// hash the space (or +) separated list of numbers in this string
//template<class Key_t, class Val_t>
//bool HashTableT<Key_t,Val_t>::hashFromString ( HashTableT *ht , char *x ) {
bool hashFromString ( HashTableT<long long,char> *ht , char *x ) {
if ( ! x ) return true;
char *xend = x + gbstrlen(x);
long n = 1;
for ( char *s = x ; s < xend ; s++ )
// i am assuming this is ascii here!
if (is_wspace_a(*s)||*s == '+') n++;
// double # slots to nd*2 so that hashtable is somewhat sparse --> fast
if ( ! ht->set ( n * 2 , NULL , 0 , false ) ) return false;
// now populate with the docids
for ( char *s = x ; s < xend ; ) {
// skip the plusses
while ( s < xend && (is_wspace_a(*s) || *s == '+') ) s++;
// are we done?
if ( s >= xend ) break;
// get the docid, a long long (64 bits)
long long d = atoll ( s );
// add it, should never fail!
if ( ! ht->addKey ( d , 1 ) ) return false;
// skip till +
while ( s < xend && (*s != '+' && !is_wspace_a(*s)) ) s++;
// are we done?
if ( s >= xend ) break;
}
return true;
}
double atof2 ( const char *s, int32_t len ) {
// skip over spaces
const char *end = s + len;
while ( s < end && is_wspace_a ( *s ) ) { s++; len--; }
// return 0 if all spaces
if ( s == end ) return 0;
char tmpBuf[128];
if ( len >= 128 ) len = 127;
//strncpy ( dst , s , len );
const char *p = s;
const char *srcEnd = s + len;
char *dst = tmpBuf;
// remove commas
for ( ; p < srcEnd ; p++ ) {
// skip commas
if ( *p == ',' ) continue;
// otherwise store it
*dst++ = *p;
}
// null term
*dst = '\0';
//buf[len] = '\0';
return atof ( tmpBuf );
}
bool HttpMime::getField(const char **field, size_t *fieldLen) {
size_t currentLinePos = m_valueStartPos;
const char *colonPos = (const char *)memchr(m_currentLine + currentLinePos, ':', m_currentLineLen);
// no colon
if (colonPos == NULL) {
return false;
}
currentLinePos = colonPos - m_currentLine;
m_valueStartPos = currentLinePos + 1;
*field = m_currentLine;
*fieldLen = currentLinePos;
// strip ending whitespaces
while (*fieldLen > 0 && is_wspace_a(m_currentLine[*fieldLen - 1])) {
--(*fieldLen);
}
logTrace(g_conf.m_logTraceHttpMime, "field='%.*s'", static_cast<int>(*fieldLen), *field);
return (*fieldLen > 0);
}
bool loadUrls ( ) {
static bool s_loaded = false;
if ( s_loaded ) return true;
s_loaded = true;
// use injectme3 file
s_ubuf1.load("./injectme3");
// scan for +++URL: xxxxx
char *s = s_ubuf1.getBufStart();
for ( ; *s ; s++ ) {
if ( strncmp(s,"+++URL: ",8) ) continue;
// got one
// \0 term it for s_contentPtrs below
*s = '\0';
// find end of it
s += 8;
char *e = s;
for ( ; *e && ! is_wspace_a(*e); e++ );
// null term it
if ( *e ) *e = '\0';
// store ptr
s_ubuf2.pushLong((long)s);
// skip past that
s = e;
// point to content
s_cbuf2.pushLong((long)(s+1));
}
// make array of url ptrs
s_urlPtrs = (char **)s_ubuf2.getBufStart();
s_contentPtrs= (char **)s_cbuf2.getBufStart();
return true;
}
bool isInWhiteSpaceList ( char *p , char *buf ) {
if ( ! p ) return false;
char *match = strstr ( buf , p );
if ( ! match ) return false;
int32_t len = gbstrlen(p);
// ensure book-ended by whitespace
if ( match &&
(match == buf || is_wspace_a(match[-1])) &&
(!match[len] || is_wspace_a(match[len])) )
return true;
// no match
return false;
}
bool endsInCurly ( char *s , int32_t slen ) {
char *e = s + slen - 1;
// don't backup more than 30 chars
char *m = e - 30;
if ( m < s ) m = s;
// \0?
if ( e > m && *e == '\0' ) e--;
// scan backwards, skipping whitespace
for ( ; e > m && is_wspace_a(*e) ; e-- );
// should be a } now to be valid json
if ( e >= m && *e == '}' ) return true;
return false;
}
int64_t atoll2 ( const char *s, int32_t len ) {
// skip over spaces
const char *end = s + len;
while ( s < end && is_wspace_a ( *s ) ) s++;
// return 0 if all spaces
if ( s == end ) return 0;
int32_t i = 0;
int64_t val = 0LL;
bool negative = false;
if ( s[0] == '-' ) { negative = true; i++; }
while ( i < len && is_digit(s[i]) ) val = val * 10LL + ( s[i++] - '0');
if ( negative ) return -val;
return val;
}
long HttpRequest::getLong ( char *field , long defaultLong ) {
long len;
char *value = getValue ( field, &len, NULL );
// return default if no match
if ( ! value || len == 0 ) return defaultLong;
// otherwise, it's a match
char c = value[len];
value[len] = '\0';
long res = atol ( value );
value[len] = c;
if ( res == 0 ) {
// may be an error. if so return the default
long i = 0;
while ( i < len && is_wspace_a(value[i]) ) i++;
if ( i < len && (value[i] == '-' || value[i] == '+') ) i++;
if ( i >= len || !is_digit(value[i]) ) return defaultLong;
}
return res;
}
bool HttpMime::getValue(const char **value, size_t *valueLen) {
// strip starting whitespaces
while (is_wspace_a(m_currentLine[m_valueStartPos]) && (m_valueStartPos < m_currentLineLen)) {
++m_valueStartPos;
}
*value = m_currentLine + m_valueStartPos;
*valueLen = m_currentLineLen - m_valueStartPos;
const char *semicolonPos = (const char *)memchr(*value, ';', *valueLen);
if (semicolonPos) {
// value should end at semicolon if present
*valueLen = semicolonPos - *value;
m_attributeStartPos = semicolonPos - m_currentLine + 1;
}
logTrace(g_conf.m_logTraceHttpMime, "value='%.*s'", static_cast<int>(*valueLen), *value);
return (*valueLen > 0);
}
double HttpRequest::getDouble ( char *field , double defaultDouble ) {
long len;
char *value = getValue ( field, &len, NULL );
// return default if no match
if ( ! value || len == 0 ) return defaultDouble;
// otherwise, it's a match
char c = value[len];
value[len] = '\0';
double res = strtod ( value , NULL );
value[len] = c;
if ( res == +0.0 ) {
// may be an error. if so return the default
long i = 0;
while ( i < len && is_wspace_a(value[i]) ) i++;
if ( i < len &&
(value[i] == '-' ||
value[i] == '+' ||
value[i] == '.') ) i++;
if ( i >= len || !is_digit(value[i]) ) return defaultDouble;
}
return res;
}
bool SearchInput::set ( TcpSocket *sock , HttpRequest *r , Query *q ) {
// store list of collection #'s to search here. usually just one.
m_collnumBuf.reset();
// zero out everything, set niceness to 0
clear ( 0 ) ;
// save it now
m_socket = sock;
// get coll rec
long collLen9;
char *coll9 = r->getString ( "c" , &collLen9 );
m_firstCollnum = -1;
CollectionRec *cr = NULL;
// now convert list of space-separated coll names into list of collnums
char *p = coll9;
// if no collection list was specified look for "token=" and
// use those to make collections. hack for diffbot.
