trick/trick_source/codegen/Interface_Code_Gen/CommentSaver.cpp

#include <iostream>
#include <sstream>
#include <regex.h>
#include <algorithm>

#include "clang/Basic/FileManager.h"

#include "CommentSaver.hh"
#include "Utilities.hh"

CommentSaver::CommentSaver(clang::CompilerInstance & in_ci , HeaderSearchDirs & in_hsd ) : ci(in_ci) , hsd(in_hsd) {}

bool CommentSaver::HandleComment(clang::Preprocessor &PP, clang::SourceRange Comment) {
    //Comment.getBegin().dump(sm) ;

    //if ( ! sm.isInSystemHeader(Comment.getBegin()) ) {
    if ( isInUserOrTrickCode( ci , Comment.getBegin() , hsd ) ) {
        std::string file_name = ci.getSourceManager().getBufferName(Comment.getBegin()) ;
        unsigned int line_no = ci.getSourceManager().getSpellingLineNumber(Comment.getBegin()) ;
        comment_map[file_name][line_no] = Comment ;
    }

    // returning false means we did not push any text back to the stream for further reads.
    return false ;
}

std::string CommentSaver::getComment( clang::SourceRange sr ) {

    if ( sr.isValid() ) {
        /* fsl_begin and fsl_end are two pointers into the header file.  We want the text between the two pointers. */
        clang::FullSourceLoc fsl_begin(sr.getBegin() , ci.getSourceManager()) ;
        clang::FullSourceLoc fsl_end(sr.getEnd() , ci.getSourceManager()) ;
        std::string comment_text( fsl_begin.getCharacterData() ,
         (size_t)(fsl_end.getCharacterData() - fsl_begin.getCharacterData())) ;
        return comment_text ;
    }

    /* return an empty string if no comment found on this line of the file. */
    return std::string() ;
}

std::string CommentSaver::getComment( std::string file_name , unsigned int line_no ) {
    return getComment(comment_map[file_name][line_no]) ;
}

std::string CommentSaver::getTrickHeaderComment( std::string file_name ) {

    std::map < unsigned int , clang::SourceRange >::iterator cit ;

    if ( trick_header_comments.find(file_name) == trick_header_comments.end() ) {
        trick_header_comments[file_name] = std::string() ;
        for ( cit = comment_map[file_name].begin() ; cit != comment_map[file_name].end() ; cit++ ) {
            std::string comment_str = getComment((*cit).second) ;
            if ( comment_str.find("@trick_parse") != std::string::npos or
                 comment_str.find("\\trick_parse") != std::string::npos ) {
                trick_header_comments[file_name] = getComment((*cit).second) ;
                break ;
            } else {
                std::transform(comment_str.begin(), comment_str.end(), comment_str.begin(), ::toupper) ;
                if ( comment_str.find("PURPOSE") != std::string::npos ) {
                    trick_header_comments[file_name] = getComment((*cit).second) ;
                    break ;
                }
            }
        }
    }

    return trick_header_comments[file_name] ;
}

void CommentSaver::getICGField( std::string file_name ) {

    /* default the icg_no flags to false */
    icg_no_found[file_name] = false ;
    icg_no_comment_found[file_name] = false ;

    std::string th_str = getTrickHeaderComment(file_name) ;
    if ( ! th_str.empty() ) {

        size_t trick_parse_keyword ;
        if ( (trick_parse_keyword = th_str.find("trick_parse")) != std::string::npos ) {
            size_t trick_parse_everything = th_str.find("everything", trick_parse_keyword) ;
            size_t trick_parse_attributes = th_str.find("attributes", trick_parse_keyword) ;
            size_t trick_parse_dep_only = th_str.find("dependencies_only", trick_parse_keyword) ;
            size_t closing_brace = th_str.find("}", trick_parse_keyword) ;
            if ( closing_brace != std::string::npos ) {
                if ( trick_parse_everything != std::string::npos and
                     trick_parse_everything < closing_brace ) {
                    icg_no_comment_found[file_name] = false ;
                    icg_no_found[file_name] = false ;
                } else if ( trick_parse_attributes != std::string::npos and
                     trick_parse_attributes < closing_brace ) {
                    icg_no_comment_found[file_name] = true ;
                    icg_no_found[file_name] = false ;
                } else if ( trick_parse_dep_only != std::string::npos and
                     trick_parse_dep_only < closing_brace ) {
                    icg_no_comment_found[file_name] = false ;
                    icg_no_found[file_name] = true ;
                } else {
                    std::cout << file_name << ": trick_parse directive found but no argument matched." << std::endl ;
                    std::cout << "Valid arguments are (everything|attributes|dependencies_only)" << std::endl ;
                    exit(-1) ;
                }
            }
        } else {

            std::transform(th_str.begin(), th_str.end(), th_str.begin(), ::toupper) ;
            int ret ;
            regex_t reg_expr ;
            regmatch_t pmatch[10] ;
            memset(pmatch , 0 , sizeof(pmatch)) ;

            /* POSIX regular expressions are always greedy, making our job harder.
               We have to use several regular expressions to get the types.  This was
               so much easier in perl! */

            /* find the ICG field */
            ret = regcomp( &reg_expr , "(ICG:)" , REG_EXTENDED | REG_ICASE ) ;
            ret = regexec( &reg_expr , th_str.c_str() , 10 , pmatch , 0 ) ;
            regfree(&reg_expr) ;
            if ( ret != 0 ) {
                return ;
            }
            th_str = th_str.substr(pmatch[1].rm_eo) ;

