Add wicked_pdf_url_base64.

[joshuapinter]

We're using this in production to solve our problem. Looking to get some feedback from @mileszs before finalizing this properly and, hopefully, getting it merged to help others in similar situations.

Using URLs to reference images can cause a lot of problems with wkhtmltopdf, particularly when used in the header and footer. It produces errors such as "Too many open files" as well as buffer/stack overflows. These issues can be seen here, among other places:

• Large pdf generation header and footer fails #288
• *** buffer overflow detected ***: wkhtmltopdf terminated wkhtmltopdf/wkhtmltopdf#1819
• QT crashes when callin wkhtmltopdf with large number of inputs wkhtmltopdf/wkhtmltopdf#3019
• Buffer Overflow with large PDF wkhtmltopdf/wkhtmltopdf#2093
• Wkhtmltopdf exited with status 134 *** buffer overflow detected *** wkhtmltopdf/wkhtmltopdf#4762
• Too many open files issue wkhtmltopdf/wkhtmltopdf#3081
• Crash with html footers and headers on huge pdfs. Increase count of open files per user not possible for my case. wkhtmltopdf/wkhtmltopdf#4296

As described here, one of the key solutions to avoiding this issue is providing the image encoded as base64. This helper method takes a URL of an image, opens the URL, reads the image data and encodes it to base64.

If the URL cannot be open (OpenURI::HTTPError is raised), it will log a warning and return nil.

TODOs:

• Better placement (since this is not technically part of the asset pipeline)?
• Better naming?
• Tests.

[unixmonkey]

Does read_from_uri work here instead? This project has been cautioned against using URI#open for security reasons.

[joshuapinter]

Let me try it out. I original just had open() and Rubocop cautioned against it and in favour of URI.parse. Hang on...

[joshuapinter]

Okay, using read_from_uri won't work because it just returns a String and we need the proper response in order to get the content_type.

I was able to get it to work using:

response = Net::HTTP.get_response(URI(url))
base64 = Base64.encode64(response.body).gsub(/\s+/, '')
"data:#{response.content_type};base64,#{Rack::Utils.escape(base64)}"

Does that work?

P.s. We're still on Rails 4.2 but this made me open up an Issue with our own codebase to replace all uses of URI.parse.

[joshuapinter]

Okay, updated the code and also had to change how failed responses were handled.

Note, I wanted to setup a guard like:

if !response.is_a?(Net::HTTPSuccess)
  Rails.logger.warn("[wicked_pdf] #{response.code} #{response.message}: #{url}")
  return nil
end

But your Rubocop config wanted me to change it to:

unless response.is_a?(Net::HTTPSuccess)
  Rails.logger.warn("[wicked_pdf] #{response.code} #{response.message}: #{url}")
  nil
end

And I despise that with a passion (we have that Cop changed in our Rubocop config) so went with the if/else instead as it's much more clear.

[unixmonkey]

Thanks for this!

[joshuapinter]

👍 Thanks for a great library.