CODEWITHSUNDEEP
pdffontsitextitext7
scott13
scott13

Reputation: 77

When using iText to generate a PDF, if I need to switch fonts many times the file size becomes too large

I have a section of my PDF in which I need to use one font for its unicode symbol and the rest of the paragraph should be a different font. (It is something like "1. a 2. b 3. c" where "1." is the unicode symbol/font and "a" is another font) I have followed the method Bruno describes here: iText 7: How to build a paragraph mixing different fonts? and it works fine to generate the PDF. The issue is that the file size of the PDF goes from around 20MB to around 100MB compared to using only one font and one Text element. This section is used repeatedly in the document thousands of times. I am wondering if there is a way to reduce the impact of switching fonts or to reduce the file size of the entire document in some way.

Style creation pseudocode:

Style style1 = new Style();
Style style2 = new Style();
PdfFont font1 = PdfFontFactory.createFont(FontProgramFactory.createFont(fontFile1), PdfEncodings.IDENTITY_H, true);
style1.setFont(font1).setFontSize(8f).setFontColor(Color.DARK_GRAY);
PdfFont font2 = PdfFontFactory.createFont(FontProgramFactory.createFont(fontFile2), "", false);
style2.setFont(font2).setFontSize(8f).setFontColor(Color.DARK_GRAY);

Writing text/paragraph pseudocode:

Div div = new Div().setPaddingLeft(3).setMarginBottom(0).setKeepTogether(true);
Paragraph paragraph = new Paragraph();
loop up to 25 times: {
    Text unicodeText = new Text(unicodeSymbol + " ").addStyle(style1);
    paragraph.add(unicodeText);
    Text plainText = new Text(plainText + " ").addStyle(style2);
    paragraph.add(plainText);
}
div.add(paragraph);

This writing of text/paragraph is done thousands of times and makes up most of the document. Basically the document consists of thousands of "buildings" that have corresponding codes and the codes have categories. I need to have the index for the category as the unicode symbol and then all of the corresponding codes within the paragraph for the building.

Here is reproducable code:

    float offSet = 50;
    Integer leading = 10;
    DateFormat format = new SimpleDateFormat("yyyy_MM_dd_kkmmss");
    String formattedDate = format.format(new Date());
    String path = "/tmp/testing_pdf_"+formattedDate + ".pdf";
    File targetPdfFile = new File(path);
    PdfWriter writer = new PdfWriter(path, new WriterProperties().addXmpMetadata());
    PdfDocument pdf = new PdfDocument(writer);
    pdf.setTagged();
    PageSize pageSize = PageSize.LETTER;
    Document document = new Document(pdf, pageSize);
    document.setMargins(offSet, offSet, offSet, offSet);
    byte[] font1file = IOUtils.toByteArray(FileUtility.getInputStreamFromClassPath("fonts/Garamond-Premier-Pro-Regular.ttf"));
    byte[] font2file = IOUtils.toByteArray(FileUtility.getInputStreamFromClassPath("fonts/Quivira.otf"));
    PdfFont font1 = PdfFontFactory.createFont(FontProgramFactory.createFont(font1file), "", true);
    PdfFont font2 = PdfFontFactory.createFont(FontProgramFactory.createFont(font2file), PdfEncodings.IDENTITY_H, true);
    Style style1 = new Style().setFont(font1).setFontSize(8f).setFontColor(Color.DARK_GRAY);
    Style style2 = new Style().setFont(font2).setFontSize(8f).setFontColor(Color.DARK_GRAY);
    float columnGap = 5;
    float columnWidth = (pageSize.getWidth() - offSet * 2 - columnGap * 2) / 3;
    float columnHeight = pageSize.getHeight() - offSet * 2;
    Rectangle[] columns = {
            new Rectangle(offSet, offSet, columnWidth, columnHeight),
            new Rectangle(offSet + columnWidth + columnGap, offSet, columnWidth, columnHeight),
            new Rectangle(offSet + columnWidth * 2 + columnGap * 2, offSet, columnWidth, columnHeight)};
    document.setRenderer(new ColumnDocumentRenderer(document, columns));
    for (int j = 0; j < 5000; j++) {
        Div div = new Div().setPaddingLeft(3).setMarginBottom(0).setKeepTogether(true);
        Paragraph paragraph = new Paragraph().setFixedLeading(leading);
//            StringBuilder stringBuilder = new StringBuilder();
        for (int i = 0; i < 26; i++) {
            paragraph.add(new Text("\u3255 ").addStyle(style2));
            paragraph.add(new Text("test ").addStyle(style1));
//                stringBuilder.append("\u3255 ").append(" test ");
        }
//            paragraph.add(stringBuilder.toString()).addStyle(style2);

        div.add(paragraph);
        document.add(div);
    }
    document.close();

In creating the reproducible code I have found this this is related to the document being tagged. If you remove the line that marks it as tagged it reduces the file size greatly.

You can also reduce the file size by using the commented out string builder with one font instead of two. (Comment out the two "paragraph.add"s in the for-loop) This mirrors the issue I have in my code.

Upvotes: 0

Views: 758

Answers (1)

Alexey Subach
Alexey Subach

Reputation: 12312

The problem is not in fonts themselves. The issues comes from the fact that you are creating a tagged PDF. Tagged documents have a lot of PDF objects in them that need a lot of space in the file.

I wasn't able to reproduce your 20MB vs 100MB results. On my machine whether with one font or with two fonts, but with two Text elements, the resultant file size is ~44MB.

To compress file when creating large tagged documents, you should use full compression mode which compresses all PDF objects, not only streams.

To activate full compression mode, create a PdfWriter instance with WriterProperties:

PdfWriter writer = new PdfWriter(outFileName, 
                  new WriterProperties().setFullCompressionMode(true));

This setting reduced the file size for me from >40MB to ~5MB.

Please note that you are using iText 7.0.x while 7.1.x line has already been released and is now the main line of iText, so I recommend that you update to the latest version.

Upvotes: 2

Related Questions