Class WordExtractor

src/table.py:587–799

class WordExtractor:
    def __init__(
        self,
        x_tolerance=DEFAULT_X_TOLERANCE,
        y_tolerance=DEFAULT_Y_TOLERANCE,
        keep_blank_chars: bool = False,
        use_text_flow=False,
        horizontal_ltr=True,  # Should words be read left-to-right?
        vertical_ttb=False,  # Should vertical words be read top-to-bottom?
        extra_attrs=None,
        split_at_punctuation=False,
        expand_ligatures=True,
    ):
        self.x_tolerance = x_tolerance
        self.y_tolerance = y_tolerance
        self.keep_blank_chars = keep_blank_chars
        self.use_text_flow = use_text_flow
        self.horizontal_ltr = horizontal_ltr
        self.vertical_ttb = vertical_ttb
        self.extra_attrs = [] if extra_attrs is None else extra_attrs

        # Note: string.punctuation = '!"#$%&\'()*+,-./:;<=>?@[\\]^_`{|}~'
        self.split_at_punctuation = (
            string.punctuation
            if split_at_punctuation is True
            else (split_at_punctuation or "")
        )

        self.expansions = LIGATURES if expand_ligatures else {}

    def merge_chars(self, ordered_chars: list):
        x0, top, x1, bottom = objects_to_bbox(ordered_chars)
        doctop_adj = ordered_chars[0]["doctop"] - ordered_chars[0]["top"]
        upright = ordered_chars[0]["upright"]
        direction = 1 if (self.horizontal_ltr if upright else self.vertical_ttb) else -1

        matrix = ordered_chars[0]["matrix"]

        rotation = 0
        if not upright and matrix[1] < 0:
            ordered_chars = reversed(ordered_chars)
            rotation = 270

        if matrix[0] < 0 and matrix[3] < 0:
            rotation = 180
        elif matrix[1] > 0:
            rotation = 90

        word = {
            "text": "".join(
                self.expansions.get(c["text"], c["text"]) for c in ordered_chars
            ),
            "x0": x0,
            "x1": x1,
            "top": top,
            "doctop": top + doctop_adj,
            "bottom": bottom,
            "upright": upright,