generate_annotated_diffs.py

import diff_match_patch as dmp_module

import hf_data_loader


def get_annotated_diff(start_text, end_text):
    dmp = dmp_module.diff_match_patch()
    dmp_mapping = {
        -1: '-',
        0: None,
        1: '+'
    }

    diff = dmp.diff_main(start_text, end_text)
    dmp.diff_cleanupSemantic(diff)

    result = [[w, dmp_mapping[t]] for t, w in diff]

    return result


def annotated_diff_for_row_manual_df(row):
    start = row['commit_msg_start']
    end = row['commit_msg_end']
    return get_annotated_diff(start, end)


def annotated_diff_for_row_synthetic_df(row):
    start = row['initial_msg_pred']
    end = row['reference']
    return get_annotated_diff(start, end)


def manual_data_with_annotated_diffs():
    df = hf_data_loader.load_raw_rewriting_dataset_as_pandas()
    annotated = df.apply(annotated_diff_for_row_manual_df, axis=1)
    df['annotated_diff'] = annotated
    return df


def synthetic_data_with_annotated_diffs():
    df = hf_data_loader.load_synthetic_dataset_as_pandas()
    annotated = df.apply(annotated_diff_for_row_synthetic_df, axis=1)
    df['annotated_diff'] = annotated
    return df