visions

What's Changed

• Merge from develop by @ieaves in https://github.com/dylan-profiler/visions/pull/203

Full Changelog: https://github.com/dylan-profiler/visions/compare/v0.7.5...v0.7.6

What's Changed

• Adds python 3.13.0 compatibility
• removes imghdr requirement
• Improves Image type identification using puremagic mime type identification

Full Changelog: https://github.com/dylan-profiler/visions/compare/v0.7.5...v0.8.0

FIX: Numpy backend FIX: Compatibility fixes with new versions of pandas ENH: Documentation updates ENH: CI

FIX: bottleneck import removed (py3.9+ incompatible)

FEAT: Numpy and spark backends DEPS: bottleneck dependency removed (py3.9+ incompatible)

FIX: CI issues FIX: Unit tests with multiple pandas boolean types. FIX: Documentation changes

• public methods on typesets are now static
• introduces a new declarative API
• documentation changes and fixes.

ENH: swifter apply for pandas backend FIX: fix for issue #147 ENH: __version__ attribute made available ENH: improved typing and CI ENH: contrib types/typesets for a low-threshold contribution of types

ENH: Expose state using typeset.detect and typeset.infer ENH: plotting of typesets improved FIX: fix and test cases for #136 CLN: pre-commit with black, isort, pyupgrade, flake8 ENH: type relations are now accessible by type (e.g. Float.relations[Integer])

Improved default handling of sparse and empty series (+ testing)

• Introduce stateful type inference and casting
• Expose test utils to users and fix diagnostic information
• Integer consistency for the standard set
• Use pd.BooleanDtype for newer versions of pandas
• Latest black formatting

API breaking changes:

• migration to single dispatch on typeset methods
• updated API to unify detect / infer / cast against Series and DataFrames
• improvements to boolean type

Fixes big_o requirement regression.

Extended root node detection over relation graph & generic support for user customized root nodes.

Split text summary in summary of length and unicode

Support for Files and Images, rewritten summarization functions

• Renamed ExistingPath to File
• Renamed ImagePath to Image
• Version bump to 0.4.2
• Summaries: return series instead of dict
• Categorical: unicode counts now based on original character distribution instead of unique characters which are used as intermediate step for increased performance.
• Categorical: aggregate functions are included for string length (min, max, mean, median).
• Path: number of unique values for the path parts are returned
• Image: make Exif and Hash calculations optional. Also return width, height and area.
• File: in addition to the file_size, return creation, modification and access time (which were already returned).