#+TITLE: RSS Generation
#+SUBTITLE: How to generate RSS feed via command line
#+AUTHOR: Yann Esposito
#+EMAIL: yann@esposito.host
#+DATE: [2019-09-30 Mon]
#+KEYWORDS: programming, web
#+DESCRIPTION: How I generate RSS feed via command line
#+OPTIONS: auto-id:t

#+begin_notes
TL;DR: To generate an RSS file you need to provide many metadatas.
Those metadata are not part of all HTML files.
So generating RSS from a tree of HTML file is not straightforward.
Here is the script I use.
#+end_notes

* RSS Problem
:PROPERTIES:
:CUSTOM_ID: rss-problem
:END:

RSS feed is meant to declare updates and new articles for a website.
Each RSS entry must therefore have a date, an unique id, a title, maybe
some categories, etc...

For most blog platform or even static website generator, those meta infos
are clearly put in the sources or in some DB.

I use =org-mode= for generating my website, and the =ox-rss= is quite slow
when generating an RSS with the full content of each item.
Mainly, the way to achieve full content of my articles inside an RSS with
=ox-rss= is by first creating a very big org file containing all the
articles, and then transforming it in RSS. And this is very slow (many minutes).

So a simpler idea inspired by lb[fn:lb] is to generate the RSS directly
from the generated HTML files.
The only difficulty is to find the metadata inside those HTML.
Unfortunately there is no real standard for all those metas.

Has there is no standard place to have all those meta informations inside
an HTML file in order to use the HTML as source you'll need to "parse" the
HTML file.
For that purpose I use =html-xml-utils=[fn:hu].

I wrote a simple zsh script; it starts with lot of variables to fill:


#+begin_src bash
# Directory
webdir="_site" # directory containing your website html files
postsdir="$webdir/posts" # directory containing the articles
rssfile="$webdir/rss.xml" # the RSS file to generate

# maximal number of articles to put in the RSS file
maxarticles=10

# RSS Metas
rsstitle="her.esy.fun"
rssurl="https://her.esy.fun/rss.xml"
websiteurl="https://her.esy.fun"
rssdescription="her.esy.fun articles, mostly random personal thoughts"
rsslang="en"
rssauthor="yann@esposito.host (Yann Esposito)"
rssimgtitle="yogsototh"
rssimgurl="https://her.esy.fun/img/FlatAvatar.png"
#+end_src

Then I set the accessor to extract the information I want from HTML files.
It is quite unfortunate that there is no really strong convention for where
to put article dates, article author email.
There are metas for title and keywords thought.

#+begin_src bash
# HTML Accessors (similar to CSS accessors)
dateaccessor='.article-date'
contentaccessor='#content'
# title and keyword shouldn't be changed
titleaccessor='title'
keywordsaccessor='meta[name=keywords]::attr(content)'
#+end_src

A few helper functions:

#+begin_src bash
formatdate() {
    # format the date for RSS
    local d=$1
    LC_TIME=en_US date --date $d +'%a, %d %b %Y %H:%M:%S %z'
}

finddate(){ < $1 hxselect -c $dateaccessor }
findtitle(){ < $1 hxselect -c $titleaccessor }
getcontent(){ < $1 hxselect $contentaccessor }
findkeywords(){ < $1 hxselect -c $keywordsaccessor | sed 's/,//g' }
mkcategories(){
    for keyword in $*; do
        printf "\\n<category>%s</category>" $keyword
    done
}
#+end_src

The =mkcategories= will be used to add an RSS category for each keyword.
And finally the real loop doing the work:

#+begin_src bash
tmpdir=$(mktemp -d) # create a temporary work dir
typeset -a dates    # an array to save dates of all articles
dates=( )

