextWebDocumentEngine(DocumentFormat.html))
+    registry.register(TextWebDocumentEngine(DocumentFormat.htm))
+    registry.register(TextWebDocumentEngine(DocumentFormat.txt))
+    registry.register(TextWebDocumentEngine(DocumentFormat.rtf))
+    registry.register(TextWebDocumentEngine(DocumentFormat.md))
+    registry.register(PythonSourceDocumentEngine())
+    for document_format in (
+        DocumentFormat.csv,
+        DocumentFormat.tsv,
+        DocumentFormat.xml,
+        DocumentFormat.rdf,
+        DocumentFormat.ttl,
+        DocumentFormat.lod,
+        DocumentFormat.json,
+        DocumentFormat.jsonl,
+        DocumentFormat.yaml,
+        DocumentFormat.yml,
+        DocumentFormat.geojson,
+        DocumentFormat.gpx,
+        DocumentFormat.kml,
+        DocumentFormat.fasta,
+        DocumentFormat.sgml,
+        DocumentFormat.dtd,
+        DocumentFormat.hml,
+        DocumentFormat.etc,
+    ):
+        registry.register(DataFileDocumentEngine(document_format))
+    for document_format in (
+        DocumentFormat.epub,
+        DocumentFormat.zip,
+        DocumentFormat.seven_z,
+        DocumentFormat.tar,
+        DocumentFormat.gz,
+    ):
+        registry.register(ArchiveContainerDocumentEngine(document_format))
+    return registry