Function get_sorted_text

src/utils.py:165–279

Extract plain text avoiding unacceptable line breaks. Text contained in clip will be sorted in reading sequence. Some effort is also spent to simulate layout vertically and horizontally. Args: page: pymupdf.Page clip: (rect-like) only consider text inside flags:

(
    page: pymupdf.Page,
    clip: rect_like = None,
    flags: OptInt = None,
    textpage: pymupdf.TextPage = None,
    tolerance=3,
)

def get_sorted_text(
    page: pymupdf.Page,
    clip: rect_like = None,
    flags: OptInt = None,
    textpage: pymupdf.TextPage = None,
    tolerance=3,
) -> str:
    """Extract plain text avoiding unacceptable line breaks.

    Text contained in clip will be sorted in reading sequence. Some effort
    is also spent to simulate layout vertically and horizontally.

    Args:
        page: pymupdf.Page
        clip: (rect-like) only consider text inside
        flags: (int) text extraction flags
        textpage: pymupdf.TextPage
        tolerance: (float) consider words to be on the same line if their top
            or bottom coordinates do not differ more than this.

    Notes:
        If a TextPage is provided, all text is checked for being inside clip
        with at least 50% of its bbox.
        This allows to use some "global" TextPage in conjunction with sub-
        selecting words in parts of the defined TextPage rectangle.

    Returns:
        A text string in reading sequence. Left indentation of each line,
        inter-line and inter-word distances strive to reflect the layout.
    """

    def line_text(clip, line):
        """Create the string of one text line.

        We are trying to simulate some horizontal layout here, too.

        Args:
            clip: (pymupdf.Rect) the area from which all text is being read.
            line: (list) word tuples (rect, text) contained in the line
        Returns:
            Text in this line. Generated from words in 'line'. Distance from
            predecessor is translated to multiple spaces, thus simulating
            text indentations and large horizontal distances.
        """
        line.sort(key=lambda w: w[0].x0)
        ltext = ""  # text in the line
        x1 = clip.x0  # end coordinate of ltext
        lrect = pymupdf.EMPTY_RECT()  # bbox of this line
        for r, t in line:
            lrect |= r  # update line bbox
            # convert distance to previous word to multiple spaces
            dist = max(
                int(round((r.x0 - x1) / r.width * len(t))),
                0 if (x1 == clip.x0 or r.x0 <= x1) else 1,
            )  # number of space characters

            ltext += " " * dist + t  # append word string
            x1 = r.x1  # update new end position