When converting HTML to PDF using iTextPDF's html2pdf (version 4.0.5), font selection follows a two-step process based on the specified font-family in HTML styles.

Font Selection Process

1. Font Matching by Family Name The converter compares each font in the font-family list with available font packages. It assigns scores based on these criteria:

   if (!"".equals(fontFamily) 
       && (null == fontInfo.getAlias() 
       && null != fontDescriptor.getFamilyNameLowerCase() 
       && fontDescriptor.getFamilyNameLowerCase().equals(fontFamily) 
       || fontDescriptor.getFamilyNameLowerCase().startsWith(fontFamily.trim())
       || (null != fontInfo.getAlias() && fontInfo.getAlias().toLowerCase().equals(fontFamily)))) {
       score += FONT_FAMILY_EQUALS_AWARD;
   }
   

   The highest-scoring available font is selected. For example, KaiTi_GB2312 will match if its lowercase family name exactly matches the specified font.

2. Character Compatibility Check Even after selceting a font, individual characters are verified for compatibility:

   for (FontInfo f : selector.getFonts()) {
       int codePoint = isSurrogatePair(text, nextUnignorable)
             ? TextUtil.convertToUtf32(text, nextUnignorable)
             : (int) text.charAt(nextUnignorable);
   
       if (f.getFontUnicodeRange().contains(codePoint)) {
           PdfFont currentFont = getPdfFont(f);
           Glyph glyph = currentFont.getGlyph(codePoint);
           if (null != glyph && 0 != glyph.getCode()) {
               font = currentFont;
               break;
           }
       }
   }
   

   If a character (like   with Unicode 160) isn't supported by the primary font, the system falls back to another font (e.g., SimSun for Chinese text).

Script-Based Segmentation

Text is segmented by Unicode script type (e.g., HAN for Chinese, LATIN for English). Each segment uses a font that supports its script:

Character.UnicodeScript unicodeScript = nextSignificantUnicodeScript(nextUnignorable);
for (int i = nextUnignorable; i < text.length(); i++) {
    int codePoint = isSurrogatePair(text, i) 
          ? TextUtil.convertToUtf32(text, i) 
          : (int) text.charAt(i);
    Character.UnicodeScript currScript = Character.UnicodeScript.of(codePoint);
    if (isSignificantUnicodeScript(currScript) && currScript != unicodeScript) {
        break;
    }
    // Process compatible characters
}

This ensures mixed-script text (e.g., Chinese with Latin acronyms like "USDA") uses appropriate fonts for each segment.