How to extract only highlighted text?

[FRIKIdelTO]

I need to extract only highlighted text from a certain PDF file but I can't with this library.
I can do this with other libraries but I like to do it with this because I have better results.
Is there any way to do that?
Thanks

[jsvine]

Hi @FRIKIdelTO, and thanks for your interest in pdfplumber. In general, (digital) highlights are represented in PDFs via annotations. You can access annotations via page.annots. The specific representation of the annotation will likely differ based on the tool that created the highlights, but should contain useful positional information you can use in combination with page.crop(...).extract_text() to identify the highlighted text.

[FRIKIdelTO]

The box containing the highlighted text is extracted incorrectly on the Y-axis.
Code example to test it:

import pdfplumber
import fitz

pdf_path = "documento.pdf"

print('pdfplumber rects:')
with pdfplumber.open(pdf_path) as pdf:
    for page in pdf.pages:
        for annot in page.annots:
            if "Highlight" in str(annot["data"]["Subtype"]):
                print(annot["data"]["Rect"])

print()

print('PyMuPDF rects:')
with fitz.open(pdf_path) as pdf:
    for n in range(len(pdf)):
        page = pdf.load_page(n)
        for annot in page.annots():
            if annot.type[0] == 8:
                print(annot.rect)

example output:

pdfplumber rects:
[27.7573, 302.109, 784.988, 339.622]
[27.7573, 216.357, 784.988, 253.798]
[697.309, 436.54, 788.0, 462.285]

PyMuPDF rects:
Rect(27.757299423217773, 255.65399169921875, 784.9879760742188, 293.1669921875)
Rect(27.757299423217773, 341.4779968261719, 784.9879760742188, 378.91900634765625)
Rect(697.3090209960938, 132.99099731445312, 788.0, 158.73599243164062)

If I use PyMuPDF rects it works fine, but with pdfplumber rects extracts wrong text

[jsvine]

Thanks for the additional information and context. The values within the ["data"] sub-dictionary are not modified from the raw values provided via pdfminer.six. Those unmodified positional values are based on a bottom-left 0,0-origin, rather than top-left (which is the origin used for pdfplumber's chars, etc.). For positional information using the top-left origin, you'll want to check out the annotation's main positional keys: x0, top, x1, and bottom.