char *token = r->getString("token",NULL);
// find all collections under this token
for ( long i = 0 ; i < g_collectiondb.m_numRecs ; i++ ) {
// must not have a "&c="
if ( p ) break;
// must have a "&token="
if ( ! token ) break;
// skip if empty
CollectionRec *tmpcr = g_collectiondb.m_recs[i];
if ( ! tmpcr ) continue;
// skip if does not match token
if ( strcmp(token,tmpcr->m_diffbotToken.getBufStart()) )
continue;
// . we got a match
// . set initial junk
if ( ! cr ) {
cr = tmpcr;
m_firstCollnum = tmpcr->m_collnum;
}
// save the collection #
if ( ! m_collnumBuf.safeMemcpy ( &tmpcr->m_collnum,
sizeof(collnum_t) ) )
return false;
}
// if we had a "&c=..." in the GET request process that
if ( p ) {
loop:
char *end = p;
for ( ; *end && ! is_wspace_a(*end) ; end++ );
// temp null
char c = *end;
*end = '\0';
CollectionRec *tmpcr = g_collectiondb.getRec ( p );
// set defaults from the FIRST one
if ( tmpcr && ! cr ) {
cr = tmpcr;
m_firstCollnum = tmpcr->m_collnum;
}
if ( ! tmpcr ) {
g_errno = ENOCOLLREC;
log("query: missing collection %s",p);
g_msg = " (error: no such collection)";
return false;
}
// add to our list
if (!m_collnumBuf.safeMemcpy(&cr->m_collnum,sizeof(collnum_t)))
return false;
// restore the \0 character we wrote in there
*end = c;
// advance
p = end;
// skip to next collection name if there is one
while ( *p && is_wspace_a(*p) ) p++;
// now add it's collection # to m_collnumBuf if there
if ( *p ) goto loop;
}
//if (! coll){coll = g_conf.m_defaultColl; collLen = gbstrlen(coll); }
//if ( ! coll )
// coll = g_conf.getDefaultColl(r->getHost(), r->getHostLen());
//if ( ! coll || ! coll[0] )
// coll = "main";
//if ( ! coll ) { g_errno = ENOCOLLREC; return false; }
//collLen = gbstrlen(coll);
//CollectionRec *cr = g_collectiondb.getRec ( coll9 );
//if ( ! cr ) {
// g_errno = ENOCOLLREC;
// g_msg = " (error: no such collection)";
// return false;
//}
// set all to 0 just to avoid any inconsistencies
//long size = (char *)&m_END_TEST - (char *)&m_START;
//memset ( this , 0x00 , size );
//setToDefaults( cr , 0 ); // niceness
m_cr = cr;
if ( ! cr ) {
log("si: collection does not exist");
g_errno = ENOCOLLREC;
return false;
}
//m_coll2 = m_cr->m_coll;
//m_collLen2 = gbstrlen(m_coll2);
// from ::reset()
m_languageWeightFactor = 0.33;
// Set IP for language detection.
// (among other things)
if ( sock ) m_queryIP = sock->m_ip;
else m_queryIP = 0;
m_hr = r;
// keep ptr to the query class to use
m_q = q;
// set this here since its size can be variable
m_sq = r->getString("sq",&m_sqLen);
// negative docids
m_noDocIds = r->getString("nodocids",&m_noDocIdsLen);
// negative sites
m_noSiteIds = r->getString("nositeids",&m_noSiteIdsLen);
// Msg5e calls Msg40 with this set to true in the searchInput
// so it can analyze the entire pages of each search result so it
// can find the article start/end tag sequence indicators
m_getTitleRec = r->getLong("gettrs",0);
m_getSitePops = r->getLong("getsitepops",0 );
// does this collection ban this IP?
/*
long encapIp = 0;
m if (! cr->hasSearchPermission ( sock, encapIp ) ) {
g_errno = ENOPERM;
g_msg = " (error: permission denied)";
return false;
}
*/
// set all search parms in SearchInput to defaults
for ( long i = 0 ; i < g_parms.m_numSearchParms ; i++ ) {
Parm *m = g_parms.m_searchParms[i];
// sanity
if ( m->m_soff < 0 ) { char *xx=NULL;*xx=0; }
char *x = (char *)this + m->m_soff;
// what is the def val ptr
char *def = NULL;
if ( m->m_off >= 0 && m->m_obj == OBJ_COLL )
def = ((char *)cr) + m->m_off;
else if ( m->m_off >= 0 && m->m_obj == OBJ_CONF )
def = ((char *)&g_conf) + m->m_off;
// set it based on type
if ( m->m_type == TYPE_LONG ) {
long v = 0;
if ( def )
v = *(long *)def;
else if ( m->m_def )
v = atol(m->m_def);
*(long *)x = v;
}
else if ( m->m_type == TYPE_BOOL ) {
long v = 0;
if ( def )
v = *(char *)def;
else if ( m->m_def )
v = atol(m->m_def);
// sanity test!
if ( v != 0 && v != 1 )
log("query: got non-bool default "
"for bool parm %s",m->m_title);
if ( v ) *(char *)x = 1;
else *(char *)x = 0;
}
else if ( m->m_type == TYPE_CHAR ) {
if ( def )
*(char *)x = *(char *)def;
else if ( m->m_def )
*(char *)x = atol(m->m_def);
}
else if ( m->m_type == TYPE_FLOAT ) {
float v = 0;
if ( def )
v = *(float *)def;
else if ( m->m_def )
v = atof(m->m_def);
*(float *)x = (float)v;
}
else if ( m->m_type == TYPE_DOUBLE ) {
double v = 0;
if ( def )
v = *(double *)def;
else if ( m->m_def )
v = atof(m->m_def);
*(double *)x = (double)v;
}
else if ( m->m_type == TYPE_LONG_LONG ) {
long long v = 0;
if ( def )
v = *(long long *)def;
else if ( m->m_def )
v = atoll(m->m_def);
*(long long *)x = (long long)v;
}
else if ( m->m_type == TYPE_STRING ||
m->m_type == TYPE_STRINGBOX ) {
//if ( m->m_cgi && strcmp ( m->m_cgi, "erpc" ) == 0 )
// log("hey1");
//if ( m->m_cgi && strcmp ( m->m_scgi, "q" ) == 0 )
// log("hey1");
char *v = NULL;
if ( def )
v = (char *)def;
else if ( m->m_def )
v = m->m_def;
*(char **)x = v;
// set the length
if ( ! v ) *(long *)(x-4) = 0;
else *(long *)(x-4) = gbstrlen(v);
}
}
// this is just used to determine in PageResults.cpp if we should
// show admin knobs next to each result...
// default to off for now. default back on.
m_isAdmin = r->getLong("admin",1);
//if ( m_isAdmin ) m_isAdmin = g_users.hasPermission ( r,PAGE_MASTER);
// local ip?
if ( ! r->isLocal() ) m_isAdmin = 0;
// default set does not take into account g_conf,
// so we will take care of that here ourselves...
m_adFeedEnabled = g_conf.m_adFeedEnabled;
//m_excludeLinkText = g_conf.m_excludeLinkText;
//m_excludeMetaText = g_conf.m_excludeMetaText;
// we need to get some cgi values in order to correct the defaults
// based on if we're doing an xml feed, have a site: query, etc.
