Test other options for HTML -> PDF

[amafehr]

Summary

Most tools that do this either have many dependencies or rely on a browser driver (slow performance, potentially some browser issues). We would like to find some middle ground tool if it exists.

Details

This has to be lightweight; many options open a headless browser reliant on a browser driver (e.g. selenium uses the Chrome driver) which takes a few seconds per conversion (see posted issue for follow-up on this separately). We explored numerous options that do not seem possible, so this area remains a task anyone is free to take on. Specifically, we tried pandoc, weasyprint, node.js html-to-pdf, and a less restrictive DOM linkedom.

• pandoc: would have been a good option but had many texlive (latex universe) and system dependencies that ultimately did not work out.
• weasyprint: we think it could work but may require some tinkering with the CSS elements in the HTML plot template. Library logging showed weasyprint fails to render SVG image and an option wasn't apparent.
• linkedom: As with the current DOM (jsdom), a richer browser was required (headless browser to render styles and flexboxes), making html-to-pdf a non-option.

TODO

• test out the following options for ease of install (focusing on dependencies and whether users on different OS can install) and performance.
• make a function that can implement the current option of your new favored option. Put this outside generate_svg.

Options (non-exhaustive) and what we know

• pyhtml2pdf: depends on selenium (browser driver) and ghostscript. This is the current implementation.
• weasyprint: looks alright, no evidence of browser driver. SVG support seems troublesome. I experimented and could not get a perfect PDF from it.
• fpdf2:
• html-to-browser" npm, generates image from DOM node (discovered this after getting 1st draft working). Doing all this in JS using sometihng like this may be more seamless than image generation on the python side.
• pandoc: I tried this route (HTML > LaTeX > PDF) and it has an incredible amount of installs and nested dependencies. I could not get a well-formatted LaTeX such that it could produce a PDF. As far as research took me, it seemed the hangs-up are around processing the images.
• imgkit: wkhtmltopdf is a common dependency many Python libraries wrap that homebrew is deprecating (2024)--this could prevent a future difficulty for install. imgkit is a wrapper around this library.
• html2image: python, in browser
• puppeteer / pyppeteer
• playwright

[amafehr]

I'm working to implement js's html-to-browser this week to remove the chromium dependency. Otherwise, install instructions are up-tp-date to reflect current installs needed.

[amafehr]

Womp womp, it didn't work out. See notes above.