C Programming Language parser (ISO/IEC 9899:2018)

Basic C parser (ISO/IEC 9899:2018 standard review) on Flex/Yacc converting programs to JSON.
Authors: Denis Chernikov, Vladislav Kuleykin

Purposes

This project was done during the Compiler Construction course (primary instructor: Eugene Zouev) at Fall 2018 semester at Innopolis University (Innopolis, Russia).
This one does very basic things:

Parsing some basic programs on C Programming Language;
Telling if a given program is syntactically correct or not;
If the program is correct, printing generated AST to the specified file in JSON format.

How to run the parser

Download and unzip the content of a repository.
On Windows: start PARSE.BAT file.
On Unix (not tested): start parse.sh file.

Note: On Windows, we assume that you have a gcc compiler accessible via command gcc source.c -o source.exe, also all the required tools can be downloaded by this link (put flex and bison folders near the batch file). On Unix, we assume that you have gcc compiler, and also that you have already installed tools flex and bison. If you don't, you can install them simply using commands:

sudo apt install flex
sudo apt install bison

How to run tests

Download and unzip the content of a repository.
On Windows: start TESTS.BAT file.
On Unix (not tested): start tests.sh file.

Project assumptions

Our lexical analyzer is done using Flex tool (not just Lex because of using start conditions %x). But we do not have a corresponding preprocessor. That's why lexical analyzer is a bit complicated (digraphs, trigraphs and newline escapes are considered and #include directives are handeled, but #define, #undef, #if, #else and other conditional directives are just ommited).
Lexical analyzer is required to convert all literals to the correct internal representation (like correct sequence of bits). In our case, only string literals and character constants are converted (de-escaped).
Error recovery in syntax is not complete (more detailed analyzis is required to put nonterminal error without conflicts). Parsing stops at the first found syntax error.
If syntax (or lexical) error occured - nothing will be printed to the specified output file. Actually, this file will not be opened for writing at all.
In case of error, no concrete description is printed yet. Easier to see errors in "debug mode" (add flag -t to bison command, -d to flex command, and in main function set yydebug = 1;).
Syntax analyzis is done for the whole set of possible programs that are written according to the ISO/IEC 9899:2018. As a reference, 2017's draft was used, but since that version no critical changes were applied. Just some obvious mistakes in standard were fixed (like - used as decrement).
There is no official code style convention for the C Programming Language. We wanted to follow a convention about maximum 80 characters per line and code style from the examples from the standard. Actually, some cases required to use much more than 80 characters in a single line (separation looked worse sometimes), that's why our code style convention is about 130 characters per line.

Additional notes

Despite the fact that we use bison tool (which is for C++), actually the project is implemented in terms of classical Yacc (flag -y). Therefore our project is written fully on C Programming Language (note some bootstrapping possibilities). Actually, do NOT try to bootstrap the project on this stage (regards to weak preprocessing).
Actual executable's usage is with one or two arguments to it's call: c_parser.exe %input_file_name% %output_file_name% OR with just one argument, assuming the input from the command line: c_parser.exe %output_file_name%
To see a PDA description of Yacc's automata add -v flag to the bison call. See generated file y.output.
In ISO/IEC 9899:2018 there is no preprocessing directive called #warning, but a lot of resources do note it, that's why we enabled this small feature in our lexical analyzer.

BNF description

Actually, the whole syntax description without code is placed in pure_c_grammar.y. It is too huge to convert it into the original BNF notation. Lexical units are explained good enough almost... everywhere! We will ommit it...

deiuch / c-to-json