almostANSI

lightweight almost ANSI C Parser for javascript in jison. Parses plain C (no preprocessor, just C) source files and returns a javascript-object based representation of its AST with location information.

what's missing ('cause it's too good to be true for 3 days of work):

• error recovery
• general robustness against malicious C programming habits :-)
• uses of tagged structs, enums and unions are not checked against declarations of the same tag
• multiple uses of tagnames for structs, enums, unions is not prohibited
• enum values are neither tracked in the environment, nor is their value determined in any way
• struct bit-fields is not supported ( struct foo { int bar:5, baz:13; } )
• type qualifiers (volatile / const)
• legacy K&R function declarations ( int f(a,b) int a; double b; {} )

see a demonstration of the code in action:

We have a small HTML-demonstrator running here

install dependencies & generate parser:

To generate the parser, execute the following piece once:

npm install jison 
node node_modules/jison/lib/cli.js  ansic.jison ansic.jisonlex 

The resulting ansic.js is now a standalone javascript file, that can be used without any dependencies.

test parser CLI (will usually only output errors on faulty input):

node ansic.js test.c

test parser from within javascript code via commonjs:

// test.js
var parser = require("./ansic").parser

console.log(
    JSON.stringify(
        parser.parse("int main() { x = 25+x; } "),
        null, 
        2
    )
)

test parser from within javascript code via import:

// test.mjs
import * as parser from "./ansic.js"

console.log(
    JSON.stringify(
        parser.parse("int main() { x = 25+x; } "),
        null, 
        2
    )
)

test parser from within typescript code via import:

// tsconfig.json : add  "allowJs": true,  

// test.ts
import * as almostANSI from './ansic.js';

console.log(almostANSI.parse("int main() { return 5; }"));

Program Analysis:

We have added a frontend to perform abstract interpretation based static program analysis via the simple Python script analyzeANSI.py - in It comes with a predefined output to graphviz in dot format, as well as with a few example constraint systems that are generated when the program is started with particular parameters.

A start point into setting up custom constraints for a static analysis via abstract interpretation can be to provide a new edge collector for the iterateEdges and write out constraints for a fixpoint solver.

./analyzeANSI.py [input.c] | xdot /dev/stdin