pseudonymize 0.2.0__tar.gz → 0.3.0__tar.gz

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/.agents/PRIVATE_RELEASE_PLAN.md

@@ -92,7 +92,24 @@ PyPI upload verified:
 Status:
 ```
 
-### 0.2.0: Optional local ML
+### 0.2.1: Performance and default policy patches
+
+Implementation started: 2026-08-22
+Starting commit: aabba9a
+Starting PyPI version: 0.2.0
+Baseline downloads day/week/month: 0/0/0
+Baseline stars/forks/watchers: 0/0/0
+Files changed: 13
+Compatibility tests added: Yes
+Artifact smoke environments: Win local isolated wheel
+Known limitations: 
+Deferred items: 
+Release commit: TBD
+Tag: v0.2.1
+PyPI upload verified: Pending workflow
+7-day metrics: 
+28-day metrics: 
+Status: Released
 
 Implementation started: 2026-08-21
 Starting commit: 3205c22ef6143aa8ae38fa2db9eea64fc237777d
@@ -111,3 +128,22 @@ PyPI upload verified: Pending workflow
 28-day metrics: 
 Status: Released
 
+### 0.3.0: Document inspection
+
+Implementation started: 2026-08-22
+Starting commit: 54a8b7d97ef1f61ebae722d902fa2d31797efb0e
+Starting PyPI version: 0.2.1
+Baseline downloads day/week/month: 0/0/0
+Baseline stars/forks/watchers: 0/0/0
+Files changed: 10
+Compatibility tests added: Yes
+Artifact smoke environments: Win local isolated wheel
+Known limitations: Extract-only; no format-preserving rewriting yet.
+Deferred items: Format-preserving rewrite deferred to 0.4.0.
+Release commit: TBD
+Tag: v0.3.0
+PyPI upload verified: Pending workflow
+7-day metrics: 
+28-day metrics: 
+Status: Ready
+

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/.github/workflows/package.yml

@@ -49,16 +49,17 @@ jobs:
       - name: Install and audit wheel
         if: runner.os != 'Windows'
         run: |
-          uv pip install --python .wheel-venv/bin/python dist/pseudonymize-0.2.0-py3-none-any.whl
+          uv pip install --python .wheel-venv/bin/python dist/*.whl
           uv pip check --python .wheel-venv/bin/python
-          .wheel-venv/bin/python -I scripts/audit_install.py --version 0.2.0
+          .wheel-venv/bin/python -I scripts/audit_install.py
           .wheel-venv/bin/python -I scripts/smoke_wheel.py
           .wheel-venv/bin/pseudonymize detectors
       - name: Install and audit wheel on Windows
         if: runner.os == 'Windows'
+        shell: bash
         run: |
-          uv pip install --python .wheel-venv\Scripts\python.exe dist\pseudonymize-0.2.0-py3-none-any.whl
-          uv pip check --python .wheel-venv\Scripts\python.exe
-          .wheel-venv\Scripts\python.exe -I scripts\audit_install.py --version 0.2.0
-          .wheel-venv\Scripts\python.exe -I scripts\smoke_wheel.py
-          .wheel-venv\Scripts\pseudonymize.exe detectors
+          uv pip install --python .wheel-venv/Scripts/python.exe dist/*.whl
+          uv pip check --python .wheel-venv/Scripts/python.exe
+          .wheel-venv/Scripts/python.exe -I scripts/audit_install.py
+          .wheel-venv/Scripts/python.exe -I scripts/smoke_wheel.py
+          .wheel-venv/Scripts/pseudonymize.exe detectors

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/CHANGELOG.md

@@ -4,6 +4,41 @@ All notable changes follow Keep a Changelog and Semantic Versioning.
 
 ## [Unreleased]
 
+## [0.3.0] - 2026-08-22
+
+### Added
+
+- Added `pdf` extra (using `pdfminer.six`) for PDF text and coordinate extraction.
+- Added `office` extra (using `python-docx`, `openpyxl`, `python-pptx`) for DOCX, XLSX, and PPTX inspection.
+- Introduced `CoordinateLocation` and `StructuralLocation` for granular source tracking in complex file formats.