            /* find the end of the ICG field */
            memset(pmatch , 0 , sizeof(pmatch)) ;
            ret = regcomp( &reg_expr , "(\\))" , REG_EXTENDED ) ;
            ret = regexec( &reg_expr , th_str.c_str() , 10 , pmatch , 0 ) ;
            regfree(&reg_expr) ;

            if ( ret != 0 ) {
                return ;
            }
            th_str = th_str.substr(0 , pmatch[1].rm_so) ;

            /* test for NoComment */
            memset(pmatch , 0 , sizeof(pmatch)) ;
            ret = regcomp( &reg_expr , "(NOCOMMENT)$" , REG_EXTENDED ) ;
            ret = regexec( &reg_expr , th_str.c_str() , 10 , pmatch , 0 ) ;
            regfree(&reg_expr) ;

            if ( ret == 0 ) {
                icg_no_comment_found[file_name] = true ;
            }

            /* test for No */
            memset(pmatch , 0 , sizeof(pmatch)) ;
            ret = regcomp( &reg_expr , "(NO)$" , REG_EXTENDED ) ;
            ret = regexec( &reg_expr , th_str.c_str() , 10 , pmatch , 0 ) ;
            regfree(&reg_expr) ;

            if ( ret == 0 ) {
                icg_no_found[file_name] = true ;
                icg_no_comment_found[file_name] = true ;
            }
        }
    }

}

bool CommentSaver::hasTrickHeader( std::string file_name ) {
    std::string th_str = getTrickHeaderComment(file_name) ;
    return (! th_str.empty()) ;
}

bool CommentSaver::hasICGNo( std::string file_name ) {

    if ( icg_no_found.find(file_name) == icg_no_found.end() ) {
        getICGField(file_name) ;
    }

    return icg_no_found[file_name] ;
}

bool CommentSaver::hasICGNoComment( std::string file_name ) {

    if ( icg_no_comment_found.find(file_name) == icg_no_comment_found.end() ) {
        getICGField(file_name) ;
    }

    return icg_no_comment_found[file_name] ;
}

std::set< std::string > CommentSaver::getIgnoreTypes( std::string file_name ) {
    std::set< std::string > ignore_types ;
    std::string th_str = getTrickHeaderComment(file_name) ;
    if ( ! th_str.empty() ) {
        int ret ;
        regex_t reg_expr ;
        regmatch_t pmatch[10] ;
        //std::cout << "here in getIgnoreTypes\n" << th_str << std::endl ;

        size_t start = 0 ;
        start = th_str.find( "trick_exclude_typename") ;
        while ( start != std::string::npos ) {
            start += strlen("trick_exclude_typename") ;
            std::string temp_str = th_str.substr(start) ;
            memset(pmatch , 0 , sizeof(pmatch)) ;
            ret = regcomp( &reg_expr , "^\\s*\\{\\s*(\\S+)\\s*\\}" , REG_EXTENDED ) ;
            ret = regexec( &reg_expr , temp_str.c_str() , 10 , pmatch , 0 ) ;
            regfree(&reg_expr) ;
            if ( ret == 0 ) {
                std::string item = temp_str.substr(pmatch[1].rm_so, pmatch[1].rm_eo) ;
                // regular expression leaving trailing space and brace. Why?
                item.erase(item.find_first_of(" \t}")) ;
                //std::cout << "icg_ignore_types " << item << std::endl ;
                ignore_types.insert(item) ;
            }
            start = th_str.find( "trick_exclude_typename", start ) ;
        }

        std::transform(th_str.begin(), th_str.end(), th_str.begin(), ::toupper) ;

        memset(pmatch , 0 , sizeof(pmatch)) ;

        /* POSIX regular expressions are always greedy, making our job harder.
           We have to use several regular expressions to get the types.  This was
           so much easier in perl! */

        /* find the start of the ICG_IGNORE_TYPES field */
        ret = regcomp( &reg_expr , "(ICG[ _]IGNORE[ _]TYPE(S)?:)" , REG_EXTENDED | REG_ICASE ) ;
        ret = regexec( &reg_expr , th_str.c_str() , 10 , pmatch , 0 ) ;
        regfree(&reg_expr) ;
        if ( ret != 0 ) {
            return std::set< std::string >() ;
        }
        th_str = th_str.substr(pmatch[1].rm_eo) ;

        /* find the end of the ICG_IGNORE_TYPES field */
        memset(pmatch , 0 , sizeof(pmatch)) ;
        ret = regcomp( &reg_expr , "(\\)\\s*\\))" , REG_EXTENDED ) ;
        ret = regexec( &reg_expr , th_str.c_str() , 10 , pmatch , 0 ) ;
        regfree(&reg_expr) ;
        if ( ret != 0 ) {
            return std::set< std::string >() ;
        }
        th_str = th_str.substr(0 , pmatch[1].rm_so) ;
        std::replace( th_str.begin(), th_str.end(), '(', ' ');
        std::replace( th_str.begin(), th_str.end(), ')', ' ');
        std::replace( th_str.begin(), th_str.end(), '\n', ' ');

        std::stringstream ss(th_str);
        std::string item;
        while (std::getline(ss, item, ' ')) {
            if ( ! item.empty() ) {
                //std::cout << "icg_ignore_types " << item << std::endl ;
                ignore_types.insert(item) ;
            }
        }

    }

    return ignore_types ;
}