Conversation
Member
seperman
commented
Mar 11, 2026
seperman
commented
- Adding support for python 3.14
If `key` is not in result, then there's no point in trying to delete it anyway
This reverts commit eec4a54.
Fix logarithmic_similarity return type hint
When comparing two lists where all items are removed (t2 becomes an empty list), the colored view would display an empty list [] instead of showing the removed items in red. The bug occurred because `_colorize_json()` returned '[]' immediately if the list was empty, without checking for items that were removed and should be displayed. Fix: Check for removed items before returning empty list representation. Added tests: - test_colored_view_list_all_items_removed - test_colored_compact_view_list_all_items_removed
Update lambda function to check for attribute existence in __slots__.
Add tests for DeepHash with classes using __slots__
…e when serializing integers larger than 9223372036854775807 (2^63-1) or smaller than -9223372036854775808 (-2^63). This is an orjson limitation since it only supports signed 64-bit integers natively. Fix in deepdiff/serialization.py: 1. Added _convert_oversized_ints() helper (line ~739) that recursively walks a data structure and converts any integer outside the signed 64-bit range to a string. 2. Wrapped the orjson.dumps() call in a try/except — on the specific "Integer exceeds 64-bit range" TypeError, the data is pre-processed through _convert_oversized_ints() and retried. Tests added in tests/test_serialization.py: - test_json_dumps_large_int_exceeding_64bit — basic case with a large positive int - test_json_dumps_large_int_in_nested_structure — large int nested inside dicts - test_json_dumps_large_negative_int — large negative int
… calling type(obj)(converted_list) to reconstruct them. This fails because NamedTuple constructors require positional arguments matching their field definitions, not a single list. Solution: Before reconstructing a tuple subclass, check for the _fields attribute (the NamedTuple marker). If present, reconstruct via type(obj)(**dict(zip(obj._fields, converted))) so each field is passed by keyword. Plain tuples and lists continue to use the original path.
… it returned a dict containing opaque DiffLevel objects rather than a readable Python dictionary. The workaround was to_dict(view_override='text'), but that was locked to whatever verbose_level was set at init time (default 1), meaning you couldn't get the full detail that the tree view had access to. Change: Replaced the view_override parameter with verbose_level on both to_dict() and to_json(). These methods now always produce a text-view dictionary — an actual usable Python dict. The default verbosity is context-aware: - If the original view is 'text': uses the verbose_level from initialization (preserving existing behavior). - If the original view is 'tree': defaults to verbose_level=2 (the most detailed output), since tree-view users chose that view for its richness. In both cases, you can override the verbosity explicitly: to_dict(verbose_level=0), to_dict(verbose_level=1), etc.
- pyproject.toml: requires-python changed to >=3.10, removed 3.9 classifier, cleaned up all python_version < '3.10' / python_version >= '3.10' conditional dependencies (coverage, click, numpy, orjson, polars, pytest, pytest-cov, python-dotenv) - .github/workflows/main.yaml: Removed 3.9 from matrix, added 3.14, changed DEFAULT_PYTHON and all conditional checks from 3.12 to 3.14 - noxfile.py: Removed 3.9, added 3.14 to pytest session - README.md: Updated to "Python 3.10+" and added "Dropping support for Python 3.9" to What is New - CLAUDE.md: Updated to "3.10+" - .github/ISSUE_TEMPLATE/bug_report.md: Updated example version from 3.9.12 to 3.10.12 Changelog/docs updates: - CHANGELOG.md: Added "Dropping support for Python 3.9" to v8-7-0 - docs/changelog.rst: Added "Dropping support for Python 3.9" to v8-7-0 - docs/index.rst: Added new 8-7-0 section with all its entries (matching the pattern of existing version sections)
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #579 +/- ##
==========================================
- Coverage 96.10% 96.06% -0.04%
==========================================
Files 16 16
Lines 4488 4474 -14
==========================================
- Hits 4313 4298 -15
- Misses 175 176 +1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
Fix typo in README
Fix ignore_keys issue in detailed__dict__
Add support for callable group_by
…t-display Fix: Colored view displays empty list [] when all items are removed
…w's fields, with path and action required and the rest optional. 2. Changed the return type of to_flat_dicts from List[FlatDeltaRow] to List[FlatDeltaDict]. 3. Used cast to tell pyright the _asdict() output conforms to FlatDeltaDict. 4. Fixed the parameter bug — the original code was hardcoding include_action_in_path=False, report_type_changes=True instead of passing through the actual arguments.
…anges: 1. Hash cache collision (deephash.py): The shared self.hashes dict used raw objects as keys. Since Python treats 1 == 1.0 and hash(1) == hash(1.0), looking up 1.0 returned the cached hash for 1, silently hiding the type difference. 2. Distance over-counting (distance.py): _get_item_length counted both old_type and old_value in type_changes entries, double-counting a single operation. This inflated distances, causing the pairing algorithm to reject items that should have been matched. Fixes: - deepdiff/deephash.py: Added _make_hash_key() / _make_hash_key_for_lookup() that wrap numeric objects as (type(obj), obj) tuples when ignore_numeric_type_changes=False. Applied at cache lookup, cache store, and all public accessor methods (__getitem__, __contains__, get, get_key). Added _unwrap_hash_key() for iteration methods so the public API still shows raw objects. - deepdiff/distance.py: Skip old_type and old_value keys in _get_item_length — they're metadata about prior state, not additional operations. Also added ignore_numeric_type_changes to DistanceProtocol and threaded it through get_key calls. - Test updates: Updated expected distance values in test_distance.py to reflect the corrected distance formula.
Ensure that delta replays apply changes to correct t1/t2 based on usage of iterable_compare_func
…ersion mechanism in model.py DiffLevel.path so future developers understand why the keys are swapped. 2. Added test_moved_and_type_changed — covers the type_changes code path where items both move and change type (str -> int), which the original tests didn't exercise.
Fix bug in DeepHash for classes with uninitialised slots
…tr() swallows all AttributeErrors — including those from __getattr__. This made objects with a broken __getattr__ (like the Bad test class) appear to have zero accessible slots rather than failing the strategy. The fix uses object.__getattribute__() to check each slot directly, which bypasses __getattr__ entirely. For classes that do define __getattr__, if a slot isn't initialized via the descriptor, we fall back to the normal getattr() — letting __getattr__ either provide a value or raise (propagating the failure). This correctly handles: - Uninitialized slots (no __getattr__): skipped gracefully, empty dict is valid - Partially initialized slots: only initialized ones included - Broken __getattr__: error propagates, object becomes unprocessed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
8 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.