//long xml = r->getLong ( "xml" , 0 ); // was "raw"
long siteLen = 0; r->getString ("site",&siteLen);
long sitesLen = 0;
char *sites = r->getString ("sites",&sitesLen,NULL);
// save it if there
if ( sites && sitesLen > 0 &&
( ! m_whiteListBuf.safeStrcpy(sites)||
! m_whiteListBuf.nullTerm() ) )
return log("query: unable to strcpy whitelist");
char format = getFormatFromRequest ( r );
// now override automatic defaults for special cases
if ( format != FORMAT_HTML ) {
m_familyFilter = 0;
// this is causing me a headache when on when i dont know it
m_restrictIndexdbForQuery = false;
// this is hackish
if ( r->getLong("rt",0) ) m_restrictIndexdbForQuery=false;
m_numTopicsToDisplay = 0;
m_doQueryHighlighting = 0;
m_spellCheck = 0;
m_refs_numToGenerate = 0;
m_refs_docsToScan = 0;
// default scoring info to off
m_getDocIdScoringInfo = false;
}
else if ( m_siteLen > 0 ) {
m_restrictIndexdbForQuery = false;
m_doSiteClustering = false;
m_ipRestrictForTopics = false;
}
else if ( m_whiteListBuf.length() > 0 ) {
m_ipRestrictForTopics = false;
}
m_doIpClustering = false;
//m_sitesQueryLen = 0;
// set the user ip, "uip"
long uip = m_queryIP;
char *uipStr = m_hr->getString ("uip" , NULL );
long tmpIp = 0; if ( uipStr ) tmpIp = atoip(uipStr);
if ( tmpIp ) uip = tmpIp;
//
//
// BEGIN MAIN PARM SETTING LOOP
//
//
// loop through all possible cgi parms to set SearchInput
for ( long i = 0 ; i < g_parms.m_numSearchParms ; i++ ) {
Parm *m = g_parms.m_searchParms[i];
char *x = (char *)this + m->m_soff;
// what is the parm's cgi name?
char *cgi = m->m_scgi;
if ( ! cgi ) cgi = m->m_cgi;
// sanity check
if ( ! m->m_sparm ) {
log("query: Failed search input sanity check.");
char *xx = NULL; *xx = 0;
}
// . break it down by type now
// . get it from request and store it in SearchInput
if ( m->m_type == TYPE_LONG ) {
// default was set above
long def = *(long *)x;
// assume default
long v = def;
// but cgi parms override cookie
v = r->getLong ( cgi , v );
// but if its a privledged parm and we're not an admin
// then do not allow overrides, but m_priv of 3 means
// to not display for clients, but to allow overrides
if ( ! m_isAdmin && m->m_priv && m->m_priv!=3) v = def;
// bounds checks
if ( v < m->m_smin ) v = m->m_smin;
if ( v > m->m_smax ) v = m->m_smax;
if ( m->m_sminc >= 0 ) {
long vmin = *(long *)((char *)cr+m->m_sminc);
if ( v < vmin ) v = vmin;
}
if ( m->m_smaxc >= 0 ) {
long vmax = *(long *)((char *)cr+m->m_smaxc);
if ( v > vmax ) v = vmax;
}
// set it
*(long *)x = v;
// do not print start result num (m->m_sprop is 0 for
// "s" now)
//if ( cgi[0] == 's' && cgi[1] == '\0' ) continue;
// should we propagate it? true by default
//if ( ! m->m_sprop ) continue;
// if it is the same as its default, and the default is
// always from m_def and never from the CollectionRec,
// then do not both storing it in here! what's the
// point?
if ( v == def && m->m_off < 0 ) continue;
// if not default do not propagate
if ( v == def ) continue;
// . include for sure if explicitly provided
// . vp will be NULL if "cgi" is not explicitly listed
// as a cgi parm. otherwise, even if *vp == '\0', vp
// is non-NULL.
// . crap, it can be in the cookie now
//char *vp = r->getValue(cgi, NULL, NULL);
// if not given at all, do not propagate
//if ( ! vp ) continue;
// store in up if different from default, even if
// same as default ("def") because default may be
// changed by the admin since m->m_off >= 0
//if ( m->m_sprpg && up + gbstrlen(cgi) + 20 < upend )
// up += sprintf ( up , "%s=%li&", cgi , v );
//if ( m->m_sprpp && pp + gbstrlen(cgi) + 80 < ppend )
// pp += sprintf ( pp , "<input type=hidden "
// "name=%s value=\"%li\">\n",
// cgi , v );
}
else if ( m->m_type == TYPE_LONG_LONG ) {
// default was set above
long def = *(long long *)x;
// assume default
long long v = def;
// but cgi parms override cookie
v = r->getLongLong ( cgi , v );
// but if its a privledged parm and we're not an admin
// then do not allow overrides, but m_priv of 3 means
// to not display for clients, but to allow overrides
if ( ! m_isAdmin && m->m_priv && m->m_priv!=3) v = def;
// set it
*(long long *)x = v;
// if it is the same as its default, and the default is
// always from m_def and never from the CollectionRec,
// then do not both storing it in here! what's the
// point?
if ( v == def && m->m_off < 0 ) continue;
// if not default do not propagate
if ( v == def ) continue;
}
else if ( m->m_type == TYPE_FLOAT ) {
// default was set above
float def = *(float *)x;
// get overriding from http request, if any
float v;
// but if its a privledged parm and we're not an admin
// then do not allow overrides
if ( ! m_isAdmin && m->m_priv && m->m_priv!=3) v = def;
else v = r->getFloat( cgi , def );
// bounds checks
if ( v < m->m_smin ) v = m->m_smin;
if ( v > m->m_smax ) v = m->m_smax;
if ( m->m_sminc >= 0 ) {
float vmin = *(float *)((char *)cr+m->m_sminc);
if ( v < vmin ) v = vmin;
}
if ( m->m_smaxc >= 0 ) {
float vmax = *(float *)((char *)cr+m->m_smaxc);
if ( v > vmax ) v = vmax;
}
// set it
*(float *)x = v;
// do not print start result num
//if ( cgi[0] == 's' && cgi[1] == '\0' ) continue;
// include for sure if explicitly provided
char *vp = r->getValue(cgi, NULL, NULL);
if ( ! vp ) continue;
// unchanged from default?
if ( v == def ) continue;
// store in up different from default
//if ((vp||v!= def) && up + gbstrlen(cgi)+20 < upend )
// up += sprintf ( up , "%s=%f&", cgi , v );
//if ((vp||v!= def) && pp + gbstrlen(cgi)+20 < ppend )
// pp += sprintf ( pp , "<input type=hidden "
// "name=%s value=\"%f\">\n",
// cgi , v );
}
else if ( m->m_type == TYPE_DOUBLE ) {
// default was set above
double def = *(double *)x;
// get overriding from http request, if any
double v;
// but if its a privledged parm and we're not an admin
// then do not allow overrides
if ( ! m_isAdmin && m->m_priv && m->m_priv!=3) v = def;
else v = r->getDouble( cgi , def );
// bounds checks
if ( v < m->m_smin ) v = m->m_smin;
if ( v > m->m_smax ) v = m->m_smax;
if ( m->m_sminc >= 0 ) {
double vmin=*(double *)((char *)cr+m->m_sminc);
if ( v < vmin ) v = vmin;
}
if ( m->m_smaxc >= 0 ) {
double vmax=*(double *)((char *)cr+m->m_smaxc);
if ( v > vmax ) v = vmax;
}
// set it
*(double *)x = v;
// include for sure if explicitly provided
char *vp = r->getValue(cgi, NULL, NULL);
if ( ! vp ) continue;
// unchanged from default?
if ( v == def ) continue;
}
else if ( m->m_type == TYPE_BOOL ) {
// default was set above
long def = *(char *)x;
if ( def != 0 ) def = 1; // normalize
// assume default
long v = def;
// cgi parms override cookie
v = r->getBool ( cgi , v );
// but if no perm, use default
if ( ! m_isAdmin && m->m_priv && m->m_priv!=3) v = def;
if ( v != 0 ) v = 1; // normalize
*(char *)x = v;
// don't propagate rcache
//if ( ! strcmp(cgi,"rcache") ) continue;
// should we propagate it? true by default
//if ( ! m->m_sprop ) continue;
// if it is the same as its default, and the default is
// always from m_def and never from the CollectionRec,
// then do not both storing it in here! what's the
// point?
if ( v == def && m->m_off < 0 ) continue;
// if not default do not propagate
if ( v == def ) continue;
// . include for sure if explicitly provided
// . vp will be NULL if "cgi" is not explicitly listed
// as a cgi parm. otherwise, even if *vp == '\0', vp
// is non-NULL.
