Release notes

0.6.4 (Dec 6, 2024)

Fixes

• Fix with_regex_replacement_table mutator not behaving correctly when pattern only matches partly

0.6.3 (Dec 4, 2024)

Features

• Improve randomized selection of replacements in mutators using replacement tables

Fixes

• Fix phonetic replacement rules not being matched correctly against original data when the desired pattern occurs in multiple places

0.6.2 (Nov 27, 2024)

Features

• Add placeholder option to with_generator for inserting generated values

0.6.1 (Nov 15, 2024)

Fixes

• Fix indexing behavior in dfbitlookup when using NumPy data types

0.6.0 (Nov 15, 2024)

Breaking changes

• Change mutator type definition from Callable[[list[pd.Series]], list[pd.Series]] to Callable[[list[pd.Series], Optional[float]], list[pd.Series]] to delegate the selection of rows to mutate to the mutators themselves
• Generator and Mutator type definitions are now exported at the top level of the module
• Replace D option in favor of d for unit parameter in with_datetime_offset
• Remove strategy parameter from with_missing_value
• Remove rng parameter from mutate_data_frame
• Remove with_edit in favor of with_group

Features

• with_replacement_table, with_regex_replacement_table and with_phonetic_replacement_table now favor rare replacements over common ones
• Add rng parameter to with_function, with_lowercase, with_missing_value, with_noop, with_repeat, with_uppercase
• with_permute now permutes series contents in a way that values are guaranteed to not remain in their original series
• Add days, hours, minutes and seconds to list of permitted unit values for with_datetime_offset
• Add list[str] as option to charset parameter of with_cldr_keymap_file, with_insert and with_substitute

Fixes

• When providing a list of mutators to a column in mutate_data_frame, all mutators are now applied to all rows instead of with a 1 / mutator_count probability
• Fix with_regex_replacement_table interpreting numbers in pattern and substitution columns as belonging to a named capture group

Documentation

• Use section-style navigation instead of tabs in Gecko docs

0.5.2 (Nov 5, 2024)

Features

• Add generator.with_group for grouping multiple (weighted) generators together

Internal

• Remove automated benchmarks

0.5.1 (Oct 30, 2024)

Features

• Add the option to use data frames for all generators and mutators that accept paths to CSV files

0.5.0 (Oct 23, 2024)

Breaking changes

• to_data_frame has a new call signature that ensures that it's consistent with mutate_data_frame

df_generated = generator.to_data_frame(
    [
        (("fruit", "type"), generator.from_multicolumn_frequency_table(
            "fruit-types.csv",
            value_columns=["fruit", "type"],
            freq_column="count",
            rng=rng,
        )),
        ("weight", generator.from_uniform_distribution(
            low=20,
            high=100,
            rng=rng,
        )),
    ], 
    10_000
)

Features

• Add mutator.with_group for grouping multiple mutators together
• Add support for Python 3.13

Documentation

• Fix creation and modification timestamps in documentation

0.4.2 (Sep 20, 2024)

Fixes

• Fix NaNs produced by generators and mutators that take in CSV files with empty cells

0.4.1 (Sep 12, 2024)

Features

• Add inline and reverse flags to with_replacement_table mutator

0.4.0 (Sep 10, 2024)

Breaking changes

• mutate_data_frame has a new call signature which ensures the order of mutation operations

df_mutated = mutator.mutate_data_frame(
    df_original,
    [
        ("gender", (0.1, mutator.with_categorical_values(
            "./gender.csv",
            value_column="gender",
            rng=rng
        ))),
        (("given_name", "last_name"), (0.05, mutator.with_permute())),
        ("postcode", [
            mutator.with_delete(rng=rng),
            mutator.with_substitute(charset="0123456789", rng=rng)
        ])
    ],
    rng=rng
)

Features

• Add generator.from_datetime_range for generating dates and times
• Add mutator.with_lowercase and mutator.with_uppercase for case conversions
• Add mutator.with_datetime_offset for applying arbitrary offsets to dates and times
• Add mutator.with_generator for appending, prepending or replacing data in a series with values from a generator
• Add mutator.with_regex_replacement_table for regex-based substitutions
• Add mutator.with_repeat for repeated values

Fixes

• Fix mutate_data_frame raising an error if probability is provided as an integer, not a float

0.3.2 (Jul 19, 2024)

Fixes

• Fix multiple mutators not being applied correctly when defined on the same column

0.3.1 (Mar 28, 2024)

Fixes

• Fix IndexError when calling with_permute on empty series
• Fix Python version range in pyproject.toml

Refactors

• Fix type hint on **kwargs in benchmarks

Documentation

• Add navigation tabs to documentation
• Fix image link in README so that it can be displayed on PyPI

Internal

• Cache dependencies in CI pipelines
• Reorganize dependencies into groups for tests, development and documentation

0.3.0 (Mar 18, 2024)

Features

• Allow corruptor.with_permute to work with more than two series at once
• Infer header parameter for functions reading CSV files
• Remove list length constraints from mutator module

Refactors

• Fix type hints on *args and **kwargs
• Rename corruptor module to mutator
• Rename Corruptor type alias to Mutator
• Rename corruptor.corrupt_dataframe to mutator.mutate_data_frame
• Rename generator.to_dataframe to generator.to_dataframe

Documentation

• Add API reference to documentation
• Update docs to use new "mutator" terminology wherever possible
• Use Google format docstrings instead of reST

Internal

• Merge documentation repository into main repository
• Move repositories from GitLab to GitHub
• Refine benchmark suite, add example based on German population dataset

0.2.0 (Feb 16, 2024)

Features

• Add generator.with_permute for swapping values between series
• Set wider version ranges for dependencies
• Fix corruptor.corrupt_dataframe to not modify original data frame
• Add tests to all corruptor functions to ensure no modifications to original data

Refactors

• Change generators to take in and return a list of series instead of single series
• Change generator.to_dataframe signature to align with corruptor.corrupt_dataframe

Internal

• Extend CI pipeline with a benchmarking step that runs on release and when manually triggered
• Add benchmark based on the "fruits" example in the docs

0.1.0 (Feb 8, 2024)

• Initial release