doccano/backend/data_export/pipeline/formatters.py

"""
Convert a dataset to the specified format.
"""
import abc

import pandas as pd

from data_export.models import DATA


class Formatter(abc.ABC):
    def __init__(self, target_column: str = "labels", **kwargs):
        self.target_column = target_column
        self.mapper = kwargs

    def format(self, dataset: pd.DataFrame) -> pd.DataFrame:
        if self.target_column not in dataset.columns:
            return dataset
        return self.apply(dataset)

    @abc.abstractmethod
    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        raise NotImplementedError("Please implement this method in the subclass.")


class JoinedCategoryFormatter(Formatter):
    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        """Format the label column to `LabelA#LabelB` format."""
        dataset[self.target_column] = dataset[self.target_column].apply(
            lambda labels: "#".join(sorted(label.to_string() for label in labels))
        )
        return dataset


class ListedCategoryFormatter(Formatter):
    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        """Format the label column to `['LabelA', 'LabelB']` format."""
        dataset[self.target_column] = dataset[self.target_column].apply(
            lambda labels: sorted([label.to_string() for label in labels])
        )
        return dataset


class FastTextCategoryFormatter(Formatter):
    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        """Format the label column to `__label__LabelA __label__LabelB` format.
        Also, drop the columns except for `data` and `self.target_column`.
        """
        dataset = dataset[[DATA, self.target_column, "Comments"]]
        dataset[self.target_column] = dataset[self.target_column].apply(
            lambda labels: " ".join(sorted(f"__label__{label.to_string()}" for label in labels))
        )
        dataset[self.target_column] = dataset[self.target_column].fillna("")
        dataset["Comments"] = dataset["Comments"].apply(
            lambda comments: " ".join(f"__comment__{comment.to_string()}" for comment in comments)
        )
        dataset = dataset[self.target_column] + " " + dataset[DATA] + " " + dataset["Comments"]
        return dataset


class TupledSpanFormatter(Formatter):
    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        """Format the span column to `(start_offset, end_offset, label)` format"""
        dataset[self.target_column] = dataset[self.target_column].apply(
            lambda spans: sorted(span.to_tuple() for span in spans)
        )
        return dataset


class DictFormatter(Formatter):
    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        """Format the column to `{key: value}` format"""
        dataset[self.target_column] = dataset[self.target_column].apply(
            lambda labels: [label.to_dict() for label in labels]
        )
        return dataset


class RenameFormatter(Formatter):
    def format(self, dataset: pd.DataFrame) -> pd.DataFrame:
        return self.apply(dataset)

    def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
        """Rename columns"""
        dataset.rename(columns=self.mapper, inplace=True)
        return dataset