Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Halwa - A Static Site Generator

Halwa is a single file static site generator written in Python. It's light-weight and the only dependencies are jinja2, hamlish-jinja, and markdown.


Using Halwa is as easy as it should be; install it, then create a configuration file and point Halwa to it:

pip install halwa

python -m halwa


Halwa is organized around two main things, content and processors. A piece of content is a data source (static HTML, HAML, or Markdown) which is eventually converted into an output page (or pages). Halwa uses the following workflow:

  1. Initialize a global data dictionary.
  2. Load each piece of content specified in sources.
  3. Run each of the processors specified in processors, generating output that goes into the data dictionary.
  4. Update each piece of content with the contents of the data dictionary.
  5. Render each piece of content, writing it out to the output folder.

A sample configuration is available at


The configuration file is a pure python file. The following variables must be set:


This is a dictionary that must contain two keys, one for the the templates directory, and another for the output directory.


A dictionary containing route definitions for various pieces of content. The key is the route name, and the value is a format string (which will be filled in by the content's metadata). For example, the route '/blog/{tag}' will have '{tag}' replaced by the tag field in the content's metadata.


A list of 4-tuples, each describing a piece of content. A sample source tuple is below:

('Page', 'content/blog*.haml', {'blogsidebar' : 'sidebar'}, ['tags'])

The entries are as follows:

  1. The type of content.
  2. The path for source files that should be used (any 'glob' expression is valid here).
  3. The set of mappings, each key in the global 'data' object will be mapped to the corresponding key given here, for the purposes of this content object; before the content's data field is updated.
  4. A list of keys in the global data object that this content object depends on. This is used for resolving caching dependencies; since templates may depend on arbitrary data items.


A list of tuples, specifying which processors should be run. A sample entry is below:

('PostArchives', {'key' : 'blogarchives'})

The first item is the name of the processor to run, the second item is a dictionary which is used to provide settings for the processor.


The initial dictionary to be used to fill the data dictionary.


There are multiple types of content that can be defined. They are described below.


This will copy the source file to the corresponding path in the output directory.


This is the base class for all the following content objects. It will read the input file, and parse the JSON dictionary at the top, which will be set as the content's metadata. The rest of the data will be parsed as content. It expects a route attribute in the metadata at the very least, to generate the output URL for the content. It will render the page as a Hamlish Jinja template and write it to the appropriate directory.


It will simply render the given template. It is intended for simple pages (such as an index).


This will take each tag in the tags dictionary that has been passed in, and render the template page with that tag.


This will take each year in the readerscorner dictionary that has been passed in, and render a copy of template page for each month in that year.


This will automatically generate a slug for the content, based on the date in the metadata and the filename. The content will be passed through markdown before being passed to the Hamlish Jinja renderer.


These process the data dictionary after all the metadata has been loaded from the content; and then put the processed results back in so that dynamic pages can be based on those.


This processor will look for the key attribute in each post (e.g. "tags"); which should be a list of values. It will make a dictionary where each unique value maps to the list of posts that contain this value. The final output will be stored in the datakey attribute of the data dictionary. It will sort the posts by the sortkey attribute (in reverse order if reverse is true).


This processor will take the routes given in the routes attribute of the data dictionary, the tags given in the tags attribute of the data dictionary, and create a dictionary that contains the routes, their URLs, and a list of tags with their counts.


This processor will generate a list of posts and put them in the key attribute of the data dictionary. It will have count posts, sorted by sortkey (in reverse order if reverse is true), use uniquekey to determine unique posts, apply each function in filters to see which posts should be included (it should pass all filters); and exclude each post in exclude.


This processor generates post archives, saving them in the key attribute of the data dictionary. The archive is a dictionary (of years), where each year is a dictionary (of months); and each month is a dictionary of posts in that month, sorted by day (in reverse order if reverse is true).


This processor generates the readers corner feed, saving it in the key attribute of the data dictionary. It will also generate a sidebar with links (using the provided route) for each year and month, saving it in the sidebarkey attribute of the data dictionary.

The archive is a dictionary (of years), where each year is a dictionary (of months); and each month is a dictionary (of days); each day is a dictionary of links, sorted by time (in reverse order if reverse is true).


This processor generates an RSSFeed, putting it in the key attribute of the data dictionary. It will put count posts, sorted by the sortkey attribute (in reverse order if reverse is true).


This processor generates a list of URLs and stores it in the key attribute of the data dictionary.


Halwa has support for some simple caching to make incremental site builds faster.

When content is initially read from disk, the loaded metadata is stored in the cache. Later reads of the metadata will come from the cache if the file on disk has not been modified.

Before content is written to disk, its dependencies are checked (the source file on disk is a default dependency). If any of the dependencies have been modified at a time later than the time the existing output file (if any) was written, the output file will be re-generated. Otherwise it will be left alone.

The caching manager can not know whether any of the templates or processors depend on some user-defined keys in the data dictionary. In order to help it out, the dependencies list can be filled in. This is especially useful for specifying which templates a piece of content depends on (if you're using jinja template inheritance) as then it will re-generate output when the source template changes. Templates are added automatically to the cache, with the key "TEMPLATE_DIR/filename".


Single file static site generator written in Python







No releases published


No packages published