Looking for help in inserting a rotated HTML box

[XHJ-TS9527]

Hi,
I'm currently trying to put some text on an image. The text includes characters in different languages, so I have to use different fonts to display them correctly. It seems that the insert_htmlbox function can meet my requirements, so I'm using this function to put the text on the image. However, the document says that the rotation angle should be a multiple of 90°, which is very inflexible. Therefore, I try to use the following solution to fulfill my requirements in putting text on an image with an arbitrary rotation angle:

1. Open the image and turn it into a PDF file

import fitz
image = fitz.open(imagePath)
imageBytes = image.convert_to_pdf()
document = fitz.open(stream=documentBytes)
image.close()
imagePage = document[0]
fontArchive = fitz.Archive(fontPath)    # fontPath refers to the directory containing font files

2. Establish a new PDF document instance

draftDocument = fitz.Document()

3. Put each text clip on the image

for textClip in textClips:
    text = textClip['text']    # textClip['text'] is the HTML string
    textCss = textClip['textCss']    # textClip['textCss'] is the CSS string
    textBox = fitz.Rect(*textClip['box'])    # textClip['box'] is a tuple with four numbers, with (0, 0) as the left-up point
    draftDocument.insert_page(len(draftDocument), width=textBox.width, height=textBox.height)
    draftPage = draftDocument[-1]
    draftPage.insert_htmlbox(textBox, text, css=textCss, archive=fontArchive)

4. Calculate the zone to paste the rotated draft page

    import numpy as np
    rotationAngle = textClip['angle']    # rotation angle in degrees, and the rotation direction is counterclockwise
    textCenterPoint = np.array(textClip['centerPoint'])    # textClip['centerPoint'] is a tuple with two numbers indicating the exact location of the center point of the text box
    roationAngleRad = np.deg2rad(rotationAngle)
    rotationMatrix = np.array([[np.cos(rotationAngleRad), -np.sin(rotationAngleRad)], 
                               [np.sin(rotationAngleRad), np.cos(rotationAngleRad)]])
    originalPoints = np.array(textClip['box']) - textCenterPoint    # originalPoints will be a 4×2 ndarray
    rotatedPoints = (rotationMatrix @ originalPoints.T).T + textCenterPoint    # rotatedPoints will indicate the outmost boundary of the text box
    rotatedPointXs = rotatedPoints[:, 0].tolist()
    rotatedPointYs = rotatedPoints[:, 1].tolist()
    leftBoundary = min(rotatedPointXs)
    upBoundary = min(rotatedPointYs)
    rightBoundary = max(rotatedPointXs)
    downBoundary = max(rotatedPointYs)
    pasteZoneRect = fitz.Rect(leftBoundary, upBoundary, rightBoundary, downBoundary)

5. Copy the page from the draft document to the image document with rotation

    imagePage.show_pdf_page(pasteZoneRect, draftDocument, len(draftDocument) - 1, rotate=rotationAngle)

6. Turn the page into an image and save it

from PIL import Image
image = Image.open(imagePath)
image = imagePage.get_pixmap(dpi=max(image.info['dpi']))
image.save(imagePath)

I think this approach should work. But when I run the program, I get an exception thrown by pymupdf.mupdf.FzErrorArgument. The exception information is code=4: source object number out of range. I don't know how this exception occurs, as I have put text on draftPage using the provided insert_htmlbox function. Can anyone give me some suggestions? Thanks in advance.

[XHJ-TS9527]

The problem can be solved if I initialize a new Document instance in each iteration and apply the insert_htmlbox method to the new page inserted into the new document. I think the only difference is the parent Document instance of Page instances, but I have no idea how this difference causes the problem.

A temporary solution:

import fitz
from PIL import Image
import numpy as np
# Step 1: Open the image and turn it into a PDF file
image = fitz.open(imagePath)
imageBytes = image.convert_to_pdf()
document = fitz.open(stream=documentBytes)
image.close()
imagePage = document[0]
fontArchive = fitz.Archive(fontPath)    # fontPath refers to the directory containing font files
# Step 2: Put each text clip on the image
for textClip in textClips:
    text = textClip['text']    # textClip['text'] is the HTML string
    textCss = textClip['textCss']    # textClip['textCss'] is the CSS string
    textBox = fitz.Rect(*textClip['box'])    # textClip['box'] is a tuple with four numbers, with (0, 0) as the left-up point
    draftDocument = fitz.Document()
    draftDocument.insert_page(len(draftDocument), width=textBox.width, height=textBox.height)
    draftPage = draftDocument[-1]
    draftPage.insert_htmlbox(textBox, text, css=textCss, archive=fontArchive)
    # Step 3: Calculate the zone to paste the rotated draft page
    rotationAngle = textClip['angle']    # rotation angle in degrees, and the rotation direction is counterclockwise
    textCenterPoint = np.array(textClip['centerPoint'])    # textClip['centerPoint'] is a tuple with two numbers indicating the exact location of the center point of the text box
    roationAngleRad = np.deg2rad(rotationAngle)
    rotationMatrix = np.array([[np.cos(rotationAngleRad), -np.sin(rotationAngleRad)], 
                               [np.sin(rotationAngleRad), np.cos(rotationAngleRad)]])
    originalPoints = np.array(textClip['box']) - textCenterPoint    # originalPoints will be a 4×2 ndarray
    rotatedPoints = (rotationMatrix @ originalPoints.T).T + textCenterPoint    # rotatedPoints will indicate the outmost boundary of the text box
    rotatedPointXs = rotatedPoints[:, 0].tolist()
    rotatedPointYs = rotatedPoints[:, 1].tolist()
    leftBoundary = min(rotatedPointXs)
    upBoundary = min(rotatedPointYs)
    rightBoundary = max(rotatedPointXs)
    downBoundary = max(rotatedPointYs)
    pasteZoneRect = fitz.Rect(leftBoundary, upBoundary, rightBoundary, downBoundary)
    # Step 4: Copy the page from the draft document to the image document with rotation
    imagePage.show_pdf_page(pasteZoneRect, draftDocument, len(draftDocument) - 1, rotate=rotationAngle)
    draftDocument.close()
# Step 5: Turn the page into an image and save it
image = Image.open(imagePath)
image = imagePage.get_pixmap(dpi=max(image.info['dpi']))
image.save(imagePath)