Convert images links from external to internal

[snowman]

Here's the script that converts image links to internal.

#!/usr/bin/env ruby

require 'open-uri'
require 'uri'
require 'pry'

Dir.mkdir('images') unless File.exist?('images')

def transform_url(url)
  url.sub('/optimized/', '/original/').sub(/_\d_\d+x\d+(\.\w+)$/, "\\1")
end

def url_to_filename(url)
  url.split('/')[-1]
end

def download_file(url, filename)
  URI.open(url) do |image|
    p "downloading #{filename}..."
    File.open(filename, 'wb') { |file| file.write(image.read) }
  end
end

def process_line(line)
  urls =
    URI.extract(line).filter { |url| url.include? 'emacs-china.org/uploads' }

  urls.each do |url|
    url = url.sub(/\)$/, "")

    transformed_url = transform_url(url)
    filename = 'images/' + url_to_filename(transformed_url)
    download_file(transformed_url, filename) unless File.exist?(filename)

    p "renaming #{url} to #{filename}"
    line.sub!(url, filename) if File.exist?(filename)
  end

  line
end

Dir['*.md'].each do |filename|
  lines = open(filename).each_line.map {|line| process_line(line) }
  File.open(filename, 'w') { |file| file.write(lines.join) }
end

[condy0919]

Maybe we can abuse GitHub issues to store images? e.g. condy0919/.emacs.d#37

[snowman]

Why have to use external links? It's a slow and bad user experience.

How I read the .md file?

I clone the repos and read it with markdown-mode, and use
imenu and narrow-to-subtree to focus on one section, and
delete the read section later with git commit per day.

To make the images links internal, we can view images locally with
image viewer like https://github.com/muennich/sxiv seamlessly with
zoom in/out, stop frame animation of GIF and manually jump between
frames, etc and without bear network latency.

And also this means integrity, rely on external service is unstable,
and GitHub is slow without using proxies like v2ray.

Use single same repos with internal images, you can deploy them
anywhere you want without limited by Image Hotlinking protection.

I think you have the original images files with good naming instead of
hash-numbering file, so by using external images links, you don't have
to host the images yourself. If you want to automatically rename those
files in markdown file, here is the utility that finds visually
similar or duplicate images (https://github.com/jhnc/findimagedupes).

Anyway, that's my two cents. You can close this pull request and
reopen it and merge it someday later when you think about it again or
not.

Have a good day! :)

[condy0919]

The original images are lost forever...

And also this means integrity, rely on external service is unstable, and
GitHub is slow without using proxies like v2ray.

Yes, it shouldn't reply on external services.