// . crap, it can be in the cookie now!
//char *vp = r->getValue(cgi, NULL, NULL);
// if not given at all, do not propagate
//if ( ! vp ) continue;
// store in up if different from default, even if
// same as default ("def") because default may be
// changed by the admin since m->m_off >= 0
//if ( m->m_sprpg && up + gbstrlen(cgi) + 10 < upend )
// up += sprintf ( up , "%s=%li&", cgi , v );
//if ( m->m_sprpp && pp + gbstrlen(cgi) + 80 < ppend )
// pp += sprintf ( pp , "<input type=hidden "
// "name=%s value=\"%li\">\n",
// cgi , v );
}
else if ( m->m_type == TYPE_CHAR ) {
// default was set above
char def = *(char *)x;
*(char *)x = r->getLong ( cgi, def );
// use this
long v = *(char *)x;
// store in up if different from default, even if
// same as default ("def") because default may be
// changed by the admin since m->m_off >= 0. nah,
// let's try to reduce cgi parm pollution...
if ( v == def ) continue;
//if ( m->m_sprpg && up + gbstrlen(cgi) + 10 < upend )
// up += sprintf ( up , "%s=%li&", cgi , v );
//if ( m->m_sprpp && pp + gbstrlen(cgi) + 80 < ppend )
// pp += sprintf ( pp , "<input type=hidden "
// "name=%s value=\"%li\">\n",
// cgi , v );
}
else if ( m->m_type == TYPE_STRING ||
m->m_type == TYPE_STRINGBOX ) {
//if ( m->m_cgi && strcmp ( m->m_cgi, "qlang" ) == 0 )
// log("hey2");
char *def = *(char **)x;
// get overriding from http request, if any
long len = 0;
char *v = NULL;
// . cgi parms override cookie
// . is this url encoded?
v = r->getString ( cgi , &len , v );
// if not specified explicitly, default it and continue
if ( ! v ) {
// sanity
if ( ! def ) def = "";
*(char **)x = def;
// length preceeds char ptr in SearchInput
*(long *)(x - 4) = gbstrlen(def);
continue;
}
// if something was specified, override, it might
// be length zero, too
*(char **)x = v;
// length preceeds char ptr in SearchInput
*(long *)(x - 4) = len;
// do not store if query, that needs to be last so
// related topics can append to it
//if ( cgi[0] == 'q' && cgi[1] == '\0' ) continue;
// should we propagate it? true by default
//if ( ! m->m_sprop ) continue;
// if not given at all, do not propagate
//if ( ! vp ) continue;
// if it is the same as its default, and the default is
// always from m_def and never from the CollectionRec,
// then do not both storing it in here! what's the
// point?
//if ( v && v == def && !strcmp(def,v) && m->m_off < 0)
// continue;
// Need to set qcs based on page encoding...
// not propagated
if (!strncmp(cgi, "qcs", 3))
continue;
// do not propagate defaults
if ( v == def ) continue;
// store in up if different from default, even if
// same as default ("def") because default may be
// changed by the admin since m->m_off >= 0
//if( m->m_sprpg && up+gbstrlen(cgi)+len+6 < upend ) {
// up += sprintf ( up , "%s=", cgi );
// up += urlEncode ( up , upend-up-2 , v , len );
// *up++ = '&';
//}
// propogate hidden inputs
//if ( m->m_sprpp && up+gbstrlen(cgi)+len+80 < upend )
// pp += sprintf ( pp , "<input type=hidden "
// "name=%s value=\"%s\">\n",
// cgi , v );
}
}
// now add the special "qh" parm whose default value changes
// depending on if we are widget related or not
long qhDefault = 1;
m_doQueryHighlighting = r->getLong("qh",qhDefault);
//
// TODO: use Parms.cpp defaults
//
TopicGroup *tg = &m_topicGroups[0];
//
//
// gigabits
//
//
tg->m_numTopics = 50;
tg->m_maxTopics = 50;
tg->m_docsToScanForTopics = m_docsToScanForTopics;
tg->m_minTopicScore = 0;
tg->m_maxWordsPerTopic = 6;
tg->m_meta[0] = '\0';
tg->m_delimeter = '\0';
tg->m_useIdfForTopics = false;
tg->m_dedup = true;
// need to be on at least 2 pages!
tg->m_minDocCount = 2;
tg->m_ipRestrict = true;
tg->m_dedupSamplePercent = 80;
tg->m_topicRemoveOverlaps = true;
tg->m_topicSampleSize = 4096;
// max sequential punct chars allowedin a topic
tg->m_topicMaxPunctLen = 1;
m_numTopicGroups = 1;
// use "&dg=1" to debug gigabits
m_debugGigabits = r->getLong("dg",0);
// override
m_format = format;
// . omit scoring info from the xml feed for now
// . we have to roll this out to gk144 net i think
//if ( m_format != FORMAT_HTML )
// m_getDocIdScoringInfo = 0;
// turn off by default!
if ( ! r->getLong("gigabits",0) ) {
m_numTopicGroups = 0;
}
//////////////////////////////////////
//
// transform input into classes
//
//////////////////////////////////////
// USER_ADMIN, ...
m_username = g_users.getUsername(r);
// if collection is NULL default to one in g_conf
//if ( ! m_coll2 || ! m_coll2[0] ) {
// //m_coll = g_conf.m_defaultColl;
// m_coll2 = g_conf.getDefaultColl(r->getHost(), r->getHostLen());
// m_collLen2 = gbstrlen(m_coll2);
//}
// reset this
m_gblang = 0;
// use gblang then!
