[user/henk/code/inspircd.git] / src / hashcomp.cpp

/*       +------------------------------------+
 *       | Inspire Internet Relay Chat Daemon |
 *       +------------------------------------+
 *
 *  InspIRCd is copyright (C) 2002-2006 ChatSpike-Dev.
 *                     E-mail:
 *              <brain@chatspike.net>
 *              <Craig@chatspike.net>
 *
 * Written by Craig Edwards, Craig McLure, and others.
 * This program is free but copyrighted software; see
 *          the file COPYING for details.
 *
 * ---------------------------------------------------
 */

using namespace std;

#include "inspircd.h"
#include "hashcomp.h"
#include <ext/hash_map>
#define nspace __gnu_cxx

/******************************************************
 *
 * The hash functions of InspIRCd are the centrepoint
 * of the entire system. If these functions are
 * inefficient or wasteful, the whole program suffers
 * as a result. A lot of C programmers in the ircd
 * scene spend a lot of time debating (arguing) about
 * the best way to write hash functions to hash irc
 * nicknames, channels etc.
 * We are lucky as C++ developers as hash_map does
 * a lot of this for us. It does intellegent memory
 * requests, bucketing, search functions, insertion
 * and deletion etc. All we have to do is write some
 * overloaded comparison and hash value operators which
 * cause it to act in an irc-like way. The features we
 * add to the standard hash_map are:
 *
 * Case insensitivity: The hash_map will be case
 * insensitive.
 *
 * Scandanavian Comparisons: The characters [, ], \ will
 * be considered the lowercase of {, } and |.
 *
 * This file also contains hashing methods for hashing
 * insp_inaddr structs, we use this if we want to cache IP
 * addresses.
 *
 ******************************************************/

using namespace std;
using namespace irc::sockets;

/* convert a string to lowercase. Note following special circumstances
 * taken from RFC 1459. Many "official" server branches still hold to this
 * rule so i will too;
 *
 *  Because of IRC's scandanavian origin, the characters {}| are
 *  considered to be the lower case equivalents of the characters []\,
 *  respectively. This is a critical issue when determining the
 *  equivalence of two nicknames.
 */
void nspace::strlower(char *n)
{
        if (n)
        {
                for (char* t = n; *t; t++)
                        *t = lowermap[(unsigned char)*t];
        }
}

size_t nspace::hash<insp_inaddr>::operator()(const insp_inaddr &a) const
{
        size_t q;
        memcpy(&q,&a,sizeof(size_t));
        return q;
}

size_t nspace::hash<string>::operator()(const string &s) const
{
        /* XXX: NO DATA COPIES! :)
         * The hash function here is practically
         * a copy of the one in STL's hash_fun.h,
         * only with *x replaced with lowermap[*x].
         * This avoids a copy to use hash<const char*>
         */
        register size_t t = 0;
        for (std::string::const_iterator x = s.begin(); x != s.end(); ++x) /* ++x not x++, as its faster */
                t = 5 * t + lowermap[(unsigned char)*x];
        return t;
}

bool irc::StrHashComp::operator()(const std::string& s1, const std::string& s2) const
{
        unsigned char* n1 = (unsigned char*)s1.c_str();
        unsigned char* n2 = (unsigned char*)s2.c_str();
        for (; *n1 && *n2; n1++, n2++)
                if (lowermap[*n1] != lowermap[*n2])
                        return false;
        return (lowermap[*n1] == lowermap[*n2]);
}

bool irc::InAddr_HashComp::operator()(const insp_inaddr &s1, const insp_inaddr &s2) const
{
#ifdef IPV6
        for (int n = 0; n < 16; n++)
                if (s2.s6_addr[n] != s1.s6_addr[n])
                        return false;
        return true;
#else
        return (s1.s_addr == s1.s_addr);
#endif
}

