codanim: Animated code & data structures with LaTeX/beamer

(C) 2020 Franck Pommereau <[email protected]>
    This is free software, see file LICENCE

codanim allows to generate LaTeX/TikZ code to animate source code and data structures for beamer presentations. So far, it is oriented to simulate C code, but there is no conceptual limitation for the simulation of other languages.

codanim is currently highly experimental and neither well documented nor well tested. You've been warned.

INSTALL

There is so far no installation procedure. Just copy directory codanim somewhere in your PYTHONPATH.

File examples/pygments.sty will be needed to compile the final LaTeX files. File examples/tpl.tex shows how the generated LaTeX code may be included in a beamer presentation.

CONCEPTS

codanim defines Python classes to simulate the data and control-flow structures of an imperative language. Data structures are defined in codanim.data and include:

• Pointer(data) a pointer to another data data
• Value(value) an arbitrary value, initialised to value. If value is None (the default), the value is considered uninitialised
• Array(init, index=[]) an array of values, that is zero-indexed, and initialised as follows:
  • if init is an int, then it is the length of the array whose values are all uninitialised
  • if init is a list then is holds the initial values of the array index is a list of variables identifiers that may be used to access the array cells (see below)
• Struct(init) a structure whose fields and initial values are defined by init that must be a dict
• Heap() a container for dynamically allocated data that provides methods new and free to do so

Control-flow structures are used to simulate within Python code that may potentially be written in any imperative language. Doing so, all the changes that are made to the data structures defined above are recorded so that they may be latter animated, consistently with the animation of the code itself. The idea is that source code in the simulated language is split into corresponding control-flow structures, and actual computation is done using equivalent Python code instead of executing the code in the source language. Control-flow structures are defined in codanim.flow and include:

• RAW(src='...') raw code from the simulated language, its simulation is no-op, but it is rendered highlighted in the final animation
• WS(src='...') just like RAW but should be only white spaces (so that it won't be highlighted)
• BLOCK(*body) a group of other structures that are simulated (and rendered) sequentially
• STMT(*steps, src='...') an arbitrary statement that is simulated by running itssteps sequentially, each of which being an arbitrary Python statement that is execed
• EXPR(expr, src='...') an arbitrary Python expression expr that simulate an expression in the simulated language
• PY(code) an arbitrary Python statement that need to be executed but is not rendered in the final animation, just like every structure that expects no src='...' argument
• ENV(name, value) a data structure that need to be defined in order to do the simulation (typically: global or external variables), and which is stored into variable name that can be used from the Python code of statements and expression
• DECL(name, init=None, animate=False, src'...') an actual variable declaration in the simulated language, that will be executed and rendered. The execution consists of evaluating init that should be an EXPR (or None) and assigning its value to name. If animate is True then the value of the variable will displayed within a comment just next to the declaration in the final animation
• XDECL(*names, src='...') several uninitialised declarations
• BREAK a break instruction for loops and switches
• RETURN(value, src='...') a return instruction from a function
• IF(cond, then, otherwise=None, src='...') simulates an if block with an optional else part (called otherwise since else is a Python keyword. Argument cond should be an EXPR instance
• WHILE(cond, body, src='...') simulates a while loop
• DO(body, cond, src='...') simulates a do/while loop
• FOR(init, cont, step, body) simulates a C-like for loop
• FUNC(body, src='...') simulates a function TODO: functions calls is not implemented yet, so basically, a function is so far only a BLOCK with source code
• SWITCH(cond, *cases, src='...') simulates a C-like switch, cases may be:
  • CASE(value, body) to simulate a case statement
  • DEFAULT() to simulate a default statement

So, all what you need is to defined some data structures, some control-flow structures with appropriate src arguments (so that code is rendered in the simulated language), and with appropriate Python code to simulate the original statements and expressions.

Writing the control-flow structures may be tedious, so codanim may be called from the command line to parse actual code to be simulated and generate the appropriate control-flow structures (in which the Python code remains to be written). Run python -m codanim for help.

EXAMPLES

File examples/Makefile can be used to build the PDF for all the examples. For instance use make heap.pdf to build the first example described below.

examples/heap.py

It defines a Heap instance and add chained Structs to it. The final picture is rendered as TikZ code. There is no animation, this is just the picture of a chained list.

examples/stack-push.py

It uses the same kind of linked lists as above, seen as stacks, to animate a push onto a stack. First the data is defined, then the control-flow structure. The latter is then simulated and finally, both code and data structures are rendered for LaTeX/beamer inclusion.

Note the use of class Box that can be used to layout several data structures.

examples/qs-partition.py and examples/qs-main.py

The partitioning and main algorithm of a quick-sort. This shows how to use Array, in particular the index argument. qs-main also shows how to define custom layout of arrays, as well as auxiliary Python functions to simulate full C-functions calls in one step.