ChromicPDF seems to be the popular choice for pdf generation at the moment. However, I wanted to get some community input on pdf generation libraries that DO NOT have a dependency on Chromium, puppeteer etc. I am trying to avoid the dev ops overhead of managing these additional dependencies. I would rather just draw out a PDF from scratch.
Libraries I looked into:
elixir-pdf - seems functional for the basic needs. elixir2pdf - The underlying erlang library erlguten is giving me compilation errors so I could not test this
I’ve been using Typst, a Rust library for generating PDFs from markdown. Typst is also a pure functional language such that you can put display logic into functions in a template to keep such logic out of Elixir.
To use the latest version, 0.12, you’ll want to use it on the command line from Elixir. There is an Elixir library that packages the Rust for version 0.3 called ExTypst and a fork that updates it to version 0.9.
Thank you. I did look at PDF generator but unfortunately WKHTML has been archived and is no longer available through brew or asdf. It seems most people are trying to migrate of it.
I find styling in in HTML and CSS then converting to PDF easier. As for generation using Chromium, the best way to do it is by using an AWS Lambda. Handles concurrency of multiple invocations perfectly.
Just done whole investigation into this for work (we are a Java house).
If you can avoid it in the first place, that’s the best approach (i.e. let the client print as pdf from their browser and use css3 styling for print media).
I did contemplate the XSL-FO route, but that seems progressively out of fashion, (I believe it’s down to the push for better CSS for printed media).
I also thought about postscript or tex, but that felt just as fiddly.
I think not, unless HTML will be page-size aware at some point, using that technology is very limited when it comes to generating more involved PDFs that actually need to follow some page structure.
I’ve used successfully latex for generating some pretty beefy PDFs before, however the way compiler was implemented makes it really hard to work with and it’s really laggy, especially if you have to do multiple passes. The best alternative on the market is mentioned above by @asianfilm, typst.
There is no great integration library in elixir, yet, but this shouldn’t stop you from using the compiler as external dependency.
Thanks to everyone who replied, I ended up choosing Typst. The markdown is pretty powerful and easy to learn. There free online editor makes it easy to experiment with the markdown as well.
I don’t have any data to back this up but figured a rust library would be more performant than weasyprint in python
I could not get ex_typst to work, so I just called the typst compiler using System.cmd.
I have a base .typ file that has the layout and template variables. I use elixir’s EEx to fill out the data and write a .typ file to a temporary location. Finally I execute the typst compiler using the shell.
@arun1 not sure if you still need a solution here, but in case it helps: we’re using Gotenberg to generate PDF’s from HTML/CSS. We’re running the default Gotenberg Docker image and use the HTTP API to convert rendered Phoenix templates to PDF.
This is especially nice because we can use all the normal view helpers and Tailwind styling that we have throughout our application, and these days things like a table of contents or page numbering is pretty much solved in CSS.
Let me know if you’re interested in some example code of our setup, and the resulting PDF file!
Hm, good question, I don’t know if Gotenberg allows creating multiple PDF’s at once, but we haven’t ran into any issues so far (and have this setup in production for over a year).
To start of, this is how we deploy Gotenberg in Kubernetes. Any other platform should work as long as it can run Docker containers. We don’t change any configuration; this is just the vanilla Gotenberg 7 image.
Then we define some options for Gotenberg that are used in the API call when creating a PDF.
We then define a module that interacts with the Gotenberg API. It does an HTTP POST to the /forms/chromium/convert/html endpoint with our HTML page as a form/multipart request body.
Finally, define a module in your application that takes some data, renders an HTML file (from a Phoenix View), and then calls the previously defined Gotenberg API module to render it to PDF.
Hope this helps! Let me know if you have any questions
