Question: how is word's origin computed?

[VeryLazyBoy]

I notice that there is a origin attribute in the TextPage content, but when I change the MediaBox and CropBox of the pdf file, the origin becomes very weird.

For example, the following code:

text = '123'

doc = fitz.Document()
page = doc.newPage()
page.setMediaBox(fitz.Rect(100, 50, 500, 500))
page.setCropBox(fitz.Rect(0, 0, 400, 400))
rc = page.insertText((0,0), text)

page.getText('dict')

outputs origin with the value: (-100, 50)

{
    'width': 2.0,
    'height': 2.0,
    'blocks': [
        {
            'number': 0,
            'type': 0,
            'bbox': (-100.0, 38.17499923706055, -81.6520004272461, 53.28900146484375),
            'lines': [
                {
                    'spans': [
                        {
                            'size': 11.0,
                            'flags': 0,
                            'font': 'Helvetica',
                            'color': 0,
                            'ascender': 1.0750000476837158,
                            'descender': -0.29899999499320984,
                            'text': '123',
                            'origin': (-100.0, 50.0),
                            'bbox': (-100.0, 38.17499923706055, -81.6520004272461, 53.28900146484375)
                        }
                    ],
                    'wmode': 0,
                    'dir': (1.0, 0.0),
                    'bbox': (-100.0, 38.17499923706055, -81.6520004272461, 53.28900146484375)
                }
            ]
        }
    ]
}

Could you help me to figure out how this origin is computed?

[JorjMcKie]

Your combination of mediabox / cropbox data make no sense: cropbox must be inside mediabox, which in turn must start at (0,0).

I will add checks to prevent things you have done.

[VeryLazyBoy]

Thanks for your reply, but I do encounter some pdf with abnormal mediabox and cropbox. For example, the file I attached:

123.pdf

file_name = '123.pdf'
doc = fitz.open(file_name)

page = doc.loadPage(0)
print(page.MediaBox, page.CropBox)
# gives Rect(9.0, 9.0, 621.0, 801.0), Rect(9.0, 0.0, 621.0, 792.0)

if I want to insert a text to this file, I find it would be inserted to a weird place

[JorjMcKie]

if I want to insert a text to this file, I find it would be inserted to a weird place

This has other reasons:
Inside the page definitions (/Contents object), geometry changes happen in a "non-localized" way - meaning they are not encapsulated properly by stacking / unstacking commands "q" / "Q".
Heal this by doing page.clean_contents() or page.wrap_contents() before your inserts and all is roses 😎:

>>> doc=fitz.open("123.pdf")
>>> page=doc[0]
>>> page.wrap_contents()
>>> page.insertText((100,100),"Hello world")
1
>>> doc.saveIncr()
>>> 

[JorjMcKie]

you can check for potential issues like that by looking at page.is_wrapped.

[JorjMcKie]

For more background of this read section "Misplaced Item Insertions on PDF Pages" of this documentation chapter.

[VeryLazyBoy]

This pdf seems to have multiple problems. I have tried your commands:

>>> doc=fitz.open("123.pdf")
>>> page=doc[0]
>>> page.wrap_contents()
>>> page.insertText((100,100),"Hello World!")
1
>>> page.getText('dict')['blocks'][1]['lines']
[{'spans': [{'size': 11.0,
    'flags': 0,
    'font': 'Helvetica',
    'color': 0,
    'ascender': 1.0750000476837158,
    'descender': -0.29899999499320984,
    'text': 'Hello World!',
    'origin': (100.0, 109.0),
    'bbox': (100.0,
     97.17500305175781,
     159.89498901367188,
     112.28900146484375)}],
  'wmode': 0,
  'dir': (1.0, 0.0),
  'bbox': (100.0, 97.17500305175781, 159.89498901367188, 112.28900146484375)}]

I am wondering if I need to correct the media box and crop box first to insert a text properly? Given the media box is Rect(9.0, 9.0, 621.0, 801.0) and crop box is Rect(9.0, 0.0, 621.0, 792.0), the text's bottom left is located at (100, 109) instead of (100, 100).

[JorjMcKie]

I am wondering if I need to correct the media box and crop box first to insert a text properly?

No, don't do that. This will break the position of other stuff on the page.
I will need to adjust PyMuPDF to accomodate this weird situation.
In the meantime subtract page.MediaBox.y0 from the desired insertion y-position as a workaround. In "normal" cases this value will be zero and do no harm.
So in your case, page.insertText((100, 100 - page.MediaBox.y0), "hello") should make it.

[VeryLazyBoy]

Thanks for your kindly help!