Commit 0390fbf9 authored by Dick Hollenbeck's avatar Dick Hollenbeck
Browse files

speedups

parent f280c737

common/dsnlexer.cpp

+40 −56
Original line number Diff line number Diff line
@@ -374,11 +374,11 @@ int DSNLEXER::NeedNUMBER( const char* aExpectation ) throw( IO_ERROR ) 
 * i.e. no bytes with MSB on can be considered whitespace, since they are likely part

 * of a multibyte UTF8 character.

 */

static bool isSpace( int cc )

static bool isSpace( char cc )

{

    // cc was signed extended from signed char, so it is often negative.

    // cc is signed, so it is often negative.

    // Treat negative as large positive to exclude rapidly.

    if( unsigned( cc ) <= ' ' )

    if( (unsigned char) cc <= ' ' )

    {

        switch( cc )

        {
@@ -394,47 +394,41 @@ static bool isSpace( int cc ) 
}





/**

 * Function isNumber

 * return true if the next sequence of s-expression text is a number:

 * either an integer, fixed point, or float with exponent.

 *

 * @param cp is the start of the current token.

 * @param limit is the end of the current line of text.

 * @param next is where to put a pointer to the start of the token after this one, but is only

 *    updated if the return value is true.

 * @return bool - true if this is a number, else false.

 */

static bool isNumber( const char* cp, const char* limit, const char** next )

inline bool isDigit( char cc )

{

#if 0   // no exponents supported

    return '0' <= cc && cc <= '9';

}



    if( strchr( "+-.0123456789", *cp ) )

    {

        ++cp;

        while( cp<limit && strchr( ".0123456789", *cp )  )

            ++cp;



        if( cp==limit || isSpace( *cp ) || *cp==')' || *cp=='(' )

/// return true if @a cc is an s-expression separator character

inline bool isSep( char cc )

{

            *next = cp;

            return true;

        }

    return isSpace( cc ) || cc=='(' || cc==')';

}



    return false;



#else   // support numbers with exponents.



    // regex for a float: "^[-+]?[0-9]*\\.?[0-9]+([eE][-+]?[0-9]+)?" i.e. any number

    // is traversed, coded manually here:

/**

 * Function isNumber

 * returns true if the next sequence of text is a number:

 * either an integer, fixed point, or float with exponent.  Stops scanning

 * at the first non-number character, even if it is not whitespace.

 *

 * @param cp is the start of the current token.

 * @param limit is the end of the current line of text.

 * @return const char* - pointer to the start of the token after this one, but if

 *    this token is a number, else NULL

 */

static const char* isNumber( const char* cp, const char* limit )

{

    // regex for a float: "^[-+]?[0-9]*\\.?[0-9]+([eE][-+]?[0-9]+)?" i.e. any number,

    // code traversal manually here:



    bool sawNumber = false;



    if( cp < limit && strchr( "-+", *cp ) )

    if( cp < limit && ( *cp=='-' || *cp=='+' ) )

        ++cp;



    while( cp < limit && strchr( "0123456789", *cp ) )

    while( cp < limit && isDigit( *cp ) )

    {

        ++cp;

        sawNumber = true;
@@ -444,7 +438,7 @@ static bool isNumber( const char* cp, const char* limit, const char** next ) 
    {

        ++cp;



        while( cp < limit && strchr( "0123456789", *cp ) )

        while( cp < limit && isDigit( *cp ) )

        {

            ++cp;

            sawNumber = true;
@@ -453,16 +447,16 @@ static bool isNumber( const char* cp, const char* limit, const char** next ) 


    if( sawNumber )

    {

        if( cp < limit && strchr( "Ee", *cp ) )

        if( cp < limit && ( *cp=='E' || *cp=='e' ) )

        {

            ++cp;



            sawNumber = false;  // exponent mandates at least one digit thereafter.



            if( cp < limit && strchr( "-+", *cp ) )

            if( cp < limit && ( *cp=='-' || *cp=='+' )  )

                ++cp;



            while( cp < limit && strchr( "0123456789", *cp ) )

            while( cp < limit && isDigit( *cp ) )

            {

                ++cp;

                sawNumber = true;
@@ -470,18 +464,7 @@ static bool isNumber( const char* cp, const char* limit, const char** next ) 
        }

    }



    if( sawNumber )

    {

        // token can only be a number if not adjoined with other non-number token text

        if( cp==limit || isSpace( *cp ) || *cp==')' || *cp=='(' )

        {

            *next = cp;

            return true;

        }

    }



    return false;

#endif

    return sawNumber ? cp : NULL;

}
@@ -577,7 +560,7 @@ L_read: 


            head = cur+1;



            if( head<limit && *head!=')' && *head!='(' && !isSpace(*head) )

            if( head<limit && !isSep( *head ) )

            {

                THROW_PARSE_ERROR( errtxt, CurSource(), CurLine(), CurLineNumber(), CurOffset() );

            }
@@ -598,9 +581,10 @@ L_read: 
            goto exit;

        }



        if( isNumber( cur, limit, &head ) )

        if( ( head = isNumber( cur, limit ) ) != NULL &&

            // token can only be a number if trailing text is a separator or line end

            ( head==limit || isSep( *head ) ) )

        {

            // handle DSN_NUMBER

            curText.clear();

            curText.append( cur, head );

            curTok = DSN_NUMBER;
@@ -726,7 +710,7 @@ L_read: 
        // If not, then call it a DSN_SYMBOL.

        {

            head = cur+1;

            while( head<limit && !isSpace( *head ) && *head!=')' && *head!='(' )

            while( head<limit && !isSep( *head ) )

                ++head;



            curText.clear();