In this post it is documented how this blog was constructed as a static site, meaning, the webpages are served from storage and not generated dynamically upon each visit using scripts (written for example in php) and databases (e. g., MySQL).

Why a static site?

The obvious and ubiquitous software choice for blogging is wordpress, a dynamic engine, which IMHO, is in several ways superior to other dynamic blogging services such as Google Blogger, Livejournal, Tumblr, etc., if nothing else for control of content.

It would have been natural to stop there and take that route. However in a more insidious manifestation of technophilia, a static blog was decided upon. There are many reasons, which all basically comes down to simplicity, control, cost savings, and, because I can do it.

The environment issue

In wordpress, everything is web-based, and the content storage, database and scripts are all intermixed. Editing must take place in the wordpress editor or in custom html/javascript/css/php. This is all too much detail for the writer to be bothered about, if s/he wants any degree of customization. Once the design is fixed, s/he should be able to focus on the content without any distractions, and if design is to be changed, it should be doable without disturbing the content files.

The architectural issue with wordpress etc. here is that they confound the production and development environments. These should ideally be separate for a neat workflow. Instead, the content writer is forced into undesired technical paraphernalia.

In contrast, static blogs compose their content using schema such as Markdown or reStructuredText, which consist of: (1), a plain text structured syntax for document markup, and, (2), a tool for conversion to HTML or other web markup languages. These are much simpler and much easier to control, and less susceptible to version changes.

There is no PHP or database setup required. Deploying is simply copying the files to a server, and so pre-publication testing is easy and can be done offline. Usually a small server script is provided alongside the static blog engine for testing. Otherwise a low-footprint server such as nginx can be run locally for this purpose.

Thus the problem of confounding development and production platforms is solved, with nice developer and testing sandboxes.

Also all this is very well-integrated with Linux, my favorite operating system, a very important reason. Command line and text editing is my pasture and I get to live in that habitat.

The hosting resources required for public deployment are very small, allowing one to use very cheap hosts such as NearlyFreeSpeech. Outside the domain registration costs, the running costs come to less than $1 per month, which is a great saving!

This migration to static blogs is gaining more traction and it is increasingly more popular, especially among developers and hackers (the latter not meaning computer security criminals, but, followers of hacker news). There are many static-site engines available by now, for example as described at staticgen.

From among these, I was highly impressed by hexo based on Node.js. Nevertheless pelican was chosen instead, for, although being older, it is based on Python, a language which is of interest to me for a long time, given its ease of reading, comparatively easy coding via libraries and above all its popularity for scientific and large-scale computing, in adapting to platforms such as SAGE. The other static blog platforms such as Jekyll and Octopress were discarded because I do not have time to learn another involved language such as Ruby, when I had already been fixated on Python. For good or bad, one can't vacillate for too long and pelican it is for the near future.

Markdown was chosen vs reStructuredText because various math and programming related sites such as Stackoverflow, Math Stack Exchange and Github use it, and I am used to posting in it for this reason.

Setting up pip and virtualenv

It is expected that:

  • The user is proficient in Linux and the terminal.
  • User is proficient in bash and can work with python if necessary.
  • User knows utilities such as git, make, ssh, rsync etc., and is familiar with web servers and markup.

Useful references are: for a general UNIX programming environment: Kernighan and Pike1 and for bash usage and scripting: Robbins and Beebe2.

pip is the python package installer. It takes care of all the Pythonic dependency hell. Use:

~$ sudo apt-get install -y python-pip python-virtualenv virtualenvwrapper

We do not want to do any global python package installations from now onwards. The method is to use virtual environments in which the user installs packages inside a local directory and runs them through wrapper scripts. There can be multiple such environments and the user can switch back and forth between them, with parameters being different each time. This gives much flexibility for development.

Now we create the virtualenv named pelican for our blogging purposes.

$ mkvirtualenv pelican
Running virtualenv with interpreter /usr/bin/python2
New python executable in pelican/bin/python2
Also creating executable in pelican/bin/python
Installing setuptools, pip...done.

Note the change in the bash prompt. Use workon and deactivate to activate/deactivate the python environment:

~$ workon pelican
(pelican)~$ deactivate

This sequence will be part of the blogging workflow whenever required.

