How to extract texts between two coordinates in a page?

[StephenZKCurry]

I want to extract texts between two coordinates on a page use the PDF's underlying flow of characters as a guide for ordering and segmenting the words, rather than presorting the characters by x/y position, mimics dragging a cursor highlights text in a PDF, How can I do that?

[JorjMcKie]

You can supply an arbitrary rectangle ("clip") inside which your desired text lives. If you only have top and bottom values, make a rectangle clip = pymupdf.Rect(0, top, page.rect.width, bottom).
Then execute text = page.get_text(sort=True, clip=clip).
This will (pymupdf v1.24.11+) extract the text in reading order.

[patsnap-liujin]

clip = pymupdf.Rect(0, top, page.rect.width, bottom)
text = page.get_text(sort=True, clip=clip).
This will (pymupdf v1.24.11+) extract the text in reading order.

Thanks for the answer! @JorjMcKie I have some further questions😄. The page.get_text get the text intersecting the rect, but is there some encapsulated method or function to get the text fully contained in the rect ? I have read textbox-extract-1.py and code an example below for my useage, and I'm looking for some practice more efficient and standard. Can you give some suggestions?

# page = doc[page_num]
# my_rect = fitz.Rect(bbox)
current_textpage = page.get_textpage()
words = page.get_text("words",clip=my_rect, textpage=current_textpage) # here is the point: when using param `textpage` to reduce execution time significantly, the param clip is ignored and the result `words` contains whole page but I want the content only in my_rect.
fullly_containded_words = [w for w in words if fitz.Rect(w[:4]) in my_rect] # if I have many rects, the loop will  execute many times
fullly_containded_text = make_text(fullly_containded_words) # make_text is a function from textbox-extract-1.py

[patsnap-liujin]

After read the source code about Page.get_textbbox, I find the function JM_rects_overlap is the judgment of overlap. I change it from

def JM_rects_overlap(a, b):
    if (0
            or a.x0 >= b.x1
            or a.y0 >= b.y1
            or a.x1 <= b.x0
            or a.y1 <= b.y0
            ):
        return 0
    return 1

to

def JM_rects_overlap(a, b):
    if (a.x0 < (b.x1+b.x0)/2 < a.x1
        and a.y0 < (b.y1+b.y0)/2 < a.y1):
        return 1
    return 0

and it works as I want. The change is judge from center point ,not edge point.