/******************************************************
 *
 * This is the implementation of our special irc::string
 * class which is a case-insensitive equivalent to
 * std::string which is not only case-insensitive but
 * can also do scandanavian comparisons, e.g. { = [, etc.
 *
 * This class depends on the const array 'lowermap'.
 *
 ******************************************************/

bool irc::irc_char_traits::eq(char c1st, char c2nd)
{
        return lowermap[(unsigned char)c1st] == lowermap[(unsigned char)c2nd];
}

bool irc::irc_char_traits::ne(char c1st, char c2nd)
{
        return lowermap[(unsigned char)c1st] != lowermap[(unsigned char)c2nd];
}

bool irc::irc_char_traits::lt(char c1st, char c2nd)
{
        return lowermap[(unsigned char)c1st] < lowermap[(unsigned char)c2nd];
}

int irc::irc_char_traits::compare(const char* str1, const char* str2, size_t n)
{
        for(unsigned int i = 0; i < n; i++)
        {
                if(lowermap[(unsigned char)*str1] > lowermap[(unsigned char)*str2])
                        return 1;

                if(lowermap[(unsigned char)*str1] < lowermap[(unsigned char)*str2])
                        return -1;

                if(*str1 == 0 || *str2 == 0)
                        return 0;

                str1++;
                str2++;
        }
        return 0;
}

std::string operator+ (std::string& leftval, irc::string& rightval)
{
        return leftval + std::string(rightval.c_str());
}

irc::string operator+ (irc::string& leftval, std::string& rightval)
{
        return leftval + irc::string(rightval.c_str());
}

bool operator== (std::string& leftval, irc::string& rightval)
{
        return (leftval == std::string(rightval.c_str()));
}

bool operator== (irc::string& leftval, std::string& rightval)
{
        return (rightval == std::string(leftval.c_str()));
}

const char* irc::irc_char_traits::find(const char* s1, int  n, char c)
{
        while(n-- > 0 && lowermap[(unsigned char)*s1] != lowermap[(unsigned char)c])
                s1++;
        return s1;
}

/* See hashcomp.h if you care about these... */
std::ostream& operator<<(std::ostream &os, const irc::string &str)
{
        return os << str.c_str();
}

std::istream& operator>>(std::istream &is, irc::string &str)
{
        std::string tmp;
        is >> tmp;
        str = tmp.c_str();
        return is;
}

irc::tokenstream::tokenstream(const std::string &source) : tokens(source), last_pushed(false)
{
        /* Record starting position and current position */
        last_starting_position = tokens.begin();
        n = tokens.begin();
}

irc::tokenstream::~tokenstream()
{
}

const std::string irc::tokenstream::GetToken()
{
        std::string::iterator lsp = last_starting_position;

        while (n != tokens.end())
        {
                /** Skip multi space, converting "  " into " "
                 */
                while ((n+1 != tokens.end()) && (*n == ' ') && (*(n+1) == ' '))
                        n++;

                if ((last_pushed) && (*n == ':'))
                {
                        /* If we find a token thats not the first and starts with :,
                         * this is the last token on the line
                         */
                        std::string::iterator curr = ++n;
                        n = tokens.end();
                        return std::string(curr, tokens.end());
                }

                last_pushed = false;

                if ((*n == ' ') || (n+1 == tokens.end()))
                {
                        /* If we find a space, or end of string, this is the end of a token.
                         */
                        last_starting_position = n+1;
                        last_pushed = true;

                        std::string strip(lsp, n+1 == tokens.end() ? n+1  : n++);
                        while ((strip.length()) && (strip.find_last_of(' ') == strip.length() - 1))
                                strip.erase(strip.end() - 1);

                        return strip;
                }

                n++;
        }
        return "";
}

irc::sepstream::sepstream(const std::string &source, char seperator) : tokens(source), sep(seperator)
{
        last_starting_position = tokens.begin();
        n = tokens.begin();
}

const std::string irc::sepstream::GetToken()
{
        std::string::iterator lsp = last_starting_position;

        while (n != tokens.end())
        {
                if ((*n == sep) || (n+1 == tokens.end()))
                {
                        last_starting_position = n+1;
                        std::string strip = std::string(lsp, n+1 == tokens.end() ? n+1  : n++);

                        while ((strip.length()) && (strip.find_last_of(sep) == strip.length() - 1))
                                strip.erase(strip.end() - 1);

                        return strip;
                }

                n++;
        }

        return "";
}

irc::sepstream::~sepstream()
{
}

std::string irc::hex(const unsigned char *raw, size_t rawsz)
{
        if (!rawsz)
                return "";