long gglen;
char *gg = r->getString ( "clang" , &gglen , NULL );
if ( gg && gglen > 1 )
m_gblang = getLanguageFromAbbr(gg);
// allow for "qlang" if still don't have it
//long gglen2;
//char *gg2 = r->getString ( "qlang" , &gglen2 , NULL );
//if ( m_gblang == 0 && gg2 && gglen2 > 1 )
// m_gblang = getLanguageFromAbbr(gg2);
// fix query by removing lang:xx from ask.com queries
//char *end = m_query + m_queryLen -8;
//if ( m_queryLen > 8 && m_query && end > m_query &&
// strncmp(end," lang:",6)==0 ) {
// char *asklang = m_query+m_queryLen - 2;
// m_gblang = getLanguageFromAbbr(asklang);
// m_queryLen -= 8;
// m_query[m_queryLen] = 0;
//
//}
// . returns false and sets g_errno on error
// . sets m_qbuf1 and m_qbuf2
if ( ! setQueryBuffers (r) )
return log("query: setQueryBuffers: %s",mstrerror(g_errno));
/* --- Virtual host language detection --- */
if(r->getHost()) {
bool langset = getLanguageFromAbbr(m_defaultSortLanguage);
char *cp;
if(!langset && (cp = strrchr(r->getHost(), '.'))) {
uint8_t lang = getLanguageFromUserAgent(++cp);
if(lang) {
// char langbuf[128];
// sprintf(langbuf, "qlang=%s\0", getLanguageAbbr(lang));
//m_defaultSortLanguage = getLanguageAbbr(lang);
char *tmp = getLanguageAbbr(lang);
strncpy(m_defaultSortLanguage, tmp, 6);
// log(LOG_INFO,
// getLanguageString(lang), r->getHost(), this);
}
}
}
/* --- End Virtual host language detection --- */
//char *qs1 = m_defaultSortLanguage;
// this overrides though
//long qlen2;
//char *qs2 = r->getString ("qlang",&qlen2,NULL);
//if ( qs2 ) qs1 = qs2;
//m_queryLang = getLanguageFromAbbr ( qs1 );
m_queryLang = detectQueryLanguage();
char *qs1 = getLangAbbr(m_queryLang);
log("query: using default lang of %s",getLangAbbr(m_queryLang));
if ( qs1 && qs1[0] && ! m_queryLang )
log("query: qlang of \"%s\" is NOT SUPPORTED",qs1);
char *prepend = r->getString("prepend",NULL,NULL);
// . the query to use for highlighting... can be overriden with "hq"
// . we need the language id for doing synonyms
if ( prepend && prepend[0] )
m_hqq.set2 ( prepend , m_queryLang , true );
else if ( m_highlightQuery && m_highlightQuery[0] )
m_hqq.set2 ( m_highlightQuery , m_queryLang , true );
else if ( m_query && m_query[0] )
m_hqq.set2 ( m_query , m_queryLang , true );
// log it here
log("query: got query %s",m_sbuf1.getBufStart());
// . now set from m_qbuf1, the advanced/composite query buffer
// . returns false and sets g_errno on error (ETOOMANYOPERANDS)
if ( ! m_q->set2 ( m_sbuf1.getBufStart(),
m_queryLang ,
m_queryExpansion ) ) {
g_msg = " (error: query has too many operands)";
return false;
}
if ( m_q->m_truncated && m_q->m_isBoolean ) {
g_errno = ETOOMANYOPERANDS;
g_msg = " (error: query has too many operands)";
return false;
}
// do not allow querier to use the links: query operator unless they
// are admin or the search controls explicitly allow links:
//if ( m_q->m_hasLinksOperator && ! m_isAdmin &&
// !cr->m_allowLinksSearch ) {
// g_errno = ENOPERM;
// g_msg = " (error: permission denied)";
// return false;
//}
// miscellaneous
m_showBanned = false;
//if ( m_isAdmin ) m_showBanned = true;
// admin can say &sb=0 explicitly to not show banned results
// . if you are searching a diffbot collection, you are the admin
// i guess...
if ( m_isAdmin || cr->m_isCustomCrawl )
m_showBanned = r->getLong("sb",m_showBanned);
if ( m_q->m_hasUrlField ) m_ipRestrictForTopics = false;
if ( m_q->m_hasIpField ) {
m_ipRestrictForTopics = false;
//if( m_isAdmin ) m_showBanned = true;
}
if ( m_q->m_hasPositiveSiteField ) {
m_ipRestrictForTopics = false;
m_doSiteClustering = false;
}
if ( m_q->m_hasQuotaField ) {
m_doSiteClustering = false;
m_doDupContentRemoval = false;
}
m_familyFilter = r->getLong("ff",0);
long codeLen;
char *code = r->getString ("code",&codeLen,NULL);
// set m_endUser
if ( ! codeLen || ! code || strcmp(code,"gbfront")==0 )
m_endUser = true;
else
m_endUser = false;
if(codeLen && !m_endUser) {
m_maxResults = cr->m_maxSearchResultsForClients;
}
else {
m_maxResults = cr->m_maxSearchResults;
}
// don't let admin bewilder himself
if ( m_maxResults < 1 ) m_maxResults = 500;
// we can't get this kind of constraint from generic Parms routines
if ( m_firstResultNum + m_docsWanted > m_maxResults )
m_firstResultNum = m_maxResults - m_docsWanted;
if(m_firstResultNum < 0) m_firstResultNum = 0;
// DEBUG: temp hack
// static bool first = true;
// if ( first ) {
// first = false;
// m_firstResultNum = 10;
// }
// if useCache is -1 then pick a default value
if ( m_useCache == -1 ) {
// assume yes as default
m_useCache = 1;
// . if query has url: or site: term do NOT use cache by def.
// . however, if spider is off then use the cache by default
if ( g_conf.m_spideringEnabled ) {
if ( m_q->m_hasPositiveSiteField ) m_useCache = 0;
else if ( m_q->m_hasIpField ) m_useCache = 0;
else if ( m_q->m_hasUrlField ) m_useCache = 0;
else if ( m_siteLen > 0 ) m_useCache = 0;
else if ( m_whiteListBuf.length() ) m_useCache = 0;
else if ( m_urlLen > 0 ) m_useCache = 0;
}
}
// never use cache if doing a rerank (msg3b)
//if ( m_rerankRuleset >= 0 ) m_useCache = 0;
bool readFromCache = false;
if ( m_useCache == 1 ) readFromCache = true;
if ( m_rcache == 0 ) readFromCache = false;
if ( m_useCache == 0 ) readFromCache = false;
// if useCache is false, don't write to cache if it was not specified
if ( m_wcache == -1 ) {
if ( m_useCache == 0 ) m_wcache = 0;
else m_wcache = 1;
}
// save it
m_rcache = readFromCache;
/*
m_language = 0;
// convert m_languageCode to a number for m_language
if ( m_languageCode ) {
m_language = (unsigned char)atoi(m_languageCode);
if ( m_language == 0 )
m_language = getLanguageFromAbbr(m_languageCode);
}
*/
// a hack for buzz for backwards compatibility
//if ( strstr ( m_q->m_orig,"gbkeyword:r36p1" ) )
// m_ruleset = 36;
//
// . turn this off for now
// . it is used in setClusterLevels() to use clusterdb to filter our
// search results via Msg39, so it is not the most efficient.
// . plus i am deleting most foreign language pages from the index
// so we can just focus on english and that will give us more english
// pages that we could normally get. we don't have resources to
// de-spam the other languages, etc.
// . turn it back on, i took out the setClusterLevels() use of that
// because we got the langid in the posdb keys now
//
//m_language = 0;
// convert m_defaultSortCountry to a number for m_countryHint
m_countryHint = g_countryCode.getIndexOfAbbr(m_defaultSortCountry);
return true;
}
// Return the value of the specified "field" within this node.
// the case of "field" does not matter.
char *XmlNode::getFieldValue ( char *field , int32_t *valueLen ) {
// reset this to 0
*valueLen = 0;
// scan for the field name in our node
int32_t flen = gbstrlen(field);
char inQuotes = '\0';
int32_t i;
// scan the characters in the node, looking for the field name in ascii
for ( i = 1; i + flen < m_nodeLen ; i++ ) {
// skip the field if it's quoted
if ( inQuotes) {
if (m_node[i] == inQuotes ) inQuotes = 0;
continue;
}
// set inQuotes to the quote if we're in quotes
if ( (m_node[i]=='\"' || m_node[i]=='\'')){
inQuotes = m_node[i];
continue;
}
// a field name must be preceeded by non-alnum
if ( is_alnum_a ( m_node[i-1] ) ) continue;
// the first character of this field shout match field[0]
if ( to_lower_a (m_node[i]) != to_lower_a(field[0] )) continue;
// field just be immediately followed by an = or space
if (m_node[i+flen]!='='&&!is_wspace_a(m_node[i+flen]))continue;
// field names must match
if ( strncasecmp ( &m_node[i], field, flen ) != 0 ) continue;
// break cuz we got a match for our field name
break;
}
// return NULL if no matching field
if ( i + flen >= m_nodeLen ) return NULL;
// advance i over the fieldname so it pts to = or space
i += flen;
// advance i over spaces
while ( i < m_nodeLen && is_wspace_a ( m_node[i] ) ) i++;
// advance over the equal sign, return NULL if does not exist
if ( i < m_nodeLen && m_node[i++] != '=' ) return NULL;
// advance i over spaces after the equal sign
while ( i < m_nodeLen && is_wspace_a ( m_node[i] ) ) i++;
// now parse out the value of this field (could be in quotes)
inQuotes = '\0';
// set inQuotes to the quote if we're in quotes
if ( m_node[i]=='\"' || m_node[i]=='\'') inQuotes = m_node[i++];
// mark this as the start of the value
int start=i;
// advance i until we hit a space, or we hit a that quote if inQuotes
if (inQuotes) {
while (i<m_nodeLen && m_node[i] != inQuotes )
i++;
}
else {
while ( i<m_nodeLen &&
!is_wspace_a(m_node[i])&&
m_node[i]!='>')
i++;
}
// set the length of the value
*valueLen = i - start;
// return a ptr to the value
return m_node + start;
}
// . return the length of a node starting at "node"
int32_t getTagLen ( char *node ) { // , int32_t version ) {
// see if it's not a node
//if ( node[0] != '<' ) return 0;
// skip over first <
int32_t i ;
// . keep looping until we hit a < or > OR while we're in quotes
// . ignore < and > when they're in quotes
for ( i = 1 ; node[i] ; i++ ) {
// this switch should speed things up... no!