# for each HTML file we generate the XML for the item in a file
# named ${d}-$(basename $fic).rss that naming convention will be useful to
# sort article by date
for fic in $postsdir/**/*.html; do
    blogfile="$(echo "$fic"|sed 's#^'$postsdir'/##')"
    printf "%-30s" $blogfile
    xfic="$tmpdir/$fic.xml"
    mkdir -p $(dirname $xfic)
    hxclean $fic > $xfic # create a cleaner HTML file to help hxselect work
    d=$(finddate $xfic)
    echo -n " [$d]"
    rssdate=$(formatdate $d)
    title=$(findtitle $xfic)
    keywords=( $(findkeywords $xfic) )
    printf ": %-55s" "$title ($keywords)"
    # up until here, we extracted the informations we need for the item
    categories=$(mkcategories $keywords)
    { printf "\\n<item>"
      printf "\\n<title>%s</title>" "$title"
      printf "\\n<guid>%s</guid>" "${websiteurl}/${blogfile}"
      printf "\\n<pubDate>%s</pubDate>%s" "$rssdate"
      printf "%s" "$categories"
      printf "\\n<description><![CDATA[\\n%s\\n]]></description>" "$(getcontent "$xfic")"
      printf "\\n</item>\\n\\n"
    } >>  "$tmpdir/${d}-$(basename $fic).rss"
    # we append the date to the list of dates
    dates=( $d $dates )
    echo " [${fg[green]}OK${reset_color}]"
done

# Now we publish the items in reverse newer articles first
echo "Publishing"
for fic in $(ls $tmpdir/*.rss | sort -r | head -n $maxarticles ); do
    echo "${fic:t}"
    cat $fic >> $tmpdir/rss
done

# we get the latest publish date
rssmaxdate=$(formatdate $(for d in $dates; do echo $d; done | sort -r | head -n 1))
# we put the current date for the latest build date
rssbuilddate=$(formatdate $(date))

# we generate the RSS file
{
# Write the preamble of the RSS file
cat <<END
<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0"
	   xmlns:content="http://purl.org/rss/1.0/modules/content/"
	   xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	   xmlns:dc="http://purl.org/dc/elements/1.1/"
	   xmlns:atom="http://www.w3.org/2005/Atom"
	   xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	   xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	   xmlns:georss="http://www.georss.org/georss"
     xmlns:geo="http://www.w3.org/2003/01/geo/wgs84_pos#"
     xmlns:media="http://search.yahoo.com/mrss/"><channel>
  <title>${rsstitle}</title>
  <atom:link href="${rssurl}" rel="self" type="application/rss+xml" />
  <link>${websiteurl}</link>
  <description><![CDATA[${rssdescription}]]></description>
  <language>${rsslang}</language>
  <pubDate>${rssmaxdate}</pubDate>
  <lastBuildDate>$rssbuilddate</lastBuildDate>
  <generator>mkrss.sh</generator>
  <webMaster>${rssauthor}</webMaster>
  <image>
    <url>${rssimgurl}</url>
    <title>${rssimgtitle}</title>
    <link>${websiteurl}</link>
  </image>
END

# write all items
cat $tmpdir/rss

# close the RSS file
cat <<END
</channel>
</rss>
END
} > "$rssfile"

# cleanup temporary directory
rm -rf $tmpdir
echo "RSS Generated"
#+end_src

** Full script
:PROPERTIES:
:CUSTOM_ID: full-script
:END:

Here is the full script I use:

#+begin_src bash
#!/usr/bin/env nix-shell
#!nix-shell -i zsh

# Directory
webdir="_site"
postsdir="$webdir/posts"
rssfile="$webdir/rss.xml"

# maximal number of articles to put in the RSS file
maxarticles=10

# RSS Metas
rsstitle="her.esy.fun"
rssurl="https://her.esy.fun/rss.xml"
websiteurl="https://her.esy.fun"
rssdescription="her.esy.fun articles, mostly random personal thoughts"
rsslang="en"
rssauthor="yann@esposito.host (Yann Esposito)"
rssimgtitle="yogsototh"
rssimgurl="https://her.esy.fun/img/FlatAvatar.png"