Installing pelican, themes and plugins

Install pelican:

~$ mkdir ~/work/blog
~$ cd ~/work/blog
~/work/blog$ workon pelican
(pelican)~/work/blog$ pip install pelican

It downloads and install many packages and finishes with:

Successfully installed pelican jinja2 blinker pygments docutils python-dateutil pytz six feedgenerator unidecode markupsafe
Cleaning up...

Now we have pelican installed!


Install additional packages:

(pelican)~/work/blog$ pip install Markdown beautifulsoup4 typogrify Pillow webassets

Create directory pelican and run pelican-quickstart to set up the pelican blogging platform inside it:

(pelican)~/work/blog$ mkdir pelican
(pelican)~/work/blog$ pelican-quickstart 
Welcome to pelican-quickstart v3.5.0.

This script will help you create a new Pelican-based website.

Please answer the following questions so this script can generate the files needed by Pelican.

> Where do you want to create your new web site? [.] pelican
> What will be the title of this web site? Notions and Notes
> Who will be the author of this web site? George S.
> What will be the default language of this web site? [en] 
> Do you want to specify a URL prefix? e.g.,   (Y/n) 
> What is your URL prefix? (see above example; no trailing slash)
> Do you want to enable article pagination? (Y/n) 
> How many articles per page do you want? [10] 
> Do you want to generate a Fabfile/Makefile to automate generation and publishing? (Y/n) 
> Do you want an auto-reload & simpleHTTP script to assist with theme and site development? (Y/n) 
> Do you want to upload your website using FTP? (y/N) 
> Do you want to upload your website using SSH? (y/N) y
> What is the hostname of your SSH server? [localhost] notionsandnotes
> What is the port of your SSH server? [22] 
> What is your username on that server? [root] 
> Where do you want to put your web site on that server? [/var/www] 
> Do you want to upload your website using Dropbox? (y/N) 
> Do you want to upload your website using S3? (y/N) 
> Do you want to upload your website using Rackspace Cloud Files? (y/N) 
> Do you want to upload your website using GitHub Pages? (y/N) 
Done. Your new project is available at /home/george/work/blog/pelican

Note that, here, the full details of the ssh server are not provided. In my case, it is stored instead in the ~/.ssh/config file, as per ssh config directives. Replace it as necessary.

We do typesetting using Markdown. Under the ~/work/blog directory, we store the site files in various subdirectories of the directory raw. The published site is stored under another directory As mentioned earlier, pelican directory stores the files for pelican. So we have three subdirectories in total for the blog directory.

Now set up these two additional directories and also initialize git to enable version control and rollbacks (if required):

(pelican)~/work/blog$ mkdir
(pelican)~/work/blog$ mv pelican/content ./raw
(pelican)~/work/blog$ git init .
(pelican)~/work/blog$ git add .
(pelican)~/work/blog$ git commit -m "Initial commit"

Now to use pelican themes and plugins, clone these from github as submodules:

(pelican)~/work/blog$ cd pelican
(pelican)~/work/blog$ git submodule add themes 
(pelican)~/work/blog$ git submodule add plugins
(pelican)~/work/blog$ cd ..
(pelican)~/work/blog$ git -commit -m "Themes+Plugins"

Configuration file

There are two config files, and We mostly work with the latter. Both are python files. Set the following configuration directives:

PLUGIN_PATHS = ['./plugins']
PLUGINS = ['extract_toc','render_math','disqus_static','better_figures_and_images']
MD_EXTENSIONS = ['codehilite','extra','smarty', 'toc']

We have included plugins for: table of contents, LaTeX rendering of math, etc., and similar Markdown extensions are added too. The toc Markdown extension is for example required for the extract_toc plugin. The "Better Figures and Images" plugin is taken from Duncan Lock, whose site was most helpful for this pelican setup. Plugin assets is for minification of javascript etc..

Now set the directives:

PATH = '../raw'
STATIC_PATHS = ['extra', 'images', 'pdfs']
    'extra/robots.txt': {'path': 'robots.txt'},
    'extra/favicon.ico': {'path': 'favicon.ico'},
    'extra/htaccess': {'path': '.htaccess'}

Create a favicon and robots.txt as required. A custom 404 page can be embedded via a .htaccess configuration directive. This is ours!

