Howto use blogcxx


A static blog generator, written in C++17. It is a fork of blogcpp

First things first: How can I get it?

git clone

This is my public (but not main repository). It is used to publish released versions and for bug/issue tracking.

How do i use blogcxx?

My goal was to remove php from my server. I just have a somehow static blog. So i used for some time dokuwiki. This stores its data in files but still uses php.

In the i meet CrimsonKing aka tux_ on bitbucket. He introduced his blogcpp.

I setup my git repository for my blog site that has an post-update hook to run the static blog generator. The output outdir of the blog is the DocRoot of the webserver. The output gets mailed to me to get feedback. When i push data to the repo the post-update hook runs and updates the blog.

blogcpp seemned perfect for the issue but it was unstable and crashed repeatedly.


As blogcpp is licenced under WTFPL2, i fork it and do what the f*ck i want. As i rewrote so big chunks, i licence blogcxx under GPLv3. I think GPL has done much more good than harm to the open source movement.

WTFPL 2 and GPL3 are compatible.

Why forked?

blogcpp was heavyly rewritten because there were internal source code difficulties. ( There was a pattern used called spagetti code. ) Take a look at blogcpp.cpp can you follow easily where the global varibales are locked? How is the programflow?

Aside from this, it didn’t handle it well when there was no site dir or it was executed in an empty directory. It just segfaulted with no error message at all. The setup of a new blogsite was painful if you dont know the internals. The final reason to rewrite was that it just crashed with no help for the user. It is a very bad first expierience for the user of a project if it just crashes without printing anything.

As i tried to send my changes to blogcpp, the author didn’t want them because i use boost and added cmark-gfm. They are to heavy for the author of blogcpp.

No, seriously, why?

  • Just like other static site generators, blogcxx is virtually unbreakable by intruders as your server is not required to host anything but static HTML files - even JavaScript and CSS are, depending on your theme, entirely optional.
  • Unlike other static site generators, blogcxx is written in modern C++, resulting in a fast build speed, reducing the need for workarounds like incremental builds or advanced caching mechanisms while compiling. Still, blogcxx tries to use as many CPU cores as possible to achieve the maximum speed on your machine.
  • blogcxx aims to provide a decent feature set without requiring too much configuration; in fact, it would work right out of the box if your website was (which is, admittedly, rather unlikely).

What changes from a user perspective from blogcpp to blogcxx?

  • usage should be much more convienent as it reports its error much better
  • syntax and other stuff stays the same 🙂
  • more stable
  • better markdown support with libcmark-gfm
  • get source from git and not from mercurial
  • template engine got replaced by htmltemplate++, like perl HTML::template but it is a straight forward simple implementation to get rid of json and inja because inja didn’t work on freebsd. It was written from me.

Current version

The latest released version is 1-alpha. master branch is always the latest version.

what i am not sure if it works

  • highlight programming languages. The generated code looks correct.
  • colorize output for warn/error/fatal messages. (on windows, linux works)
  • ExcerptParser seems confused

TODO list + What is not yet working

  • metadata in the HTML (opengraphimg, feedmeta)
  • spawn $EDITOR on UNIX
  • OpenGraph support if you feel like sharing stuff on Facebook or whatever.
  • comments
  • download the highlight.js into the static folder
  • config collection: check the pathes if they are unique

Features common between blogcpp-9 and blogcxx-1

  • Markdown support: Write your articles and pages in Markdown, get HTML.
  • Theming support: You don’t like the available themes? Well, nor do I. Go on and make a better one!
  • RSS feeds for new articles, tags and categories.
  • Sticky articles: Just put Sticky: On on top of your article file.
  • Emoji support: Write :-) and get a real smiling face instead.
  • Commenting support: Even a static blog needs a community. 😉
  • Syntax highlighting via highlight.js.
  • Article series: If you have articles which continue other articles, you can just add them to the same series and blogcxx will make them refer to each other.

