How to get text horizontal width

[murlai]

How to get the horizontal scaled value of the text. refer below image

horizontal_scaling.pdf

[JorjMcKie]

There is no way to extract this value.
What you can do however is the following:

pprint(page.get_text("dict")["blocks"])  # get the text in dictionary format
[{'bbox': (18.8164005279541,
           17.379932403564453,
           27.15677833557129,
           22.965869903564453),
  'lines': [{'bbox': (18.8164005279541,
                      17.379932403564453,
                      27.15677833557129,
                      22.965869903564453),
             'dir': (1.0, 0.0),
             'spans': [{'ascender': 0.9052734375,
                        'bbox': (18.8164005279541,
                                 17.379932403564453,
                                 27.15677833557129,
                                 22.965869903564453),
                        'color': 22352,
                        'descender': -0.2119140625,
                        'flags': 16,
                        'font': 'Arial-BoldMT',
                        'origin': (18.8164005279541, 21.906299591064453),
                        'size': 3.535533905029297,
                        'text': '123456'}],
             'wmode': 0}],
  'number': 0,
  'type': 0},
 {'bbox': (18.8164005279541,
           25.971233367919922,
           43.837547302246094,
           40.14897155761719),
  'lines': [{'bbox': (18.8164005279541,
                      25.971233367919922,
                      35.49716567993164,
                      31.557170867919922),
             'dir': (1.0, 0.0),
             'spans': [{'ascender': 0.9052734375,
                        'bbox': (18.8164005279541,
                                 25.971233367919922,
                                 35.49716567993164,
                                 31.557170867919922),
                        'color': 22352,
                        'descender': -0.2119140625,
                        'flags': 16,
                        'font': 'Arial-BoldMT',
                        'origin': (18.8164005279541, 30.497600555419922),
                        'size': 5.0,
                        'text': '123456'}],
             'wmode': 0},
            {'bbox': (18.8164005279541,
                      34.56303405761719,
                      43.837547302246094,
                      40.14897155761719),
             'dir': (1.0, 0.0),
             'spans': [{'ascender': 0.9052734375,  # this is the text span we want to look at
                        'bbox': (18.8164005279541,
                                 34.56303405761719,
                                 43.837547302246094,  # width of the text is bbox.width
                                 40.14897155761719),
                        'color': 22352,
                        'descender': -0.2119140625,
                        'flags': 16,
                        'font': 'Arial-BoldMT',
                        'origin': (18.8164005279541, 39.08940124511719),
                        'size': 6.123724460601807,  # the font size used
                        'text': '123456'}],
             'wmode': 0}],
  'number': 1,
  'type': 0}]
pprint(page.get_fonts())  # extract the font of that text span
[(5, 'ttf', 'TrueType', 'TWVHJA+Arial-BoldMT', 'TT0', 'WinAnsiEncoding')]
font=fitz.Font(fontbuffer=doc.extract_font(5)[-1])  # create the fitz.Font of that font
font.text_length("123456",fontsize=6.123724460601807)  # check the "normal" text width of that font size
20.434342267457396
width=43.837547302246094 - 18.8164005279541  # bbox width
width / 20.434342267457396  # compute actual width divided by "normal" width
1.2244654829991415  # so it looks like our text is 22.45% wider

[JorjMcKie]

Looking at all the different spans:

spans = []
for b in page.get_text("dict")["blocks"]:
    for l in b["lines"]:
        for s in l["spans"]:
            spans.append(s)

            
for s in spans:
    fs = s["size"]
    width=s["bbox"][2]-s["bbox"][0]
    tl=font.text_length("123456",fontsize=fs)
    print(width / tl)

    
0.706945111138104
0.9997719353599649
1.2244654829991415

[murlai]

@JorjMcKie thanks for this solution. it's working perfectly.

But I have another issue.
It has a 5pt font size in illustrator. when we convert to PDF it's reduced to 3.535533905029297.

I know because of the horizontal scale. My original requirement is, how to get the correct font size as per Adobe illustrator or Adobe acrobat.

[JorjMcKie]

My original requirement is, how to get the correct font size as per Adobe illustrator or Adobe acrobat.

You could do this:
extract the text after executing fitz.TOOLS.set_small_glyph_heights(True). Then

• The standard character width in a font is called "glyph_advance". For example font.glyph_advance(ord("1")) = 0.55615234375 for character "1".
• For this character with a horizontal scale of 0.5, the character dictionary (extracted with option "rawdict") is this:

{'bbox': (18.8164005279541,
          19.041404724121094,
          20.2067813873291,
          22.57693862915039),
 'c': '1',
 'origin': (18.8164005279541, 21.906299591064453)}

• so the width w = 20.2067813873291 - 18.8164005279541 = 1.390380859375,
• the height h is the fontsize 3.535533905029297
• Then compute font.glyph_advance(ord("1"))*h*h/w = 4.999999997444684, which is the specified original fontsize.

Please also note that the formnulation "original fontsize" is not correct ... because Adobe neither reports it!
In your example, these are the commands which render the page:

print(page.read_contents().decode())
/OC /MC0 BDC 
BT
/CS0 cs 1  scn
/GS0 gs
/TT0 1 Tf  # "original" fontsize is 1!
2.5 0 0 5 18.8164 36.4541 Tm  # this text matrix scales x by 2.5 and y by 5
(123456)Tj
5 0 0 5 18.8164 27.8628 Tm
(123456)Tj
7.5 0 0 5 18.8164 19.271 Tm
(123456)Tj
ET
EMC 

So it is somewhat deliberate to say 5 is the original font size. The reported font size 3.535533905029297 is equal to sqrt(2.5 * 5), the geometric mean of the two scaling factors.