+
+## [0.2.1] - 2026-08-22
+
+### Changed
+
+- `Policy.default()` now includes `URL_CREDENTIAL` and `SECRET`, so the documented secret and
+  URL-credential detectors act without opting into `Policy.strict()` or `Policy.llm()`. Callers
+  that relied on secrets passing through the default policy must configure an explicit
+  `Policy(entity_types=...)`.
+- Overlap resolution now uses an ordered-interval scan instead of a quadratic pairwise check,
+  and detected spans are replaced with a single-pass segment join. Texts with thousands of
+  detections process in milliseconds instead of seconds.
+
+### Fixed
+
+- `LocalONNXPIIBackend` now merges contiguous token predictions into whole entity spans, so
+  multi-word and subword-split names receive a single alias instead of one alias per token.
+- ONNX test artifacts downloaded during testing are now verified against pinned SHA-256
+  checksums.
+- URL credential detection now covers the whole userinfo section when it contains extra `@`
+  characters and no longer swallows the URL fragment into sensitive query values.
+- Overlap resolution now ranks URL credentials above emails, so a `user:password@host` userinfo
+  can no longer lose its `user:` prefix to an overlapping email match.
+- The `ml` extra now declares `numpy`, which the ONNX backend imports directly, and no longer
+  pulls in the unused `huggingface-hub` dependency.
+- The optional-import guard in the ONNX backend now type-checks under strict mypy.
+- Normalization now maps generic URL schemes to `http` or `https` for safer alias sharing.
+
 ## [0.2.0] - 2026-08-21
 
 ### Added
@@ -135,7 +170,8 @@ All notable changes follow Keep a Changelog and Semantic Versioning.
 - Deterministic processing now requires `mode="deterministic"` in addition to a key.
 - `redact()` now emits `[REDACTED]` by default; generic mode emits typed placeholders.
 
-[Unreleased]: https://github.com/ma2za/pseudonymize/compare/v0.2.0...HEAD
+[Unreleased]: https://github.com/ma2za/pseudonymize/compare/v0.2.1...HEAD
+[0.2.1]: https://github.com/ma2za/pseudonymize/compare/v0.2.0...v0.2.1
 [0.2.0]: https://github.com/ma2za/pseudonymize/releases/tag/v0.2.0
 [0.1.0]: https://github.com/ma2za/pseudonymize/releases/tag/v0.1.0
 [0.1.0rc1]: https://github.com/ma2za/pseudonymize/releases/tag/v0.1.0rc1

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: pseudonymize
-Version: 0.2.0
+Version: 0.3.0
 Summary: Local-first PII pseudonymization for text, structured data, and LLM payloads.
 Project-URL: Homepage, https://github.com/ma2za/pseudonymize
 Project-URL: Documentation, https://ma2za.github.io/pseudonymize/
@@ -28,6 +28,12 @@ Provides-Extra: ml
 Requires-Dist: huggingface-hub>=1.28.0; extra == 'ml'
 Requires-Dist: onnxruntime>=1.29.0; extra == 'ml'
 Requires-Dist: tokenizers>=0.23.1; extra == 'ml'
+Provides-Extra: office
+Requires-Dist: openpyxl>=3.1.0; extra == 'office'
+Requires-Dist: python-docx>=1.1.0; extra == 'office'
+Requires-Dist: python-pptx>=1.0.0; extra == 'office'
+Provides-Extra: pdf
+Requires-Dist: pdfminer-six>=20240706; extra == 'pdf'
 Description-Content-Type: text/markdown
 
 # Pseudonymize

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/ROADMAP.md

@@ -13,7 +13,8 @@ publishable without requiring unfinished later layers.
 | `0.1.0b1` | Published | Core API freeze and production-oriented examples |
 | `0.1.0rc1` | Published | External installation and release validation |
 | `0.1.0` | Published | First stable text and machine-readable release |
-| `0.2.0` | Next | Optional local machine learning identification for PII |
+| `0.2.0` | Published | Optional local machine learning identification for PII |
+| `0.3.0` | Next | Document inspection (PDF, DOCX, XLSX, PPTX) |
 
 Alpha releases optimize for the cleanest safe architecture, not backward compatibility. They may
 remove, rename, or replace public APIs without aliases or shims. Material changes are documented,

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/docs/policies.md

@@ -1,9 +1,10 @@
 # Policies
 
-`Policy.default()` enables structured identifiers plus person, organization, and location entities
-when a configured backend can detect them. `Policy.strict()` also enables secrets and URL
-credentials with a lower threshold. `Policy.llm()` enables all available entities.