We use the built-texts theme and the following directives are added as data for it:

THEME = 'themes/built-texts'


Edit the Makefile as follows:


Add in other parameters, such as the ssh hostname, as required.


First, enable the virtual environment 'pelican' using 'workon' (and remember to 'deactivate' it later. Posts are created and edited using plain text editors in files with *.md extensions, in Markdown format. For example, these are the first few lines of the Markdown for this page:

Title: How this blog was set up using Pelican
Date: 2015-04-10 
Category: Tech
Subcategory: Blogging
Tags: pelican, python
Slug: pelican-setup


**Table of contents**


In this post it is documented how this blog was constructed using a static 
site maker, meaning, there is no server-side processing of data.

More details are available at the Pelican Documentation website. The whole source of this website is uploaded to a github repository where more examples and details may be found.

After the articles or pages are edited, the compilation and static site generation are done by make html. Running make devserver will fire up a local server at http://localhost:8000, which can be used for pre-publication viewing. Note that this will require fixing up of the file in addition to which was edited earlier. In particular, relative URLs to be enabled in pre-publication mode in order for us to be able to follow links in the browser.

After viewing, Ctrl + C will stop the server and return to shell.

For publication, run make publish, and later upload via make rsync_upload, which uploads only the necessary files using rsync + ssh.


Here are the screen captures of a first built of this site with a (slightly modified) built-texts theme, with some css additions. The captures were done using a Google Chrome extension by Peter Cole.

Index page screenshot:

index.html Click for expanding the thumbnail to actual size.

Below is the next screenshot of how the article on Lebesgue measure and construction via Caratheodory extension theorem, was rendered.

index.html Click for actual size image.

Once or twice for small customizations I got stuck and two guys in the pelican IRC chat helped me out. Thanks, guys (if you are reading this)!

Moving to pelican-bootstrap3 theme and other improvements

The site by now looks good and is workable. I was however not satisfied yet.

The reason is that the web technologies are progressing fast and the current trend is responsive and mobile and tablet-ready pages. The built-texts theme, although very good, was not up to date. Therefore it was decided to replace it with a more modern and responsive theme. Some of the style customized for the PC screen may be lost, but the viewability in mobile and touch devices and the more modern font and design are positives.

This time the pelican-bootstrap3 theme was chosen. There were a few more responsive themes available; but this one was the most used and supported.

Quite a good amount of tinkering was done to finally make this satisfactory. While the comparatively new bootstrap3 theme was more up-to-date, I was still influenced by some of the elegance of the older built-texts theme. A few additions were my own ideas, too, on top of existing changes done by others, among the foremost of which are Tyler Hartley's and Christine Doig's sites.

The most important decision was to use the Bootswatch theme readable. This site is meant to be for expository and rigorous articles on pure mathematics and it is of foremost importance to ease the user's eye when staring the screen for a long time.

There were a lot of settings for the pelican-bootstrap3 scheme and these were pushed into a separate file and called via an import directive. So, in that file, the setting was:

BOOTSTRAP_THEME = 'readable'

The menus on the top navbar were added by a small bit of pythonic tinkering in the Jinja2 template:

(pelican)~/work/blog/pelican$ vim plugins/pelican-bootstrap3/templates/article_list.html

Write something like:

  {% for cat, cat_articles in categories %}
    <li class="dropdown {% if cat == category %}active{% endif %}">
    <a href="#" class="dropdown-toggle" id="menu-{{ cat }}" data-toggle="dropdown" >
    {{ cat }}
    <span class="caret"></span>
     <ul class="dropdown-menu" role="menu">
     {% for cat_article in cat_articles %}
        <li><a href="{{ SITEURL }}/{{ cat_article.url }}">{{ cat_article.title }}</a></li>
     {% endfor %}
  {% endfor %}
{% endif %}

Similarly, make a menu on the right side of navbar for the pages was added.

Other additions

Another thing to be enabled soon was a commenting system via disqus. This is explained on the documentation page.

The save pages were modified, and subcategories were added. A little change in the code was required for this to enable a new `subpath' metadata, as in this github patch.

