Size of PDF file generated is high #2987

shashikulk · 2024-01-07T18:30:22Z

shashikulk
Jan 7, 2024

First of all, thanks for adding insert_htmlbox() feature/function to pymupdf. It works very well for non-English script. However, I have one observation regarding file size.

I am using insert_htmlbox() to insert some text (English as well as Devanagari) to my file. The size of the file created seems very high for the amount of text content I am writing to it. Below is the code I am using as an example in which I am writing only 1 line of Devanagari text to it, and the size of the pdf file generated is 242 KB.

import fitz_new as fitz
import os
out_path = str(os.getcwd()) + '/output/'
pdffilename = out_path + 'sample.pdf'

doc = fitz.open()
page = doc.new_page()
text = "PDF तयार करण्याचा हा माझा पहिला दस्तऐवज प्रकल्प आहे"
rect = fitz.Rect(30, 30, 480, 55)
page.insert_htmlbox(rect, text)
doc.save(pdffilename)
doc.close()

If I add more lines of text using insert_htmlbox(), the file size increases proportionately. Is there any way to keep the pdf file size to low?

I am using latest pymupdf with Python 3.10.x and Win11.

regards,
Shashikant

Answered by JorjMcKie

Jan 8, 2024

The method dynamically pulls in all required fonts. These may be several. Asian fonts typically are large, but the number of glyphs that you actually use in your text is small.
Therefore you should make sure to

build font subsets which only contain the glyphs that are actually in use
save the file with all available compression options

Please try the following code version:

import fitz_new as fitz
import os
out_path = str(os.getcwd()) + '/output/'
pdffilename = out_path + 'sample.pdf'

doc = fitz.open()
page = doc.new_page()
text = "PDF तयार करण्याचा हा माझा पहिला दस्तऐवज प्रकल्प आहे"
rect = fitz.Rect(30, 30, 480, 55)
page.insert_htmlbox(rect, text)
doc.subset_fonts()
doc.ez_save(pdffil…

View full answer

JorjMcKie · 2024-01-08T07:32:36Z

JorjMcKie
Jan 8, 2024
Maintainer

The method dynamically pulls in all required fonts. These may be several. Asian fonts typically are large, but the number of glyphs that you actually use in your text is small.
Therefore you should make sure to

build font subsets which only contain the glyphs that are actually in use
save the file with all available compression options

Please try the following code version:

import fitz_new as fitz
import os
out_path = str(os.getcwd()) + '/output/'
pdffilename = out_path + 'sample.pdf'

doc = fitz.open()
page = doc.new_page()
text = "PDF तयार करण्याचा हा माझा पहिला दस्तऐवज प्रकल्प आहे"
rect = fitz.Rect(30, 30, 480, 55)
page.insert_htmlbox(rect, text)
doc.subset_fonts()
doc.ez_save(pdffilename)
doc.close()

0 replies

JorjMcKie · 2024-01-08T08:50:27Z

JorjMcKie
Jan 8, 2024
Maintainer

Going to move this to the "Discussions" tab.

10 replies

shashikulk Jan 10, 2024
Author

Thanks a lot. Will try these methods / features. May have to learn a little bit of HTML ;-)

shashikulk Jan 12, 2024
Author

Need some help.
I am trying to create a table in PDF using the approach given here
It is working fine, except a few styles (table border radius and column widths) are not reflecting in the output pdf. If I view the html (the contents of text variable I am generating through my code) in a browser, all the styles are showing correctly. What could be the reason for difference in behavior in pdf output created by insert_htmlbox vs. browser? Or, am I missing something?

Given below is my code for reference.

import random
import fitz_new as fitz

srno, ques, ans = [], [], []
filename = "<your-local-file-name>.pdf"
qlang = random.choice(['English', 'Marathi'])
print('language is ', qlang)
for i in range(15):
    srno.append(str(i+1))
    if qlang == 'English':
        ques.append("random text # This is some random text "+str(i+1))
    else:
        ques.append("न्यूयॉर्क, श्रावणबेळगोळा, कल्याण # यादृच्छिक प्रश्न " + str(i + 1))
    ans.append(str(random.randint(1000, 10000000)))

if qlang == 'English':
    mystyle = """<td style="font-family: QuicksandFM;">"""
else:
    mystyle = """<td style="font-family: NotoSansDevanagari;">"""