# HTML Accessors (similar to CSS accessors)
dateaccessor='.article-date'
contentaccessor='#content'
# title and keyword shouldn't be changed
titleaccessor='title'
keywordsaccessor='meta[name=keywords]::attr(content)'

formatdate() {
    # format the date for RSS
    local d=$1
    LC_TIME=en_US date --date $d +'%a, %d %b %Y %H:%M:%S %z'
}

finddate(){ < $1 hxselect -c $dateaccessor }
findtitle(){ < $1 hxselect -c $titleaccessor }
getcontent(){ < $1 hxselect $contentaccessor }
findkeywords(){ < $1 hxselect -c $keywordsaccessor | sed 's/,//g' }
mkcategories(){
    for keyword in $*; do
        printf "\\n<category>%s</category>" $keyword
    done
}

autoload -U colors && colors

tmpdir=$(mktemp -d)
typeset -a dates
dates=( )
for fic in $postsdir/**/*.html; do
    blogfile="$(echo "$fic"|sed 's#^'$postsdir'/##')"
    printf "%-30s" $blogfile
    xfic="$tmpdir/$fic.xml"
    mkdir -p $(dirname $xfic)
    hxclean $fic > $xfic
    d=$(finddate $xfic)
    echo -n " [$d]"
    rssdate=$(formatdate $d)
    title=$(findtitle $xfic)
    keywords=( $(findkeywords $xfic) )
    printf ": %-55s" "$title ($keywords)"
    categories=$(mkcategories $keywords)
    { printf "\\n<item>"
      printf "\\n<title>%s</title>" "$title"
      printf "\\n<guid>%s</guid>" "${websiteurl}/${blogfile}"
      printf "\\n<pubDate>%s</pubDate>%s" "$rssdate"
      printf "%s" "$categories"
      printf "\\n<description><![CDATA[\\n%s\\n]]></description>" "$(getcontent "$xfic")"
      printf "\\n</item>\\n\\n"
    } >>  "$tmpdir/${d}-$(basename $fic).rss"
    dates=( $d $dates )
    echo " [${fg[green]}OK${reset_color}]"
done
echo "Publishing"
for fic in $(ls $tmpdir/*.rss | sort -r | head -n $maxarticles ); do
    echo "${fic:t}"
    cat $fic >> $tmpdir/rss
done

rssmaxdate=$(formatdate $(for d in $dates; do echo $d; done | sort -r | head -n 1))
rssbuilddate=$(formatdate $(date))
{
cat <<END
<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0"
	   xmlns:content="http://purl.org/rss/1.0/modules/content/"
	   xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	   xmlns:dc="http://purl.org/dc/elements/1.1/"
	   xmlns:atom="http://www.w3.org/2005/Atom"
	   xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	   xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	   xmlns:georss="http://www.georss.org/georss"
     xmlns:geo="http://www.w3.org/2003/01/geo/wgs84_pos#"
     xmlns:media="http://search.yahoo.com/mrss/"><channel>
  <title>${rsstitle}</title>
  <atom:link href="${rssurl}" rel="self" type="application/rss+xml" />
  <link>${websiteurl}</link>
  <description><![CDATA[${rssdescription}]]></description>
  <language>${rsslang}</language>
  <pubDate>${rssmaxdate}</pubDate>
  <lastBuildDate>$rssbuilddate</lastBuildDate>
  <generator>mkrss.sh</generator>
  <webMaster>${rssauthor}</webMaster>
  <image>
    <url>${rssimgurl}</url>
    <title>${rssimgtitle}</title>
    <link>${websiteurl}</link>
  </image>
END
cat $tmpdir/rss
cat <<END
</channel>
</rss>
END
} > "$rssfile"

rm -rf $tmpdir
echo "RSS Generated"
#+end_src

The =nix-shell= bang pattern is a neat trick to have all the dependencies I
need when running my script, I could have added zsh, but my main concern
was about =html-xml-utils=.

Along my script I have a =shell.nix= file containing:

#+begin_src nix
{ pkgs ? import (fetchTarball https://github.com/NixOS/nixpkgs/archive/19.09-beta.tar.gz) {} }:
  pkgs.mkShell {
    buildInputs = [ pkgs.html-xml-utils ];
  }
#+end_src

If you are not already using nix[fn:nix] you should really take a look.
That =shell.nix= will work on Linux and MacOS.

[fn:lb] https://github.com/LukeSmithxyz/lb
[fn:hu] https://www.w3.org/Tools/HTML-XML-utils/
[fn:nix] https://nixos.org/nix