Some Jinja2 alternations made are: a custom article sidebar and individual sidebar for each article in the indexes, different sidebar for pages, some font-awesome and glyphicon additions, etc.. This was altogether done in a haphazard manner as this is my first encounter with any templating language.

The readable bootswatch theme had a problem, of having black links and so it was impossible to locate the links. So link colors were modified to a shade of blue using the custom.css file, which was included in as EXTRA_PATH_METADATA. Here is the whole source.

Some color and other style changes in were incorporated into extra.css. For example, the "more ..." button text for article summary ending in indexes was highlighted and enlarged. In templates/article_list.html:

<br><a class="btn btn-default btn-small" id="readmore" href="{{ SITEURL }}/{{ article.url }}"> Read more ...</a>

and in extra.css:

#readmore {
  font-size: 17px;
  font-weight: bold;
  color: #003380;
  background-color: #CCFFCC;

A few more things were added to extra.css, to change the heading font style to small caps, to change some foreground and background colors, to change font sizes for some cases, and to highlight code differently.

To the Jinja2 templates, a few other additions were made too, like a custom footer with colophon, background image, a tag cloud, and a tweets widget. Funnily, more than an hour was spent trying to fix this widget, only to realize in the end that the browser adblock plus extension was blocking it!

The whole source code is available on a github repository for examination if required. All the customizations so far were all implemented by self, with one notable exception that required intervention of a professional developer:

Barry Steyn updates pelican render_math plugin for index pages

This is supposed to be a pure mathematics blog, and so display of math is extremely important.

It was implemented using the very helpful render_math plugin to display math. It made use of MathJaX to display mathematics encoded in the LaTeX typesetting markup language.

This had one small glitch however. The article summaries in index.html of the base directory, or other indexes for authors, tags, categories and so on, failed to render math correctly using MathJaX.

That was how I made my first post in Github. I had originally imagined that I must have missed something trivial and it was a bold step for me. The author, Barry Steyn, responded promptly and offered a solution, which he also said was not completely trivial like I feared before daring to submit an issue. Moral: Don't be so shy!!

He kept to his word and updated the plugin which I cloned and verified to be working.

Thanks a lot, man!! This addition beautifies me (and this site) a lot, considering the future plans to upload many more Math articles.

See also: Barry Steyn's pelican LaTeX plugin page.

Final comments

It took altogether longer than expected, but was an intensive and beneficial (hopefully) journey. This page was titled, Long Road, to be reminiscent of Grothendieck's Long March, but of course no claims to such greatness or even mere competence are made here!! And, as for the reference, this is certainly meant to be a pure mathematics focus blog and one might go the whole mile and allude to mathematical references even in tech posts!

This website domain was registered on an impulse on July 11, 2014, but due to intervening turbulent personal life, it took me up to April 2015 to turn back to finally act upon site building. So, maybe, it was indeed a long march!

Any suggestions for improvement are welcome, from web designers and pelican aficionados in particular. Thank you!


The end (Or is it?)



(Book references added on 2016 September 13.)

Amazon links

Symbols of the form Read reviews or buy at Amazon are links to Amazon, when relevant/available.

  1. Kernighan, B. W. and R. Pike, The Unix Programming Environment Read reviews or buy at Amazon, Prentice Hall, 1984. 

  2. Robbins, A. and N. H. F. Beebe, Classic Shell Scripting: Hidden Commands that Unlock the Power of Unix Read reviews or buy at Amazon O'Reilly Media, 2005. 


  • Avatar

    Posted on

    Really well put together article.

    As an aside, you can also use GitHub Pages for hosting, which is free, and then integrate it with Travis-CI to automatically publish the blog (basically run pelican to generate the output and push the changes back online) in order to decouple the actual writing of blog posts from the publishing part.

    The above also has the advantage of enabling a history of changes done (both for the articles themselves and the output), as well as simplifying things if you want to have guest posts and so on.

  • Avatar

    Posted on

    Really enjoyed the article. I found it pretty useful and love the math

  • Avatar
    Sam Banks

    Posted on

    Thanks George, this provided a great easy start with Pelican.

  Add a Comment

Format the comment in plain text or HTML, Markdown or reStructuredText.

Comment Atom Feed

© 2015 -- 2016 George S. of