rustedf.cloud
Menu

HTML to PDF: library, headless browser or API? How to choose

Published 4 min read

There are three common ways to turn HTML into a PDF: a library inside your application, a browser that you run without a screen, and an API that does the work elsewhere. None of them is the right answer for every project. This guide compares them on what you have to run, what HTML and CSS you get, and where the document goes, and ends with the questions that decide.

The three ways

A library in your application. You add a package and call it. It lays out HTML and CSS with its own engine, so it can be small and quick to start. It usually supports a subset of CSS, and what is outside the subset is yours to work around. It runs in your process, with the memory and the failures that come with that, and you keep it updated.

A headless browser. You run a real browser without a window and ask it to print a page. You get the HTML and CSS support of the browser, including JavaScript. In exchange you install the browser in every environment, give it memory and time for each document, and keep it patched. At volume you manage a pool of browsers, queues and restarts.

An API. You send the HTML and the data in a request and get a PDF back. There is nothing to install or to keep running, and the cost grows with the number of documents. The document and its data leave your own infrastructure, you depend on the service being available, and its limits (the size of a request, the pages of a document, the rate of calls) apply to you.

What each asks of you

Aspect Library Headless browser API
What you run A package in your app A browser in every environment Nothing, a client call
HTML and CSS you get Whatever the engine supports Whatever the browser supports Whatever the service supports
JavaScript while rendering Usually not Yes Depends on the service
Where the data goes Stays with you Stays with you To the service
Who updates it You You The service
Cost grows with Your servers Your servers The number of documents

Where rustedf.cloud sits

It is the third way, and it has clear limits. Documents are laid out for paper: pages, tables that continue across pages, headers and footers, page numbers. It supports a defined subset of HTML and CSS, and it does not run JavaScript. Images must be uploaded to your account or embedded as data: URLs, because the service never fetches an address from a template. Anything outside the subset is ignored and reported, and a validation endpoint tells you what before you send it.

What you get in return is no browser to run, templates with live preview in a portal, formatting of numbers, currencies and dates for any locale, and a failed render that costs nothing.

Questions that decide

  • Do your documents need JavaScript to render? If a chart or a layout is drawn by a script, you need a browser or something that can run it. A template-based API will not do.
  • Must the data stay inside your infrastructure? If a contract or a regulation says so, use a library or your own browser.
  • How many documents, and how steady? A few thousand a month in bursts suits an API well. A very large, steady volume may be cheaper to run yourself.
  • Who will keep it running? If nobody on the team wants to own a browser pool, that is a cost that does not appear in the first week.
  • How much of the web page must be reproduced? An invoice, a quote, a report or a certificate fits the paper-oriented subset. A page copied from a complex web application may not.

If most answers point to “documents built from data, laid out for paper, nothing to maintain”, try the API with one real document and validate it. The free plan is for exactly that.

Read Supported HTML and CSS and Rendering a PDF.

← All posts