arch_OT = fitz.Archive('<your-local-fonts-directory>')
doc = fitz.Document()
page = doc.new_page()

text = """
<style>
@font-face {font-family: NotoSansDevanagari; src: url(NotoSansDevanagari-Regular.ttf);}
@font-face {font-family: QuicksandFM; src: url(Quicksand-Regular.ttf);}
@font-face {font-family: QuicksandFM; src: url(Quicksand-Bold.ttf); font-weight: bold;}
td, th {padding: 5px; text-align: center;}
table {
    border: 0.25px solid blue;
    border-spacing: 0;
    border-radius: 2px;
    table-layout: fixed;
}
table td, table th {
    border: 0.25px solid black;
}
table tr td {
    border-right: 0;
    border-bottom:0;
}
table th {
    border-right: 0;
    border-top: 0;
    border-bottom: 0;
    background-color: #E0FFFF;
}
</style>
<body>
<table>
    <tr style="font-family: QuicksandFM;">
        <th style="border-left:0; width: 30px;"><b>#</b></th>
        <th style="width: 600px;"><b>Question</b></th>
        <th style="width: 100px;"><b>Answer</b></th>
    </tr>
"""
for i in range(15):
    text = text + """<tr>
        <td style="border-left:0; font-family: QuicksandFM;">""" + srno[i] + """</td>
        """ + mystyle + ques[i] + """</td>
        <td style="font-family: QuicksandFM;">""" + ans[i] + """</td>
        </tr>"""

text = text + """
    </table>
    </body>
    """

print('text is ', text)
rect = fitz.Rect(30, 30, 680, 700)
page.insert_htmlbox(rect, text, archive=arch_OT)
doc.ez_save(filename)

JorjMcKie Jan 12, 2024
Maintainer

Column widths of HTML tables cannot be influenced - at least not easily. They are computed and optimized by the Story code itself.
You can exert some tricky influence by adding a table row with font height 0 and a desired number of non-breaking spaces   - normal double spaces are contracted to one space as usual.
Not sure about border radius yet. Maybe that it indeed is unsupported by MuPDF. I can ask if you need it or ask yourself in the public MuPDF Discord channel. MuPDF Story supports HTML4 and CSS2 roughly.

Other comment:
You shouldn't need to add font NotoSansDevanagari-Regular.ttf - it is one of the MuPDF standard ones. But won't harm either to name it that way.

JorjMcKie Jan 12, 2024
Maintainer

Ah - and you definitely should use doc.subset_fonts() immediately before save.
Also make sure to use garbage collection 3 or simply ez_save.

JorjMcKie Jan 12, 2024
Maintainer

All Asian fonts are large and will thus benefit very much from building font subsets (= remove unused glyphs).

shashikulk · 2024-01-15T05:39:13Z

shashikulk
Jan 15, 2024
Author

Thanks for your response.

Since column width and row height are not in my control as a user of insert_htmlbox(), and the   approach was getting trickier due to varying length of text in each row, I decided to use make_table() individually for each column.

Now I am using make_table(), coupled with shape.draw_rect() and insert_htmlbox(). So far it is giving me desired result and control in terms of column width and row height.

Noted your other comments.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Size of PDF file generated is high #2987

{{title}}

Replies: 3 comments 10 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

Select a reply

Size of PDF file generated is high #2987

shashikulk Jan 7, 2024

Replies: 3 comments · 10 replies

JorjMcKie Jan 8, 2024 Maintainer

JorjMcKie Jan 8, 2024 Maintainer

shashikulk Jan 10, 2024 Author

shashikulk Jan 12, 2024 Author

JorjMcKie Jan 12, 2024 Maintainer

JorjMcKie Jan 12, 2024 Maintainer

JorjMcKie Jan 12, 2024 Maintainer

shashikulk Jan 15, 2024 Author

shashikulk
Jan 7, 2024

Replies: 3 comments 10 replies

JorjMcKie
Jan 8, 2024
Maintainer

JorjMcKie
Jan 8, 2024
Maintainer

shashikulk Jan 10, 2024
Author

shashikulk Jan 12, 2024
Author

JorjMcKie Jan 12, 2024
Maintainer

JorjMcKie Jan 12, 2024
Maintainer

JorjMcKie Jan 12, 2024
Maintainer

shashikulk
Jan 15, 2024
Author