8.0.1

8.0.1 - extra import of numpy is removed

8.0.0

With the introduction of threshold_to_diff_deeper, the values returned are different than in previous versions of DeepDiff. You can still get the older values by setting threshold_to_diff_deeper=0. However to signify that enough has changed in this release that the users need to update the parameters passed to DeepDiff, we will be doing a major version update.

• use_enum_value=True makes it so when diffing enum, we use the enum's value. It makes it so comparing an enum to a string or any other value is not reported as a type change.
• threshold_to_diff_deeper=float is a number between 0 and 1. When comparing dictionaries that have a small intersection of keys, we will report the dictionary as a new_value instead of reporting individual keys changed. If you set it to zero, you get the same results as DeepDiff 7.0.1 and earlier, which means this feature is disabled. The new default is 0.33 which means if less that one third of keys between dictionaries intersect, report it as a new object.
• Deprecated ordered-set and switched to orderly-set. The ordered-set package was not being maintained anymore and starting Python 3.6, there were better options for sets that ordered. I forked one of the new implementations, modified it, and published it as orderly-set.
• Added use_log_scale:bool and log_scale_similarity_threshold:float. They can be used to ignore small changes in numbers by comparing their differences in logarithmic space. This is different than ignoring the difference based on significant digits.
• json serialization of reversed lists.
• Fix for iterable moved items when iterable_compare_func is used.
• Pandas and Polars support.

7.0.1

• When verbose=2, return new_path when the path and new_path are different (for example when ignore_order=True and the index of items have changed).
• Dropping support for Python 3.7
• Introducing serialize to flat rows for delta objects.
• fixes the issue with hashing datetime.date objects where it treated them as numbers instead of dates (fixes #445).
• upgrading orjson to the latest version
• Fix for bug when diffing two lists with ignore_order and providing compare_func
• Fixes #438
• Supporting Python 3.12 in the build process by Leo Sin
• Fixes #457 by sf-tcalhoun
• Fixes #441
• Fixes #450
• Fixes #443
• Include type info and change the "unknown" value for flat rows to something that is friendly for Postgres enums

6.7.1

• v6-7-1
  • Support for subtracting delta objects when iterable_compare_func is used.
  • Better handling of force adding a delta to an object.
  • Fix for Can't compare dicts with both single and double quotes in keys
  • Updated docs for Inconsistent Behavior with math_epsilon and ignore_order = True
• v6-7-0
  • Delta can be subtracted from other objects now.
  • verify_symmetry is deprecated. Use bidirectional instead.
  • always_include_values flag in Delta can be enabled to include values in the delta for every change.
  • Fix for Delta.add breaks with esoteric dict keys.
  • You can load a delta from the list of flat dictionaries.

6.6.0

• Numpy 2.0 support
• Adding Delta.to_flat_dicts

6.4.1

Bugfix: Numpy should be optional

6.4.0

• v6-4-0
  • Add Ignore List Order Option to DeepHash by
    Bobby Morck
  • pyyaml to 6.0.1 to fix cython build problems by Robert Bo Davis
  • Precompiled regex simple diff by cohml
  • New flag: zip_ordered_iterables for forcing iterable items to be compared one by one.

6.3.1

• Bugfix deephash for paths by maggelus
• Bugfix deephash compiled regex maggelus
• Fix tests dependent on toml by martin-kokos
• Bugfix for include_paths for nested dictionaries by kor4ik
• Use tomli and tomli-w for dealing with tomli files by martin-kokos
• Bugfix for datetime.date by Alex Sauer-Budge

6.3.0

• PrefixOrSuffixOperator: This operator will skip strings that are suffix or prefix of each other.
• include_obj_callback and include_obj_callback_strict are added by Håvard Thom.
• Fixed a corner case where numpy's np.float32 nans are not ignored when using ignore_nan_equality by Noam Gottlieb
• orjson becomes optional again.
• Fix for ignore_type_in_groups with numeric values so it does not report number changes when the number types are different.

6.2.1

Major improvement in the diff report for lists when items are all hashable and the order of items is important.