Features added compared to blogcpp

  • switch the markdown parser at compiletime to libcmark/libcmark-gfm but the blogcpp markdown parser is still available. Markdown syntax used with cmark-gfm
  • blogcxx can now run without an blog.cfg. It uses internal defaults (the one from the blog.cfg-example). Just use the blog.cfg file to override what you want to change.
    • The -help options displays also the file options
  • handle gracefully UTF BOMS in read files
  • use the configured locale everywhere.
  • All messages that get printed go through boost::locale::translate(). That means, i18n will be done in no time. All arguments are positional like %1%. different languages can rearrange the order of the parameters.
  • Articles can be parsed with hardbreaks on to format code and so on Hardbreaks: On (default is off). If hardbreaks are on, If you make new line, the generated html has a new line.
  • a folder media gets copied into the outdir.
  • colorize output for warn/error/fatal messages. This works on Unix and Win10+
    • white = standard
    • yellow = warning
    • red = error
    • cyan = fatal
  • blogcxx can now create a howto page if no blog exists yet. It also contails the default template for that issue. The internal tool bin2cxx creates cpp and h files that the cxx arrays get included in the executable.
  • You can add your own comment template. No recompilation needed. Just add a new mycomment.txt template to your comments folder.

Features changed compared to blogcpp

  • cxxopts replaced by boost::program_options (users dont care)
  • icu replaced by boost::locale (but still used as a backend on Linux) (users dont care)
  • Configurable article permalinks.
  • template engine got replaced by htmltemplate++, like perl HTML::template but it is a straight forward simple implementation to get rid of json and inja because inja didn’t work on freebsd

other good things

  • all compilation is now consistent on all platforms
  • static executable size is now just 2-3 MB
  • all path handling (including the url pathes) are handled with std::filesystem.
  • option added to switch the API for variant(std::variant/boost::variant) and filesystem (std::filesystem/std::experimental::filesystem and boost::filesystem)
  • option added to link all static
  • Installation option for blogcxx added

Features removed compared to blogcpp

  • js plugins: I think they are unneeded and were used to work around the codequality of blogcpp. If you need them, i could create a pluginsystem. This pluginsystem would have a js plugin that loads your js plugins.
  • google+ comment system: google+ closes soon
  • OEmbed support
  • maybe i will remove the comment system (why use a static web page and then hand the data of the users to others?)

Description of the internal changes (blogcpp-9 -> blogcxx-1)

In blogcpp global variables were accessed (non const) by multiple threads. No separation of collection and generation of html. The concurrency is now like Map and Reduce. This is the job of the

template <typename KeyT, typename RetValT>
class AsyncWorker;

The KeyT is used to identify the results. The RetValT can be void. No more raw threads. use std::future() and std::async(launch::async,…)

  • clangformat everything. Perfectly formatted code makes it easier for maintainance and future expansions.
  • make a metadata structure that is after the FilesToMetadata phase const
  • reduce #ifdef #else #endif over the whole codebase, but still keep the options available. This improves readability and maintainable a lot.
  • TemplateWrapper does now what its name implies. It uses the so called PIMPL pattern that is also known as compiler firewall.
  • replace struct tm by boost::posix_time::ptime over the whole code base
  • Restructure the whole src folder. Split the tasks into manageable files that are grouped accordig to their function in the process.
  • use boost::format for logging.
  • RSS (replaced tinyxml2 with htmltemplate++. we already have a templating engine)

No more

    #ifdef WITH_DEBUGLOG
        stringstream ss_debuglog;
        ss_debuglog << "Gathering ...";

Now it is typesafe and without sstream everywhere

    // these are variadic templates 🙂
    LOG_DEBUG("Gathering files from %1%.", file.string());

What stayed the same between blogcpp and blogcxx?

  • Markdown Parser (but exchangeable with libcmark-gfm)
  • Emoji Parser (clear(), but parse got reimplemented)
  • Excerpt Extractor
  • Fileparser (mostly)
  • SingleItem structure (mostly)
  • Syntax in the pages/post files

Used libraries

blogcxx uses the following third-party libraries (in 3rdparty/):

Not in 3rdparty directory because it is to big.

  • boost (ICU (as Boost::Locale Backend on Linux))

“…one of the most highly regarded and expertly designed C++ library projects in the world.” — Herb Sutter and Andrei Alexandrescu, C++ Coding Standards

Used tools

  • cmake
  • a c++17 compiler. known to work
    • Visual Studio 2019 Community Edition
    • gcc 8.2
    • clang 6


How can I use blogcxx?

The easier way:

Just start blogcxx in your blog directory. It will create a default blog with an howto article. (basicly this Readme). It will ask you for every step. 🙂

Use blog.cfg to override defaults.

  • blogcxx --new-article or blogcxx --new-post to start a new article or
  • blogcxx --new-page to start a new page. When you’re done, run
  • blogcxx and everything will be there soon.

What is a post, what is a page?