if ( node[i] != '<' &&
node[i] != '>' &&
node[i] != '\"' &&
node[i] != '\'' )
continue;
// this is about 1.3 times faster than above (with -O2 on both)
//if ( ! is_tag_control_char ( node[i] ) ) continue;
if ( node[i] == '<' ) break;
if ( node[i] == '>' ) {
break;
//if ( node[i-1]!='b') break;
//if ( i -2 < 0 ) break;
//if ( node[i-2]!='g') break;
// we had a "gb>" which means that these 3 chars
// we originally a > html encoded entity which
// we decoded for easier parsing
//continue;
}
//if (version >= 70 && version < 77) continue;
// we can have double quotes within single quotes
if ( node [ i ] == '\"' ) {
// scan back looking for equal sign...
int32_t k; for ( k = i - 1 ; k > 1 ; k-- ) {
if ( is_wspace_a(node[k]) ) continue;
break;
}
if ( k <= 1 ) continue;
// . if an equal sign did not immediately preceed
// this double quote then ignore the double quote
// . this now fixes the harwoodmuseum.org issue
// talked about below
if ( node[k] != '=' ) continue;
// skip over this first quote
i++;
while ( node[i] && node[i]!='\"' ) {
// crap some pages have unbalanced quotes.
// see /test/doc.14541556377486183454.html
if ( node[i ]=='>' &&
node[i-1]=='\"' ) {
i--;
break;
}
// like an img tag hits a </a> for
// http://www.harwoodmuseum.org/press_deta
// il.php?ID=44
// BUT this f***s up
// onclick="tb_show('<b>Community Calendar</b>'
// on the </b> which is legitamately in quotes
//if ( node[i ]=='<' &&
// node[i+1]=='/' ) {
// i--;
// break;
//}
if ( node[i ]=='>' &&
node[i-1]==' ' &&
node[i-2]=='\"' ) {
i--;
break;
}
// skip this char
i++;
}
// return the length if tag ended abuptly
if ( ! node[i] ) return i;
// back-to-back quotes? common mistake
if ( node[i+1] == '\"' ) i++;
continue;
}
// continue if we don't have a " '" or "='"
if ( node [ i ] != '\'' ) continue;
if ( node[i-1] != '=' && !is_wspace_a( node[i-1] ) ) continue;
// skip to end of quote
while ( node[i] && node[i]!='\'' ) i++;
}
// skip i over the >
if ( node[i] == '>' ) i++;
// . else we found no closure outside of quotes so be more stringent
// . look for closure with regard to quotes
else for ( i=1; node[i] && node[i] != '>' && node[i] != '<';i++);
// return the LENGTH of the whole node
return i ;
}
bool Log::logR ( int64_t now, int32_t type, const char *msg, bool forced ) {
if ( ! g_loggingEnabled ) {
return true;
}
// return true if we should not log this
if ( ! forced && ! shouldLog ( type , msg ) ) {
return true;
}
// get "msg"'s length
int32_t msgLen = strlen ( msg );
ScopedLock sl(s_lock);
// do a timestamp, too. use the time synced with host #0 because
// it is easier to debug because all log timestamps are in sync.
if ( now == 0 ) now = gettimeofdayInMillisecondsGlobalNoCore();
// . skip all logging if power out, we do not want to screw things up
// . allow logging for 10 seconds after power out though
if ( ! g_process.m_powerIsOn && now - g_process.m_powerOffTime >10000){
return false;
}
// chop off any spaces at the end of the msg.
while ( is_wspace_a ( msg [ msgLen - 1 ] ) && msgLen > 0 ) msgLen--;
// a tmp buffer
char tt [ MAX_LINE_LEN ];
char *p = tt;
if (m_logPrefix) {
if ( m_logTimestamps ) {
if( m_logReadableTimestamps ) {
time_t now_t = (time_t)(now / 1000);
struct tm tm_buf;
struct tm *stm = localtime_r(&now_t,&tm_buf);
p += sprintf ( p , "%04d%02d%02d-%02d%02d%02d-%03d %04" PRId32" ", stm->tm_year+1900,stm->tm_mon+1,stm->tm_mday,stm->tm_hour,stm->tm_min,stm->tm_sec,(int)(now%1000), g_hostdb.m_hostId );
} else {
if ( g_hostdb.getNumHosts() <= 999 )
p += sprintf ( p , "%" PRIu64 " %03" PRId32 " ", (uint64_t)now , g_hostdb.m_hostId );
else if ( g_hostdb.getNumHosts() <= 9999 )
p += sprintf ( p , "%" PRIu64" %04" PRId32" ", (uint64_t)now , g_hostdb.m_hostId );
else if ( g_hostdb.getNumHosts() <= 99999 )
p += sprintf ( p , "%" PRIu64" %05" PRId32" ", (uint64_t)now , g_hostdb.m_hostId );
}
}
// Get thread id. pthread_self instead?
unsigned tid=(unsigned)syscall(SYS_gettid);
p += sprintf(p, "%06u ", tid);
// Log level
p += sprintf(p, "%s ", getTypeString(type));
}
// then message itself
const char *x = msg;
int32_t avail = (MAX_LINE_LEN) - (p - tt) - 1;
if ( msgLen > avail ) msgLen = avail;
if ( *x == ':' ) x++;
if ( *x == ' ' ) x++;
strncpy ( p , x , avail );
// capitalize for consistency. no, makes grepping log msgs harder.
//if ( is_alpha_a(*p) ) *p = to_upper_a(*p);
p += strlen(p);
// back up over spaces
while ( p[-1] == ' ' ) p--;
// end in period or ? or !
