Rod's extensions for Pandoc, the universal document converter.


Pandox is a small collection of utilities that extend John MacFarlane's Pandoc, the "universal document converter".

The Pandox extensions are typically invoked by piping a JSON-formatted AST (generated via pandoc -t json FILENAME) in as stdin, and piping the transformed JSON-formatted AST to another pandoc invocation for rendering. For example:

pandoc -t json | coffee lib/ | pandoc -f json -t html

will generate an HTML version of the file, first applying the filter defined in up-caser.

The npm module includes directly executable scripts for each extension. Hence:

pandoc -t json | pandox-up-caser | pandoc -f json -t html

also works (following npm install -g pandox).

Pandox processes the JSON-format abstact-syntax tree that pandoc can generate when given the -t json flag. Internally, this is quite similiar to (but not exactly the same as) the Pandoc filters API that exists for Haskell and Python.

If you'd like to try your hand at writing custom JavaScript-based Pandoc filters, simply extend the PandocFilter class or supply a filtering method. For example:

var PandocFilter = require('pandoc-filter').PandocFilter
function upcase(key,value) {
  if(key==='Str') {
    return value.toUpperCase();
  } else {
    return value;
var filter = new PandocFilter(upcase);

HOWEVER one should not consider the API fully stable or settled just yet, so some of the semantics might change in future releases. We follow the semver version numbering conventions so it should be easy to tell when a breaking change is introduced, but the PandocFilter API will probably change moderately frequently until we're more satisified with it.

The CodeBlockProcessor extension adds several capabilities to the way in which Pandoc handles "fenced code blocks", such as:

This is a sample of text inside a "fenced" code block.

Pandoc supports several parameters that control the way in which a code block is rendered. The general form is:

This is a sample of text inside a "fenced" code block.


  • #THE-ID is used to identify the code block in things like HTML anchors and Latex cross-references.

  • .CLASS-ONEand .CLASS-TWO enumerate HTML classes to assign to the code block, and sometimes influence the rendering in other ways. For example, adding the class .numberLines will cause Pandoc to number the lines in the code block when rendering it.

  • NAME="VALUE" and NAME2="VALUE2" enumerate name-value pairs that can be used to modify the way in which the code block is rendered. For example, adding the pair startFrom=100 will cause Pandoc to number the lines starting with 100 rather than 1.

CodeBlockProcessor adds a few new parameters that can be controlled by name-value pairs.

  • input-file - replaces the body of the code block with the contents of the specified file.

  • input-cmd - replaces the body of the code block with output of the specified command..

  • exec - executes the body of the code block as it were a shell script

  • output-file - writes the body of the code block to the specified file.

  • output-cmd - pipes the body of the code block to the specified command.