Fuzzing with Fandango – Task Sheet

Your Task

Your task is to create a Markdown fuzzer – that is, a tool that creates inputs for a Markdown processor such as Pandoc. We start with simple Markdown syntax and end with a generator that creates full-fledged papers. Use the Fandango Fuzzer for this task.

News (March 12, 2025)

• We have released a new Fandango version 0.1.7 that fixes a number of errors, notably reporting of syntax error locations. To upgrade, enter

$ pip install --upgrade fandango-fuzzer

• We found that under certain circumstances, the form <digit>+ does not expand to a sufficient number of repetitions. As a workaround, use <digit>{N,M} instead to specify a number of repetitions between N and M.

• Updates on submission and grading (see end of page).

Installing Fandango

1. Install Python. See the specific instructions for your operating system on how to do this.
2. From the command line, run

$ pip install fandango-fuzzer

Installing Pandoc

We use Pandoc as a Markdown converter. Follow the Installation Instructions.

First Steps

Create a file Markdown.fan with these contents:

<start> ::= "Some random text: " <random_text>
<random_text> ::= <printable>+

Use Fandango to fuzz inputs:

$ fandango fuzz -f Markdown.fan -n 10

Send the input to a Markdown processing tool such as Pandoc; redirect its output to a file that you can open in a browser.

$ fandango fuzz -f Markdown.fan -n 10 pandoc --to html > output.html
$ open output.html

Or create a LaTeX document from it:

$ fandango fuzz -f Markdown.fan -n 10 pandoc --to latex > output.tex
$ pdflatex output

Task 1: Basic Markdown (10%)

Extend your Fandango spec Markdown.fan for Markdown that produces inputs with the following features:

• Random letters and digits
• Emphasis (_Text_, *Text*, __Text__, **Text**)
• Headers (# Header, ## Subheader, …)
• Links ([Text](URL)). Use a fixed URL for now.

Read the Pandoc Manual to understand which Markdown features are supported. To see the list of enabled extensions, use

$ pandoc --list-extensions=markdown

Task 2: Fakers (10%)

Enhance your Fandango spec to make use of generators. Use the Python faker module to produce

• URLs
• “Lorem Ipsum” Text blocks

Task 3: Some Natural Text (10%)

Create a sub-grammar that creates random English sentences. Have rules and plenty of alternatives for subjects, verbs, objects, conjunctions, etc. Combine this with the faker module to obtain names, places, …

Task 4: Tables (10%)

Enhance your Fandango spec to produce Markdown tables. Create tables with a configurable, but fixed number of rows and columns; use <row>{N} to repeat a row N times. Fill the tables with numbers.

Task 5: Constraints (10%)

Use constraints to ensure all numbers in tables are between 10 and 1000.

Task 6: Cross-References (10%)

Create references to headers. A Header ### My Awesome Section can be referenced by taking the title, converting all letters to lowercase, and replacing all non-letters with spaces: Read [my awesome section](my-awesome-section).

For this,

1. create a generator for headers that saves the reference in a table;
2. create a second generator for references that takes a reference from the table.

Task 7: Math (10%)

Because it’s fun: Create a sub-grammar that produces LaTeX math formulas ($...$). Use pandoc --to latex to create output that you can run through LaTeX.

Be sure to create the most complex formulas ever seen.

Task 8: Citations (10%)

Have your text include citations, rendered as footnotes:

Zamudio et al. introduced Fandango.[^1]

[^1]: Zamudio, Smytzek, Zeller: The Fandango Book, 2025

• As with references, you will need generators for this task.
• Use the faker module to produce author names.
• Use constraints to produce citation years (say, 1950-2025).

Task 9: A Paper Generator (20%)

Put all the above together to create a random scientific paper in Markdown (and then LaTeX, and then PDF).

Papers will be graded on creativity as well as syntactic and semantic similarity with real papers. The best generated paper submitted by Thursday, 16:30 (end of course) will get an award.

Task 10: Large Language Models (Bonus)

For the abstract of the above paper, create a grammar that creates a random scientific-sounding title; then ask an LLM of your choice to create an abstract for it. Integrate the interaction into the Markdown.fan file.

Submitting your Work

When you’re done, submit the Markdown.fan file (as well as any generated PDFs) at https://forms.gle/vaGX976kqtKQEoo96.

Please submit

• before Wednesday, March 12, 16:30 to take part in the best paper award.
• before Monday, March 17, 23:59 to obtain a grade.

Your grade will be a percentage 0-100%; see the percentages for the individual tasks.