We'll respond shortly.
Validating that a web app’s content is rendered correctly is an integral part of testing web apps. Displaying user-submitted input in HTML is the core functionality of almost any website. For example, the last couple of web sites you used most likely had you enter information in a text box that was later shown to you on another page. A common way Rails developers achieve this is via RSpec and Capybara. Capybara provides lots of matchers to narrow down how to find elements on the page. But what if you aren’t displaying HTML to your users? What if the output is in a different format, like a PDF? How do you test PDFs with Capybara and RSpec?
Under one of my feature specs I attempted to call
page.should have_content('some content') on a rendered PDF. Unfortunately, an error is thrown:
invalid byte sequence in UTF-8. This error signals that the document contains invalid encoding. The problem could be a couple of unknown characters or that the document is missing an encoding type. Either way, Capybara needs a way to grab the actual contents in a format it can understand.
My first attempt to test content of a PDF used wicked_pdf’s debug option. By adding
:show_as_html => params[:debug].present? to my rendering line I accessed the HTML (pre-PDF rendering) by appending
?debug=1 to my request. Now that the content was standard HTML Capybara’s
page worked normally.
However, the PDF may not line up with the HTML 100% using this method. This discrepancy causes problems when using complex CSS to render your pages; just because the spec passes via HTML doesn’t guarantee the PDF will do the same. For example problems arise when using certain combinations of the
How the accompanying CSS is referenced is another issue with this approach. Providing an absolute reference to any assets is required, since the wkhtmltopdf binary is run outside of your Rails application. For example, a CSS file is referenced with the
wicked_pdf_stylesheet_link_tag "pdf" helper method.
As a result, the HTML-rendered views grab CSS via a relative path while the PDF grabs CSS via an absolute path. While this is not an issue for a small app, this can become difficult to maintain once there are multiple CSS modules for different output formats.
A different approach is to leverage pdf-reader‘s text rendering. You can then set the content of the Capybara
page directly. First, render the PDF document and save it to a temporary local file. Then tell pdf-reader to parse the text to a standard format for Capybara to use. Finally, directly set Capybara’s
@body variable to the (now valid) PDF text contents.
def convert_pdf_to_page temp_pdf = Tempfile.new('pdf') temp_pdf << page.source.force_encoding('UTF-8') reader = PDF::Reader.new(temp_pdf) pdf_text = reader.pages.map(&:text) page.driver.response.instance_variable_set('@body', pdf_text) end
Once your page has loaded, call
convert_pdf_to_page and use Capybara’s
page normally. All of the text matchers should work.
page.should have_content('some PDF content')
Now the code stays succinct by using Capybara’s built-in matchers for both HTML pages and PDF files. Now you don’t need to worry about learning a new DSL for PDF specs.
This post was heavily inspired by Matthew O’Riordan’s gist. It uses an older version of pdf-reader so it needs to set up a bunch of manual string parsing to work. Lucky for us one of pdf-reader’s new API method now makes it easy to test PDFs with Capybara.
spec_helper.rbto gain access to the helper method in all of your specs.