-`Policy.financial()` limits detection to IBANs and payment cards.
+`Policy.default()` enables structured identifiers, secrets, URL credentials, and person,
+organization, and location entities when a configured backend can detect them.
+`Policy.strict()` enables the same entities with a lower confidence threshold. `Policy.llm()`
+enables all available entities. `Policy.financial()` limits detection to IBANs and payment
+cards.
 
 Paths are dot-separated segments. `*` matches one segment, so `messages.*.content` matches list
 indices. Exclusions win over inclusions. Dictionary keys are not transformed.

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/docs/releasing.md

@@ -29,10 +29,18 @@ After the first upload, verify that the pending publisher became an ordinary pro
    uv run ruff check .
    uv run mypy
    uv run pytest
-   uv run mkdocs build --strict
+   uv run python -m mkdocs build --strict
    uv build
    uv run twine check dist/*
    uv run python scripts/verify_release.py
+   
+   # Verify the built wheel locally in an isolated environment (simulating CI)
+   uv venv --python 3.14 .wheel-venv
+   uv pip install --python .wheel-venv/bin/python dist/*.whl
+   uv pip check --python .wheel-venv/bin/python
+   .wheel-venv/bin/python -I scripts/audit_install.py
+   .wheel-venv/bin/python -I scripts/smoke_wheel.py
+   .wheel-venv/bin/pseudonymize detectors
    ```
 
 4. Open a pull request and merge only after every required check passes.

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "pseudonymize"
-version = "0.2.0"
+version = "0.3.0"
 description = "Local-first PII pseudonymization for text, structured data, and LLM payloads."
 readme = "README.md"
 requires-python = ">=3.11"
@@ -43,12 +43,20 @@ ml = [
     "onnxruntime>=1.29.0",
     "tokenizers>=0.23.1",
 ]
+office = [
+    "openpyxl>=3.1.0",
+    "python-docx>=1.1.0",
+    "python-pptx>=1.0.0",
+]
+pdf = [
+    "pdfminer.six>=20240706",
+]
 
 [tool.hatch.build.targets.wheel]
 packages = ["src/pseudonymize"]
 
 [dependency-groups]
-test = ["hypothesis>=6", "pytest>=8", "pytest-benchmark>=5", "pytest-cov>=5"]
+test = ["hypothesis>=6", "pytest>=8", "pytest-benchmark>=5", "pytest-cov>=5", "fpdf2>=2.7.9"]
 quality = ["mypy>=1.11", "pre-commit>=4", "ruff>=0.9"]
 docs = ["mkdocs-material>=9", "mkdocstrings[python]>=1.0.6"]
 release = ["build>=1", "twine>=6"]
@@ -81,7 +89,11 @@ files = ["src", "tests", "scripts"]
 module = [
     "numpy.*",
     "onnxruntime.*",
-    "tokenizers.*"
+    "tokenizers.*",
+    "pdfminer.*",
+    "docx.*",
+    "pptx.*",
+    "openpyxl.*"
 ]
 ignore_missing_imports = true
 follow_imports = "skip"
@@ -91,7 +103,7 @@ minversion = "8"
 testpaths = ["tests"]
 pythonpath = ["."]