A post is an article that you post on your blog. It can contain tags, categories and be part of an series.

A page is a information page: It is added to the page navigation bar and it is always there. It get used for download pages, impressum and so on.

The nerd way:

Set up your blog.cfg file first, follow the documentation in the blog.cfg-example file in this repository. Then start to add Markdown-formatted posts (in /posts) and/or pages (in /pages) formatted like this:

Author: Your name here.
Date: 2016-01-01 12:00:00
Title: This is the title.
Tags: tag 1;tag 2;tag 3
Categories: some;random;categories


(Everything that follows will be a part of the contents.)

When you’re done, run blogcxx and everything will be there soon.

Which meta data types are allowed?

You mean, except the Title/Tags/Date part? Well:

  • You can additionally set Markdown: Off to disable Markdown parsing for this specific post or page. Note that pages don’t have categories, every category added to a page will be ignored.
  • If you want your newly created article (for logical reasons, only posts are supported) to be “sticked” to the top of your index page, just add Sticky: On into its header part.
  • Articles will always be sorted by their creation date; you can indicate the latest change (which will be displayed or not, depending on your theme) with Changed: yyyy-mm-dd hh:mm:ss somewhere above or below the Date: field.
  • If you don’t want to have emojis on this post or page without removing them everywhere, just add Emojis: Off.
  • In case you want to have a specific OpenGraph image for this page or article, you can set OpenGraphImage: http://my/host/image.png or whatever you want it to show.
  • You can also disable comments per post (pages don’t have comments): Comments: Off.
  • If you use Content plug-ins but you want to disable them for this item exclusively, you can set Plugins: Off.
  • If you want to add an article to a certain series, it is enough to set Series: Name Of Your Series here. Note that currently only one series is supported per article.
  • Pages are usually sorted alphabetically. If you don’t want to do that, you can set the Position parameter with any number larger than 0. Note that the “start page” is not a real page, so the template decides where to put it.
  • Articles can be parsed with hardbreaks on to format code and so on Hardbreaks: On (default is off)

OK, but how can I compile blogcxx first?

It’s not hard to compile blogcxx yourself though:

Boost API is most suitable for ‘older’ compilers like gcc 6.3 and will build on almost every compiler.

blogcxx has been proven to compile on:

OS Compiler Status filesystem variant tested
Windows MSVC 2019 OK std std 1-alpha
Windows MSVC 2019 OK boost std 1-alpha
Windows MSVC 2019 OK std boost 1-alpha
Windows MSVC 2019 OK boost boost 1-alpha
Windows MSYS 8.2 std std
Windows MSYS 8.2 boost std
Windows MSYS 8.2 std boost
Windows MSYS 8.2 OK boost boost 1-alpha
Stretch gcc 6.3 FAIL std std 1-alpha
Stretch gcc 6.3 FAIL boost std 1-alpha
Stretch gcc 6.3 FAIL std boost 1-alpha
Stretch gcc 6.3 FAIL boost boost 1-alpha
Buster gcc 8.2 OK std std 1-alpha
Buster gcc 8.2 OK boost std 1-alpha
Buster gcc 8.2 OK std boost 1-alpha
Buster gcc 8.2 OK boost boost 1-alpha
FBSD 12 clang 6 OK std std 1-alpha
FBSD 12 clang 6 boost std
FBSD 12 clang 6 std boost
FBSD 12 clang 6 boost boost

Build on Windows:


  • Visual Studio 2019
  • cmake 3.13
  • Boost 1.68.0

Manual compilation instructions:

  • Get boost from and unzip it to your build directory.
  • Start a Commandline prompt from Visualstudio VS2019 x64 Native Tools-Commandprompt
  • cd into …/boost-1.68.0
  • bootstrap.bat
  • b2 -j8 --build-type=complete address-model=64 --stagedir=stage64 stage Depending on your machine it takes about 15 min.

Start cmake gui:

  • setup source directory to …/blogcxx-src and build directory to …/blogcxx-src/build
  • configure: choose Visualstudio 2017 x64
  • Add an Entry: BOOST_ROOT (path) pointing to …/boost-1.68.0
  • Add an Entry: BOOST_LIBRARYDIR (path) pointing to …/boost-1.68.0/stage64
  • Press generate.
  • you can choose the options and generate again (options see below) Here i recommend STATIC_LINKING=ON (just on windows)
  • press “Open project”
  • choose the build type. Probally Release
  • build

Build on Linux/FreeBSD:


  • clang6 or gcc 8.2
  • cmake 3.9
  • Boost 1.62.0

Build boost 1.68.0 on Linux

Install cmake and gcc/clang from your package manager.

