Objectives

• to compile C programs using the Unix command line
• to use C I/O functions
• to use finite state machines to model a process

Introduction

Profs. Dawn Lawrie and Dave Binkley are working on a project with the aim of splitting and expanding identifers to aid program comprehension and search. For example, if a programmer wants to find the parts of a program where the radius of a sphere is computed or used, they might not know that the code code uses the identifier sphrad for that value and so would not know what to search for. Profs. Lawrie and Binkley use nearby comments in order to determine how to split and expand identifiers, so that if sphrad appears near the comment "compute the radius of the sphere" in one place, sphrad would match the search "sphere radius" everywhere.

You will develop a tool to extract comments from C source code as the first part of this analysis (we will leave extraction of identifiers and all subsequent steps to someone else).

Assignment

Write a program called Comments that reads C source code from standard input and writes the text of the comments to standard output. The text of a comment does not include the delimiters /* and */ at the beginning or end of a C-style comment, or the marker // that starts a C++-style comment. In both cases, leading whitespace and asterisks (*) on each line should not be output; internal and trailing whitespace should be output as is and each non-empty comment should be output with a newline character at the end ("non-empty" means contains something other than whitespace, asterisks, and tags). Furthermore, doxygen/javadoc style tags – any word within a comment that starts with an at symbol (@) – should not be output regardless of what type of comment it is in and whether it is actually a valid tag (so don't output something like @DrGlennNFA even though that is not a tag used by either tool). Additionally, note that

• tags should be treated as whitespace for the purposes of determining what is leading whitespace,
• line breaks should be output where found in comments, except
  • a backslash (\) at the end of a line immediately before a newline character is a line-continuation character and should be ignored along with the following newline; for all purposes consider those two characters to not appear in the input
  • there should be no empty lines in the output,
• the beginning of a comment should be treated as the beginning of a line regardless of whether the line contains code before the comment, and
• all other punctation must be output as it appears in the input (we leave the task of filtering out such things to others).

Other things to consider:

• within string and character literals, /* and // do not mark the beginning of a comment
• quotes within comments need not be balanced
• you may assume that string and character literals are properly terminated
• you may assume that the character after an unescaped backslash within a string or character literal creates a valid escape sequence
• line-continuation (see above) is valid within string and character literals as well as in comments
• you need not do anything special for preprocessor directives; treat them like any other code
• you may assume there are no trigraphs in the input

Your program must use only a constant amount of space (in other words, the amount of memory used must not vary with the amount of input read). So, for example, you may not use arrays or call malloc.

Your submission must include a makefile that produces an executable file named Comments when make is run with no arguments.

Example

/* This is a single-line C comment */
#include <stdio.h>
/******
 * This is a nicely formatted
 * multi-line comment.
 ******/

int main(int argc, char **argv)
{
  // This is a C++ comment.
}

$ ./Comments < input_0.txt
This is a single-line C comment
This is a nicely formatted
multi-line comment.
This is a C++ comment.