        /* EWW! This used to be using sprintf, which is WAY inefficient. -Special */
        
        const char *hex = "0123456789abcdef";
        
        char buf[rawsz*2+1];

        size_t i, j;
        for (i = 0, j = 0; j < rawsz; ++j)
        {
                buf[i++] = hex[raw[j] / 16];
                buf[i++] = hex[raw[j] % 16];
        }
        buf[i] = '\0';

        return buf;
}

const char* irc::Spacify(char* n)
{
        static char x[MAXBUF];
        strlcpy(x,n,MAXBUF);
        for (char* y = x; *y; y++)
                if (*y == '_')
                        *y = ' ';
        return x;
}


irc::modestacker::modestacker(bool add) : adding(add)
{
        sequence.clear();
        sequence.push_back("");
}

void irc::modestacker::Push(char modeletter, const std::string &parameter)
{
        *(sequence.begin()) += modeletter;
        sequence.push_back(parameter);
}

void irc::modestacker::Push(char modeletter)
{
        this->Push(modeletter,"");
}

void irc::modestacker::PushPlus()
{
        this->Push('+',"");
}

void irc::modestacker::PushMinus()
{
        this->Push('-',"");
}

int irc::modestacker::GetStackedLine(std::deque<std::string> &result)
{
        int n = 0;
        result.clear();
        result.push_back(adding ? "+" : "-");

        while (!sequence[0].empty() && (sequence.size() > 1) && (result.size() < MAXMODES+1))
        {
                result[0] += *(sequence[0].begin());
                if (!sequence[1].empty())
                        result.push_back(sequence[1]);
                sequence[0].erase(sequence[0].begin());
                sequence.erase(sequence.begin() + 1);
                n++;
        }

        return n;
}

irc::stringjoiner::stringjoiner(const std::string &seperator, const std::vector<std::string> &sequence, int begin, int end)
{
        for (int v = begin; v < end; v++)
                joined.append(sequence[v]).append(seperator);
        joined.append(sequence[end]);
}

irc::stringjoiner::stringjoiner(const std::string &seperator, const std::deque<std::string> &sequence, int begin, int end)
{
        for (int v = begin; v < end; v++)
                joined.append(sequence[v]).append(seperator);
        joined.append(sequence[end]);
}

irc::stringjoiner::stringjoiner(const std::string &seperator, const char** sequence, int begin, int end)
{
        for (int v = begin; v < end; v++)
                joined.append(sequence[v]).append(seperator);
        joined.append(sequence[end]);
}

std::string& irc::stringjoiner::GetJoined()
{
        return joined;
}

irc::portparser::portparser(const std::string &source, bool allow_overlapped) : in_range(0), range_begin(0), range_end(0), overlapped(allow_overlapped)
{
        sep = new irc::commasepstream(source);
        overlap_set.clear();
}

irc::portparser::~portparser()
{
        delete sep;
}

bool irc::portparser::Overlaps(long val)
{
        if (!overlapped)
                return false;

        if (overlap_set.find(val) == overlap_set.end())
        {
                overlap_set[val] = true;
                return false;
        }
        else
                return true;
}

long irc::portparser::GetToken()
{
        if (in_range > 0)
        {
                in_range++;
                if (in_range <= range_end)
                {
                        if (!Overlaps(in_range))
                        {
                                return in_range;
                        }
                        else
                        {
                                while (((Overlaps(in_range)) && (in_range <= range_end)))
                                        in_range++;
                                
                                if (in_range <= range_end)
                                        return in_range;
                        }
                }
                else
                        in_range = 0;
        }

        std::string x = sep->GetToken();

        if (x == "")
                return 0;

        while (Overlaps(atoi(x.c_str())))
        {
                x = sep->GetToken();

                if (x == "")
                        return 0;
        }

        std::string::size_type dash = x.rfind('-');
        if (dash != std::string::npos)
        {
                std::string sbegin = x.substr(0, dash);
                std::string send = x.substr(dash+1, x.length());
                range_begin = atoi(sbegin.c_str());
                range_end = atoi(send.c_str());