Build boost:

  • cd …./boost/1.68.0
  • ./
  • time ./b2 -j8 –build-type=complete –layout=tagged –stagedir=stage64 address-model=64 stage

This took on my machine (Amd 8350) about 8 min 3179,97s user 212,00s system 730% cpu 7:44,02 total

build blogcxx

  • cd …/blogcxx-src
  • mkdir build
  • cd build
  • export CC=/usr/bin/clang-6.0
  • export CXX=/usr/bin/clang++-6.0
  • cmake -DBOOST_ROOT=${HOME}/Dokumente/boost/1.68.0
  • ccmake .. (now choose your options)
  • time make -j8 clean all

It took again on my machine (Amd 8350) about 1 min make -j8 clean all 278,61s user 7,67s system 520% cpu 55,023 total

Optional preprocessor definitions while compiling:


  • blogcxx is the traditional and default MarkdownParser.
  • libcmark-gfm is the implementation used at github with table support


  • std::filesystem
  • boost::filesystem


  • std::variant
  • boost::variant


Create a static binary.

Which directories need to be there?

Here’s a site’s minimal directory structure:

  • ./templates/TPLNAME/
  • ./INDIR/posts/
  • ./INDIR/pages/
  • ./OUTDIR/

This directory structure can be created by blogcxx for you 🙂

Of course, the capital letters mean that the values are indeed variables. By default, TPLNAME is default, INDIR is site and OUTDIR is outdir. Please use the configuration file when building your site in order to change these values.

How does syntax highlighting work?

Four spaces at the beginning of a line mark the beginning of a code block. blogcxx tries to determine the language of the code automatically. If you want to set the language explicitly, you can use the Pelican syntax for code blocks; example:

 print("This is Python code.")

Which parameters are supported?

You can call blogcxx with the following parameters:

  • -h / --help
    • Prints this list of parameters. It contains the description of all parameters which are part of the blog.cfg-example
  • -v / --version
    • Prints the version number of blogcxx.
  • --console-verbosity --file-verbosity
    • Generates your site, but set the amount of info you want to see. Quiet = 5 / Verbose = 0
  • --new-article
    • Generates a new article stub and opens it in your default editor.
  • --new-page
    • Generates a new page stub and opens it in your default editor.
  • --config [filename]
    • Uses a different configuration file than blog.cfg. The file must exist.

Which emojis are supported?

Given that you have actually enabled Emoji support in your blog.cfg file and not disabled it for the page or article you want to process, the following smiley codes will automatically be replaced:

Code Emoji
;-) 😉
:-D 😀
:'( 😭
``:- ``
>:) 😈
>:-) 😈
>:( 😠
>:-( 😠
:-* 😘
:-O 😮
:-o 😮
:-S 😕
:-s 😕
:-# 😶
0:-) 😇
:o) 🤡
<_< 😒
^^ 😊
^_^ 😊
m( 🤦

Which comment systems does blogcxx support?

While blogcxx does not have its own commenting system, you can easily integrate existing external commenting systems via your blog.cfg (please refer to the blog.cfg-example file in this repository). Currently supported are:

  • Disqus (comments = disqus, you need to set your commentdata to your Disqus shortname)
  • isso (comments = isso, you need to set your commentdata to your isso folder)
  • Hashover (comments = hashover, you need to set your commentdata to your Hashover folder) – currently, version 2.0 only
  • Discourse (comments = discourse, you need to set your commentdata to your Discourse domain)
  • Commento (comments = commento, you need to set your commentdata to your Commento server URI)

Can I use raw HTML in my Markdown-enabled article or page?

blogcpp markdown parser

Yes, you can! Everything between <rawhtml> and </rawhtml> will be ignored by blogcpp Markdown parser.

cmark-gfm markdown parser

Inline HTML with cmark-gfm

In short, just add the raw html to the Markdown and it will be embedded. It must just be correct. Opening and closing divs and so on.

Which files does a template need to work?

blogcxx needs index.txt, post.txt, page.txt and archives.txt in order to be able to process your site correctly. All other template files are optional. CSS and image files can be put into a subfolder named static, blogcxx will automatically copy this folder to your output directory then.

Categories blogcxx howto
Tags howto