html-to-favicon-url

#!/bin/sh
#
# Parse HTML to extract the favicon URL.
#
# Syntax:
#
#     ... | html-to-favicon-href
#
# Example:
#
#     wget http://example.com | html-to-favicon-href
#     => http://example.com/favicon.ico
#
# ## Implementation
#
# This implementation requires Ruby and the Nokogiri gem.
#
# We have tried other implementations that haven't worked so far,
# such as using hxselect, or xmlstarlet, or pup, but these failed
# on various real-world web pages e.g. www.kickstarter.com:
#
#   * The `pup` implementation finds the link, but the
#     selector doesn't seem to be able to do `contains`:
#
#         ... | pup '//link[contains(concat(" ", @rel, " "), " icon ")]'
#
#   * The `tidy` approach works to clean up the HTML, but nothing
#     later in the pipe is able to find the link and its href:
#
#         ... | tidy -q --show-warnings no -numeric -asxhtml
#
#   * The `hxselect` approach works to find the link,
#     but the CSS selector is unable to get the href:
#
#         ... | hxclean 2>/dev/null | hxselect 'link[rel~="icon"]'
#
#   * An implmentation using ruby succeeds, but needs nokogiri:
#
#         ... | ruby -rnokogiri -p -e \
#         'd=Nokogiri::HTML(STDIN.read); $_=""; puts d.xpath("//link[contains(concat(\" \", @rel, \" \"), \" icon \")]").first["href"]'
#
# Command: html-to-favicon-url
# Version: 1.3.0
# Contact: Joel Parker Henderson (joel@joelparkerhenderson.com)
# License: GPL
# Created: 2015-05-31
# Updated: 2016-02-09
##
set -euf

pup 'link[rel~="icon"] attr{href}' | head -1