                if ((range_begin > 0) && (range_end > 0) && (range_begin < 65536) && (range_end < 65536) && (range_begin < range_end))
                {
                        in_range = range_begin;
                        return in_range;
                }
                else
                {
                        /* Assume its just the one port */
                        return atoi(sbegin.c_str());
                }
        }
        else
        {
                return atoi(x.c_str());
        }
}

irc::dynamicbitmask::dynamicbitmask() : bits_size(4)
{
        /* We start with 4 bytes allocated which is room
         * for 4 items. Something makes me doubt its worth
         * allocating less than 4 bytes.
         */
        bits = new unsigned char[bits_size];
        memset(bits, 0, bits_size);
}

irc::dynamicbitmask::~dynamicbitmask()
{
        /* Tidy up the entire used memory on delete */
        delete[] bits;
}
                          
irc::bitfield irc::dynamicbitmask::Allocate()
{
        /* Yeah, this isnt too efficient, however a module or the core
         * should only be allocating bitfields on load, the Toggle and
         * Get methods are O(1) as these are called much more often.
         */
        unsigned char* freebits = this->GetFreeBits();
        for (unsigned char i = 0; i < bits_size; i++)
        {
                /* Yes, this is right. You'll notice we terminate the  loop when !current_pos,
                 * this is because we logic shift our bit off the end of unsigned char, and its
                 * lost, making the loop counter 0 when we're done.
                 */
                for (unsigned char current_pos = 1; current_pos; current_pos = current_pos << 1)
                {
                        if (!(freebits[i] & current_pos))
                        {
                                freebits[i] |= current_pos;
                                return std::make_pair(i, current_pos);
                        }
                }
        }
        /* We dont have any free space left, increase by one */

        if (bits_size == 255)
                /* Oh dear, cant grow it any further */
                throw std::bad_alloc();

        unsigned char old_bits_size = bits_size;
        bits_size++;
        /* Allocate new bitfield space */
        unsigned char* temp_bits = new unsigned char[bits_size];
        unsigned char* temp_freebits = new unsigned char[bits_size];
        /* Copy the old data in */
        memcpy(temp_bits, bits, old_bits_size);
        memcpy(temp_freebits, freebits, old_bits_size);
        /* Delete the old data pointers */
        delete[] bits;
        delete[] freebits;
        /* Swap the pointers over so now the new 
         * pointers point to our member values
         */
        bits = temp_bits;
        freebits = temp_freebits;
        this->SetFreeBits(freebits);
        /* Initialize the new byte on the end of
         * the bitfields, pre-allocate the one bit
         * for this allocation
         */
        bits[old_bits_size] = 0;
        freebits[old_bits_size] = 1;
        /* We already know where we just allocated
         * the bitfield, so no loop needed
         */
        return std::make_pair(old_bits_size, 1);
}

bool irc::dynamicbitmask::Deallocate(irc::bitfield &pos)
{
        /* We dont bother to shrink the bitfield
         * on deallocation, the most we could do
         * is save one byte (!) and this would cost
         * us a loop (ugly O(n) stuff) so we just
         * clear the bit and leave the memory
         * claimed -- nobody will care about one
         * byte.
         */
        if (pos.first < bits_size)
        {
                this->GetFreeBits()[pos.first] &= ~pos.second;
                return true;
        }
        /* They gave a bitfield outside of the
         * length of our array. BAD programmer.
         */
        return false;
}

void irc::dynamicbitmask::Toggle(irc::bitfield &pos, bool state)
{
        /* Range check the value */
        if (pos.first < bits_size)
        {
                if (state)
                        /* Set state, OR the state in */
                        bits[pos.first] |= pos.second;
                else
                        /* Clear state, AND the !state out */
                        bits[pos.first] &= ~pos.second;
        }
}

bool irc::dynamicbitmask::Get(irc::bitfield &pos)
{
        /* Range check the value */
        if (pos.first < bits_size)
                return (bits[pos.first] & pos.second);
        else
                /* We can't return false, otherwise we can't
                 * distinguish between failure and a cleared bit!
                 * Our only sensible choice is to throw (ew).
                 */
                throw ModuleException("irc::dynamicbitmask::Get(): Invalid bitfield, out of range");
}

unsigned char irc::dynamicbitmask::GetSize()
{
        return bits_size;
}