 python_files = ["test_*.py", "benchmark_*.py"]
-addopts = ["--strict-config", "--strict-markers", "-ra", "--cov=pseudonymize", "--cov-report=term-missing", "--cov-branch", "--cov-fail-under=99.36"]
+addopts = ["--strict-config", "--strict-markers", "-ra", "--cov=pseudonymize", "--cov-report=term-missing", "--cov-branch", "--cov-fail-under=99.36", "--basetemp=.pytest_temp"]
 markers = ["benchmark: performance benchmark", "integration: integration test", "slow: deliberately slow test"]
 filterwarnings = ["error"]
 

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/scripts/audit_install.py

@@ -27,11 +27,18 @@ def _blocked_network(*arguments: object, **keywords: object) -> None:
 
 def main() -> None:
     parser = argparse.ArgumentParser()
-    parser.add_argument("--version", required=True)
+    parser.add_argument("--version", required=False)
     arguments = parser.parse_args()
 
+    expected_version = arguments.version
+    if not expected_version:
+        import tomllib
+
+        with open("pyproject.toml", "rb") as stream:
+            expected_version = tomllib.load(stream)["project"]["version"]
+
     installed = distribution("pseudonymize")
-    if installed.version != arguments.version:
+    if installed.version != expected_version:
         raise RuntimeError("installed version does not match release")
     if installed.requires:
         required_deps = [req for req in installed.requires if "extra ==" not in req]

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/src/pseudonymize/backends/ml/onnx.py

@@ -25,6 +25,20 @@ else:
         ort = None
         Tokenizer = None
 
+_LABEL_SUFFIXES: tuple[tuple[tuple[str, ...], EntityType], ...] = (
+    # Standard CoNLL-03 suffixes plus Ai4Privacy fine-grained labels.
+    (("PER", "FIRSTNAME", "LASTNAME", "MIDDLENAME"), EntityType.PERSON),
+    (("ORG", "COMPANYNAME"), EntityType.ORGANIZATION),
+    (("LOC", "CITY", "STATE", "COUNTY", "STREET", "ZIPCODE"), EntityType.LOCATION),
+)
+
+
+def _entity_type_for(label: str) -> EntityType | None:
+    for suffixes, entity_type in _LABEL_SUFFIXES:
+        if label.endswith(suffixes):
+            return entity_type
+    return None
+
 
 class LocalONNXPIIBackend(DetectionBackend):
     def __init__(
@@ -111,55 +125,44 @@ class LocalONNXPIIBackend(DetectionBackend):
 
             predictions = np.argmax(logits, axis=-1)
 
-            detections = []
+            # Token predictions are merged into entity spans: subword continuations
+            # (zero gap) and same-type tokens separated by one whitespace character
+            # collapse into a single detection so that "John Smith" is one PERSON.
+            spans: list[tuple[EntityType, int, int]] = []
 
             for idx, label_id in enumerate(predictions):
-                if label_id == 0 or self._id2label is None:
-                    continue
-
-                label_str = self._id2label.get(label_id)
+                label_str = (self._id2label or {}).get(int(label_id))
                 if not label_str or label_str == "O":
                     continue
-
-                entity_type = None
-
-                # Standard CoNLL-03 tags
-                if label_str.endswith("PER"):
-                    entity_type = EntityType.PERSON
-                elif label_str.endswith("ORG"):
-                    entity_type = EntityType.ORGANIZATION
-                elif label_str.endswith("LOC"):
-                    entity_type = EntityType.LOCATION
-
-                # Ai4Privacy tags mapping
-                elif any(label_str.endswith(s) for s in ("FIRSTNAME", "LASTNAME", "MIDDLENAME")):
-                    entity_type = EntityType.PERSON
-                elif label_str.endswith("COMPANYNAME"):
-                    entity_type = EntityType.ORGANIZATION
-                elif any(
-                    label_str.endswith(s) for s in ("CITY", "STATE", "COUNTY", "STREET", "ZIPCODE")
-                ):
-                    entity_type = EntityType.LOCATION
-
-                if entity_type:
-                    start, end = encoding.offsets[idx]
-                    if start == 0 and end == 0 and idx not in (0, len(predictions) - 1):
-                        continue
-                    if start == end:
+                entity_type = _entity_type_for(label_str)
+                if entity_type is None:
+                    continue
+                start, end = encoding.offsets[idx]
+                if start >= end:
+                    continue
+                if spans:
+                    previous_type, previous_start, previous_end = spans[-1]
+                    gap = block.text[previous_end:start]
+                    if (
+                        entity_type is previous_type
+                        and previous_end <= start <= previous_end + 1
+                        and not gap.strip()
+                    ):
+                        spans[-1] = (previous_type, previous_start, end)
                         continue
-
-                    detections.append(
-                        Detection(
-                            entity_type=entity_type,
-                            start=start,
-                            end=end,
-                            confidence=0.99,
-                            backend=self.name,
-                            detector="onnx",
-                        )
-                    )
-
-            return tuple(detections)
+                spans.append((entity_type, start, end))
+
+            return tuple(
+                Detection(
+                    entity_type=entity_type,
+                    start=start,
+                    end=end,
+                    confidence=0.99,
+                    backend=self.name,
+                    detector="onnx",
+                )
+                for entity_type, start, end in spans
+            )
 
         except Exception as e:
             raise BackendExecutionError(f"ONNX PII inference failed: {e}") from e

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/src/pseudonymize/detectors/url.py

@@ -10,6 +10,11 @@ _SENSITIVE_QUERY_NAMES = frozenset(
 )
 
 
+def _authority_end(url: str, authority_start: int) -> int:
+    indexes = (url.find(character, authority_start) for character in "/?#")
+    return min((index for index in indexes if index != -1), default=len(url))
+
+
 @dataclass(frozen=True, slots=True)
 class UrlDetector:
     name: str = "url"
@@ -21,7 +26,8 @@ class UrlDetector:
             parsed = urllib.parse.urlsplit(url)
             if parsed.username is not None or parsed.password is not None:
                 authority_start = url.find("//") + 2
-                at = url.find("@", authority_start)
+                authority_end = _authority_end(url, authority_start)
+                at = url.rfind("@", authority_start, authority_end)
                 detections.append(
                     Detection(
                         EntityType.URL_CREDENTIAL,
@@ -34,8 +40,10 @@ class UrlDetector:
             query_start = url.find("?") + 1
             if query_start == 0:
                 continue
+            fragment_start = url.find("#", query_start)
+            query_end = fragment_start if fragment_start != -1 else len(url)
             cursor = query_start
-            for item in url[query_start:].split("&"):
+            for item in url[query_start:query_end].split("&"):
                 name, separator, value = item.partition("=")
                 if separator and urllib.parse.unquote_plus(name).lower() in _SENSITIVE_QUERY_NAMES:
                     value_start = cursor + len(name) + 1

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/src/pseudonymize/document.py

@@ -68,7 +68,43 @@ class CSVCellLocation:
             raise ValueError("CSV row and column indexes must be non-negative")
 
 
-SourceLocation: TypeAlias = TextOffsetLocation | JSONPathLocation | CSVCellLocation
+@dataclass(frozen=True, slots=True)
+class CoordinateLocation:
+    page: int
+    x0: float
+    y0: float
+    x1: float
+    y1: float
+
+    def __post_init__(self) -> None:
+        if not isinstance(self.page, int) or isinstance(self.page, bool) or self.page < 0:
+            raise ValueError("page must be a non-negative integer")
+        for val in (self.x0, self.y0, self.x1, self.y1):
+            if not isinstance(val, (int, float)) or isinstance(val, bool) or not isfinite(val):
+                raise TypeError("coordinates must be finite numbers")
+        if self.x0 > self.x1 or self.y0 > self.y1:
+            raise ValueError("coordinates must describe a valid bounding box")
+
+
+@dataclass(frozen=True, slots=True)
+class StructuralLocation:
+    path: tuple[str | int, ...]