//if ( p[-1] != '?' && p[-1] != '.' && p[-1] != '!' )
// *p++ = '.';
*p ='\0';
// the total length, not including the \0
int32_t tlen = p - tt;
// . filter out nasty chars from the message
// . replace with ~'s
char cs;
char *ttp = tt;
char *ttpend = tt + tlen;
for ( ; ttp < ttpend ; ttp += cs ) {
cs = getUtf8CharSize ( ttp );
if ( is_binary_utf8 ( ttp ) ) {
for ( int32_t k = 0 ; k < cs ; k++ ) *ttp++ = '.';
// careful not to skip the already skipped bytes
cs = 0;
continue;
}
}
// . if filesize would be too big then make a new log file
// . should make a new m_fd
if ( m_logFileSize + tlen+1 > MAXLOGFILESIZE && g_conf.m_logToFile )
makeNewLogFile();
if ( m_fd >= 0 ) {
write ( m_fd , tt , tlen );
write ( m_fd , "\n", 1 );
m_logFileSize += tlen + 1;
}
else {
// print it out for now
fprintf ( stderr, "%s\n", tt );
}
return false;
}
// come here once per second i guess
void Test::initTestRun ( ) {
g_errno = 0;
// . all hosts should have their g_conf.m_repairMode parm set
// . it is global now, not collection based, since we need to
// lock down titledb for the scan and there could be recs from
// the collection we are repairing in titledb's rdbtree, which,
// when dumped, would mess up our scan.
if ( ! g_conf.m_testSpiderEnabled && ! g_conf.m_testParserEnabled ) {
char *xx=NULL;*xx=0; }
// if both enabled, core
if ( g_conf.m_testSpiderEnabled && g_conf.m_testParserEnabled ) {
char *xx=NULL;*xx=0; }
// if the power went off
if ( ! g_process.m_powerIsOn ) return;
// return if currently running
// no, admin can re-init even if running now
//if ( m_isRunning ) { char *xx=NULL;*xx=0; }//return;
// must be host #0 only
if ( g_hostdb.m_myHost->m_hostId != 0 ) return;
// if was initially in this mode, don't do anything
//if ( m_testSpiderEnabledSaved ) return;
//if ( m_testParserEnabledSaved ) return;
// you must have the "test" coll already setup!
CollectionRec *cr = g_collectiondb.getRec("test");
if ( ! cr ) {
// note it
log("test: please add a collection named \"test\" first.");
// stop the test
g_conf.m_testParserEnabled = false;
g_conf.m_testSpiderEnabled = false;
// all done
return;
}
char *testDir = getTestDir();
// scan for file named "run.start.%li.txt" which is a dump of all
// the conf and parms
char filename[100];
File f;
long i; for ( i = 0 ; i < 9999 ; i++ ) {
// make filename. base it off working dir, g_hostdb.m_dir
sprintf ( filename,"%s/%s/run.%li.collparms.txt",
g_hostdb.m_dir,testDir,i );
// exist?
f.set ( filename );
// open files
long status = f.doesExist();
// error?
if ( status == -1 ) {
// note it in the log
log("test: doesExist() returned -1");
// end the test
g_conf.m_testParserEnabled = false;
g_conf.m_testSpiderEnabled = false;
// all done
return;
}
// try next i if this one in use
if ( status ) continue;
// got one
break;
}
// close it
f.close();
// create the run.%li.version.txt file
char cmd[1000];
char vfile[200];
sprintf(vfile,"%s/%s/run.%li.version.txt",g_hostdb.m_dir,testDir,i);
sprintf(cmd,
"%s/gb -v >& %s ; "
"echo -n \"RUN START TIME: \" >> %s ; "
"date >> %s",
g_hostdb.m_dir,vfile,
vfile,
vfile);
system(cmd);
// save it
m_runId = i;
cr = g_collectiondb.getRec ( "test" );
if ( ! cr ) {
// and no more of this
g_conf.m_testParserEnabled = false;
g_conf.m_testSpiderEnabled = false;
return;
}
// set these
m_coll = cr->m_coll;
// turn on spiders
//cr->m_spideringEnabled = 1;
// crap i guess this too!!!
//g_conf.m_spideringEnabled = 1;
//
// log out the global parms
//
char fbuf[100];
// print our global parms into a file called run.%li.start.txt
sprintf(fbuf,"%s/%s/run.%li.confparms.txt",g_hostdb.m_dir,testDir,i);
// this saves it as xml i think
g_parms.saveToXml ( (char *)&g_conf , fbuf );
//
// log out the coll specific parms
//
// update name
sprintf(fbuf,"%s/%s/run.%li.collparms.txt",g_hostdb.m_dir,testDir,i);
// save that
g_parms.saveToXml ( (char *)cr , fbuf );
// get the list of urls to download and inject in order
sprintf(fbuf,"%s/%s/urls.txt",g_hostdb.m_dir,testDir);
// set it
f.set ( fbuf ) ;
// read it in
long fsize = f.getFileSize();
// add one for \0 termination
long need = fsize + 1;
// read it in
char *buf = (char *)mmalloc ( need ,"qatest");
// error?
if ( ! buf ) {
// note it
log("test: failed to alloc %li bytes for url buf",fsize);
// disable testing
g_conf.m_testParserEnabled = false;
g_conf.m_testSpiderEnabled = false;
// all done
return;
}
// open it
f.open ( O_RDONLY );
// read it in
long rs = f.read ( buf , fsize , 0 ) ;
// check it
if ( rs != fsize ) {
// note it
log("test: failed to read %li bytes of urls.txt file",fsize);
// disable testing
g_conf.m_testParserEnabled = false;
g_conf.m_testSpiderEnabled = false;
// all done
return;
}
// save it
m_urlBuf = buf;
// null term it just in case
buf[need-1] = '\0';
// end of it, including the terminating \0
m_urlEnd = buf + need;
// init url offset
m_urlPtr = m_urlBuf;
// reset just in case
//m_spiderLinks = false;
m_bypassMenuElimination = false;
// first check for spiderlinks=1|true
for ( char *p = m_urlBuf ; p < m_urlEnd ; p++ ) {
//if ( p[0] != 's' ) continue;
//if ( p[1] != 'p' ) continue;
//if ( ! strncmp(p,"spiderlinks",11) )
// m_spiderLinks = true;
//if ( ! strncmp(p,"bypassmenuelimination",21) )
// m_bypassMenuElimination = true;
}
// force max spiders to one because one page is often dependent
// on the previous page!
//if ( ! m_spiderLinks ) cr->m_maxNumSpiders = 1;
// need to make it 6 since some priorities essentially lock the
// ips up that have urls in higher priorities. i.e. once we dole
// a url out for ip X, then if later we add a high priority url for
// IP X it can't get spidered until the one that is doled does.
