Experiments in PDF generation purely client-side using JavaScript, Canvas, etc.

At work, we're looking into ways to use less-expensive technologies for simpler sites. And one of the "small details" that ends up being a breaker, is PDF generation. Our current technique is WkHTMLtoPDF to generate very customized PDFs (statistics, charts, maps) of content not necessarily like that on the web page. This means a server so we can use command-line tools (e.g. pdfkit for Python).

But if we can generate PDFs wholly client-side using static files, we could make strides in our quest to host at Github pages and other lower-cost and no-cost providers.

https://gregallensworth.github.io/clientside-pdf-printing/rasterized/

https://gregallensworth.github.io/clientside-pdf-printing/pdfmake/

https://gregallensworth.github.io/clientside-pdf-printing/pdfkit/

• Export of an element of HTML
  • Basic header, some content block and footer
  • Images
  • Hyperlinks
  • Bootstrap responsive table + real table
  • Multi-page content
  • Font usage
  • Ability to select text in the output PDF (copy-paste non-rasterized text)
• PDF export of an element not visible on the page
  • Some statistical readout, even if in HTML
  • Alternative header simulating letterhead
• PDF export of a Leaflet map
  • Basic map with tiles
  • Marker using SVG icon
  • GeoJSON overlay
• PDF export of charts
  • D3
  • Highcharts
• Testing of page break behavior e.g. lengthy lists that would run off the bottom, so we know what workarounds or caveats are required

Generally speaking, there are two techniques:

• Those based on html2canvas essentially render a visual representation of the HTML elements onto a Canvas, then save that Canvas as an image.
  • Examples: html2pdf, jsPDF, canvg
  • Some libraries can parse HTML effectively, making for more rapid development.
  • Leaflet maps still TBD, no good outcomes yet.
  • Resulting PDF is basically a static picture in a PDF container. It cannot have text selected (copy-paste), nor contain PDF metadata such as author info, page links and table of contents, and contains no text. For real PDF purposes this is effectively worthless.
• Those based on composing a true PDF binary structure, using lower-level API such as addPage(), setXY(), drawPath().
  • Examples: PDFkit, and pdfmake (a higher-level wrapper for common HTML structures)
  • Resulting PDFs are true PDFs with text, PDF metadata, etc.
  • Low-level approach denies embedding of maps and charts without additional significant work, e.g. using Canvas techniques to save the chart or map to an image, then embedding that image. Significantly more work, taking a few hours for even simple pretty layouts, and requiring effectively-infiniite labor for pixel-perfect maps, charts, etc.
    • Example: Using pdfmake, adding new fonts means rebuilding the entire distribution from scratch. https://github.com/bpampuch/pdfmake/wiki/Custom-Fonts---client-side
  • Kits lacks other fundamental features, e.g. triggering download or fetching content as a Blob for use with a hyperlink.

This is simply tests and documentation of our usage of someone else's work, [https://github.com/eKoopmans/html2pdf the html2pdf system by eKoopmans] Donations and kudos should be directed to that author, and to those of the underlying libraries html2canvas and jsPDF.