Project Gutenberg
The Project Gutenberg is a project gathering public domain books in different language, its web site is http://www.gutenberg.org. The purpose of this project is to create a sustainable solution to create a ZIM file providing the Gutenberg project ebooks in the similar manner like gutenberg.org
Goals
- A script (python/perl/nodejs) able to create quickly a ZIM file with all books in all languages.
- The data should be scraped from www.gutenberg.org.
- The texts should be available in HTML and EPUB.
- The ZIM should provide a simple filtering/search solution to find content (by author, language, title, ....)
One way to achieve it
- Retrieve the list of books is published by the Gutenberg project in XML/RDF format
- Parse the XML/RDF and put the data in a structured manner (memory or local DB)
- Download the necessary HTML+EPUB data from Gutenberg.org based on the XML/RDF Catalog in a target directory
- Create the necessary templates of the index web pages (For the search/filter feature, a javascript client side solution should be tried)
- Fill the HTML templates with the data from the XML/RDF and write the index pages in a target directory
- Run zimwriterfs to create the corresponding ZIM file of your target directory
First investigation results
Work done by didier chez google.com and cniekel chez google.com
The RDF index is at http://www.gutenberg.org/cache/epub/feeds/rdf-files.tar.bz2
wget works, contains 30k directories with each an rdf-file: every directory has 1 file with the rdf-description of one book.
Emmanuel suggests the scraper should download everything into one dir, then converting the data into an output dir, then zim-ifying that directory.
- Getting data
Gutenberg supports rsync ( rsync -av --del ftp@ftp.ibiblio.org::gutenberg /var/www/gutenberg ) so using that to get data seems most simple. In 2011, the archive was 650GB. That was source, the generated data: rsync -av --del ftp@ftp.ibiblio.org::gutenberg-epub /var/www/gutenberg-generated
If I cd gutenberg-generated, there is stuff like: ./10126/pg10126.rdf ./10126/pg10126.mobi ./10126/pg10126.xapian ./10126/pg10126.plucker.pdb ./10126/pg10126.qrcode.mobile.png ./10126/pg10126.log ./10126/pg10126.txt.utf8 ./10126/pg10126.qrcode.desktop.png ./10126/pg10126.epub ./10126/pg10126.txt.utf8.gzip ./10126/pg10126.converter.log ./10126/pg10126.qrcode.png ./10126/pg10126.qioo.jar
To get epub+text+html, you'll need both rsync-trees, which seems quite inconvenient.
So a caching fetch-by-url seems more convenient, the rdf-file contains the timestamp, which could be compared so updates to a book will be caught.
So a on-disk-caching, robots-obeying url-retriever needs to be made/reused. If you can somehow filter which books to fetch (language-only, book-range), that will be convenient
- Converting data
...
- Zimming
one of the rdflib readers for python: https://github.com/RDFLib/rdflib.git
The best Goobuntu packaged option seems to be: python-rdflib
Building zimwriterfs
sudo apt-get install libzim-dev liblzma-dev libmagic-dev autoconf automake git clone git://git.code.sf.net/p/kiwix/other kiwix-other cd kiwix-other/zimwriterfs/ ./autogen.sh ./configure make sudo make install
gives you
/usr/local/bin/zimwriterfs
Driver steps
rsync -av --del ftp@ftp.ibiblio.org::gutenberg-epub /var/www/gutenberg-generated
copy-data-to-outputdir
build-index
Scripting Stages
- [one-time only for mirror] rsync all data to zimfarm.kiwix.org/gutenberg
- Download & Extract rdf-files.tar.bz2
- Loop through folder/files and parse RDF
- Fill the Database with all data
- Query the database to reflect filters and get list of books
- Download the books based on filters (formats, languages)
- [MLB] Generate a static folder repository of all ePUB files
- Generate zimwriterfs-friendly folder of static HTML files based on templates and list of books.
- Generate zim file from static folder
- Prepare the templates
- Article template
- HomePage template
- Index template?
Next steps
One of the problem is that even on Gutenberg, we don't have all the most important books of the French litterature. We should help to fix this. Here is the coordination page.