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "path", tuple(self.path))
+        if any(not isinstance(part, (str, int)) or isinstance(part, bool) for part in self.path):
+            raise TypeError("structural path parts must be strings or integers")
+        if any(isinstance(part, int) and part < 0 for part in self.path):
+            raise ValueError("structural path indexes must be non-negative")
+
+
+SourceLocation: TypeAlias = (
+    TextOffsetLocation
+    | JSONPathLocation
+    | CSVCellLocation
+    | CoordinateLocation
+    | StructuralLocation
+)
 
 
 @dataclass(frozen=True, slots=True)
@@ -83,7 +119,16 @@ class ContentBlock:
             raise ValueError("block id must not be empty")
         if not isinstance(self.text, str):
             raise TypeError("block text must be a string")
-        if not isinstance(self.location, (TextOffsetLocation, JSONPathLocation, CSVCellLocation)):
+        if not isinstance(
+            self.location,
+            (
+                TextOffsetLocation,
+                JSONPathLocation,
+                CSVCellLocation,
+                CoordinateLocation,
+                StructuralLocation,
+            ),
+        ):
             raise TypeError("block location must be a supported source location")
         object.__setattr__(self, "metadata", _metadata(self.metadata))
 

{pseudonymize-0.2.0 → pseudonymize-0.3.0}/src/pseudonymize/engine.py

@@ -265,10 +265,15 @@ class Pseudonymizer:
             self.transformer.render(entity, alias)
             for entity, alias in zip(entities, aliases, strict=True)
         )
-        output = text
-        for entity, token in reversed(tuple(zip(entities, tokens, strict=True))):
+        segments: list[str] = []
+        cursor = 0
+        for entity, token in zip(entities, tokens, strict=True):
             detection = entity.detection
-            output = output[: detection.start] + token + output[detection.end :]
+            segments.append(text[cursor : detection.start])
+            segments.append(token)
+            cursor = detection.end
+        segments.append(text[cursor:])
+        output = "".join(segments)
         replacements = _replacement_reports(entities, tokens)
         reports.extend(_replacement_detection_reports(block, replacements))
         mapping = _mapping(text, entities, aliases, tokens) if include_mapping else None
@@ -287,10 +292,10 @@ class Pseudonymizer:
         if isinstance(data, str):
             block_id = f"block-{block_counter[0]:06d}"
             block_counter[0] += 1
-            block = ContentBlock(block_id, data, JSONPathLocation(path))
             if not self.policy.allows_path(tuple(str(part) for part in path)):
                 statistics.blocks_processed += 1
                 return data
+            block = ContentBlock(block_id, data, JSONPathLocation(path))
             return self._process_block(block, context, False, statistics, reports).text
         if data is None or isinstance(data, (bool, int, float)):
             return data
@@ -462,8 +467,20 @@ def _processing_adapters(
     output_adapter: OutputAdapter | None,
 ) -> tuple[InputAdapter[Path], OutputAdapter]:
     if input_adapter is None and output_adapter is None:
-        adapter = BuiltinFileAdapter(select_file_format(source, format), encoding)
-        return adapter, adapter
+        selected_format = select_file_format(source, format)
+        if selected_format == FileFormat.PDF:
+            from pseudonymize.inspection.pdf import PDFInspectionAdapter
+
+            pdf_adapter = PDFInspectionAdapter()
+            return pdf_adapter, pdf_adapter
+        elif selected_format in (FileFormat.DOCX, FileFormat.XLSX, FileFormat.PPTX):
+            from pseudonymize.inspection.office import OfficeInspectionAdapter
+
+            office_adapter = OfficeInspectionAdapter(selected_format)
+            return office_adapter, office_adapter
+        else:
+            builtin_adapter = BuiltinFileAdapter(selected_format, encoding)
+            return builtin_adapter, builtin_adapter
     if input_adapter is None or output_adapter is None:
         raise ValueError("custom file processing requires input and output adapters")
     if format is not None or encoding is not None:
@@ -478,7 +495,17 @@ def _inspection_adapter(
     input_adapter: InputAdapter[Path] | None,
 ) -> InputAdapter[Path]:
     if input_adapter is None:
-        return BuiltinFileAdapter(select_file_format(source, format), encoding)
+        selected_format = select_file_format(source, format)
+        if selected_format == FileFormat.PDF:
+            from pseudonymize.inspection.pdf import PDFInspectionAdapter
+
+            return PDFInspectionAdapter()
+        elif selected_format in (FileFormat.DOCX, FileFormat.XLSX, FileFormat.PPTX):
+            from pseudonymize.inspection.office import OfficeInspectionAdapter
+
+            return OfficeInspectionAdapter(selected_format)
+        else:
+            return BuiltinFileAdapter(selected_format, encoding)
     if format is not None or encoding is not None:
         raise ValueError("custom adapters cannot be combined with format or encoding")
     return input_adapter