Implement TIFF Predictor 2

[helpmefindaname]

Currently, the TIFF Predictor 2 is not implemented, so you cannot read the image stream out of the PDF attached below.

The tiff predictor 2 is specified in the TIFF Revision 6.0 (Section 14, page 64) and referenced as part of the PDF reference 1.7 (Section 3.3, page 75)

In my pdf-corpus I have so far only encountered the TIFF Predictor with exactly 8 bitspercomponent. For that the following code represents a valid implementation:

def apply_tiff_predictor(colors: int, columns: int, bitspercomponent: int, data: bytes) -> bytes:
    if bitspercomponent != 8:
        error_msg = f"Unsupported `bitspercomponent': {bitspercomponent}"
        raise ValueError(error_msg)
    bpp = colors * (bitspercomponent // 8)
    nbytes = columns * bpp
    buf: list[int] = []
    for scanline_i in range(0, len(data), nbytes + 1):
        raw: list[int] = []
        for i in range(nbytes):
            new_value = data[scanline_i + i]
            if i >= bpp:
                new_value += raw[i - bpp]
                new_value %= 256
            raw.append(new_value)
        buf.extend(raw)

    return bytes(buf)

test_pdf_with_tiff_predictor.pdf

[dhdaines]

Fixed in PLAYA-PDF: dhdaines/playa#18 (but it may be less useful to you as it doesn't save images at the moment)