//else cr->m_maxNumSpiders = 6;
// . first space out all comments
// . comments are nice because we know why the url is in urls.txt
for ( char *p = m_urlBuf ; p < m_urlEnd ; p++ ) {
// skip if not start of a comment line
if ( *p != '#' ) continue;
// if not preceeded by a \n or start, skip
if ( p > m_urlBuf && *(p-1) != '\n' ) continue;
// ok, nuke it
for ( ; *p && *p !='\n' ; p++ ) *p = ' ';
}
// if we hit "\nSTOP\n" then white out that and all past it
for ( char *p = m_urlBuf ; p < m_urlEnd ; p++ ) {
// skip if not start of a comment line
if ( *p != '\n' ) continue;
// check it
if ( strncmp(p,"\nSTOP\n",6) ) continue;
// white out
for ( ; *p ; p++ ) {
// until we HIT RESUME
if ( *p == '\n' && ! strncmp(p,"\nRESUME\n",8) ) {
p[1] = ' ';
p[2] = ' ';
p[3] = ' ';
p[4] = ' ';
p[5] = ' ';
p[6] = ' ';
break;
}
*p = ' ';
}
// all done
//break;
}
// then NULL terminate all urls by converting all white space to \0s
for ( char *p = m_urlBuf ; p < m_urlEnd ; p++ )
// all non url chars to \0
if ( is_wspace_a(*p) ) *p = '\0';
// flag this
m_isRunning = true;
// and this
m_isAdding = true;
m_testStartTime = gettimeofdayInMilliseconds();
// set up dedup table
m_dt.set ( 8,0,0,NULL,0,false,MAX_NICENESS,"testdedup");
// remove all old files for now to avoid system diffs
log("test: beginning injection");
// . now inject each url in order, one at a time using msg7 i guess
// . returns true if all done
if ( ! injectLoop() ) return;
// close it up
//stopIt();
}
// . parse an incoming request
// . return false and set g_errno on error
// . CAUTION: we destroy "req" by replacing it's last char with a \0
// . last char must be \n or \r for it to be a proper request anyway
bool HttpRequest::set ( char *origReq , long origReqLen , TcpSocket *sock ) {
// reset number of cgi field terms
reset();
if ( ! m_reqBuf.reserve ( origReqLen + 1 ) ) {
log("http: failed to copy request: %s",mstrerror(g_errno));
return false;
}
// copy it to avoid mangling it
m_reqBuf.safeMemcpy ( origReq , origReqLen );
// NULL term
m_reqBuf.pushChar('\0');
m_reqBufValid = true;
// and point to that
char *req = m_reqBuf.getBufStart();
long reqLen = m_reqBuf.length() - 1;
// save this
m_userIP = 0; if ( sock ) m_userIP = sock->m_ip;
m_isSSL = 0; if ( sock ) m_isSSL = (bool)sock->m_ssl;
// TcpServer should always give us a NULL terminated request
if ( req[reqLen] != '\0' ) { char *xx = NULL; *xx = 0; }
// how long is the first line, the primary request
long i;
// for ( i = 0 ; i<reqLen && i<MAX_REQ_LEN &&
// req[i]!='\n' && req[i]!='\r'; i++);
// . now fill up m_buf, used to log the request
// . make sure the url was encoded correctly
// . we don't want assholes encoding every char so we can't see what
// url they are submitting to be spidered/indexed
// . also, don't de-code encoded ' ' '+' '?' '=' '&' because that would
// change the meaning of the url
// . and finally, non-ascii chars that don't display correctly
// . this should NULL terminate m_buf, too
// . turn this off for now, just try to log a different way
// m_bufLen = urlNormCode ( m_buf , MAX_REQ_LEN - 1 , req , i );
// ensure it's big enough to be a valid request
if ( reqLen < 5 ) {
log("http: got reqlen<5 = %s",req);
g_errno = EBADREQUEST;
return false;
}
// or if first line too long
//if ( i >= 1024 ) { g_errno = EBADREQUEST; return false; }
// get the type, must be GET or HEAD
if ( strncmp ( req , "GET " , 4 ) == 0 ) m_requestType = 0;
// these means a compressed reply was requested. use by query
// compression proxies.
else if ( strncmp ( req , "ZET " , 4 ) == 0 ) m_requestType = 0;
else if ( strncmp ( req , "HEAD " , 5 ) == 0 ) m_requestType = 1;
else if ( strncmp ( req , "POST " , 5 ) == 0 ) m_requestType = 2;
else {
log("http: got bad request cmd: %s",req);
g_errno = EBADREQUEST;
return false;
}
// . NULL terminate the request (a destructive operation!)
// . this removes the last \n in the trailing \r\n
// . shit, but it f***s up POST requests
if ( m_requestType != 2 ) { req [ reqLen - 1 ] = '\0'; reqLen--; }
// POST requests can be absolutely huge if you are injecting a 100MB
// file, so limit our strstrs to the end of the mime
char *d = NULL;
char dc;
// check for body if it was a POST request
if ( m_requestType == 2 ) {
d = strstr ( req , "\r\n\r\n" );
if ( d ) { dc = *d; *d = '\0'; }
else log("http: Got POST request without \\r\\n\\r\\n.");
}
// . point to the file path
// . skip over the "GET "
long filenameStart = 4 ;
// skip over extra char if it's a "HEAD " request
if ( m_requestType == 1 || m_requestType == 2 ) filenameStart++;
// are we a redirect?
i = filenameStart;
m_redirLen = 0;
if ( strncmp ( &req[i] , "/?redir=" , 8 ) == 0 ) {
for ( long k = i+8; k<reqLen && m_redirLen<126 ; k++) {
if ( req[k] == '\r' ) break;
if ( req[k] == '\n' ) break;
if ( req[k] == '\t' ) break;
if ( req[k] == ' ' ) break;
m_redir[m_redirLen++] = req[k];
}
}
m_redir[m_redirLen] = '\0';
// find a \n space \r or ? that delimits the filename
for ( i = filenameStart ; i < reqLen ; i++ ) {
if ( is_wspace_a ( req [ i ] ) ) break;
if ( req [ i ] == '?' ) break;
}
// now calc the filename length
m_filenameLen = i - filenameStart;
// return false and set g_errno if it's 0
if ( m_filenameLen <= 0 ) {
log("http: got filenameLen<=0: %s",req);
g_errno = EBADREQUEST;
return false;
}
// . bitch if too big
// . leave room for strcatting "index.html" below
if ( m_filenameLen >= MAX_HTTP_FILENAME_LEN - 10 ) {
log("http: got filenameLen>=max");
g_errno = EBADREQUEST;
return false;
}
// . decode the filename into m_filename and reassign it's length
// . decode %2F to / , etc...
m_filenameLen = urlDecode(m_filename,req+filenameStart,m_filenameLen);
// NULL terminate m_filename
m_filename [ m_filenameLen ] = '\0';
// does it have a file extension AFTER the last / in the filename?
bool hasExtension = false;
for ( long j = m_filenameLen-1 ; j >= 0 ; j-- ) {
if ( m_filename[j] == '.' ) { hasExtension = true; break; }
if ( m_filename[j] == '/' ) break;
}
// if it has no file extension append a /index.html
if ( ! hasExtension && m_filename [ m_filenameLen - 1 ] == '/' ) {
strcat ( m_filename , "index.html" );
m_filenameLen = gbstrlen ( m_filename );
}
// set file offset/size defaults
m_fileOffset = 0;
// -1 means ALL the file from m_fileOffset onwards
m_fileSize = -1;
// "e" points to where the range actually starts, if any
//char *e;
// . TODO: speed up by doing one strstr for Range: and maybe range:
// . do they have a Range: 0-100\n in the mime denoting a partial get?
//char *s = strstr ( req ,"Range:bytes=" );
//e = s + 12;
// try alternate formats
//if ( ! s ) { s = strstr ( req ,"Range: bytes=" ); e = s + 13; }
//if ( ! s ) { s = strstr ( req ,"Range: " ); e = s + 7; }
// parse out the range if we got one
//if ( s ) {
// long x = 0;
// sscanf ( e ,"%li-%li" , &m_fileOffset , &x );
// // get all file if range's 2nd number is non-existant
// if ( x == 0 ) m_fileSize = -1;
// else m_fileSize = x - m_fileOffset;
// // ensure legitimacy
// if ( m_fileOffset < 0 ) m_fileOffset = 0;
//}
// reset our hostname
m_hostLen = 0;
// assume request is NOT from local network
//m_isAdmin = false;
m_isLocal = false;
// get the virtual hostname they want to use
char *s = strstr ( req ,"Host:" );
// try alternate formats
if ( ! s ) s = strstr ( req , "host:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) !='\n' ) s = NULL;
// parse out the host if we got one
if ( s ) {
// skip field name, host:
s += 5;
// skip e to beginning of the host name after "host:"
while ( *s==' ' || *s=='\t' ) s++;
// find end of the host name
char *end = s;
while ( *end && !is_wspace_a(*end) ) end++;
// . now *end should be \0, \n, \r, ' ', ...
// . get host len
m_hostLen = end - s;
// truncate if too big
if ( m_hostLen >= 255 ) m_hostLen = 254;
// copy into hostname
memcpy ( m_host , s , m_hostLen );
}
// NULL terminate it
m_host [ m_hostLen ] = '\0';
// get Referer: field
s = strstr ( req ,"Referer:" );
// find another
if ( ! s ) s = strstr ( req ,"referer:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) !='\n' ) s = NULL;
// assume no referer
m_refLen = 0;
// parse out the referer if we got one
if ( s ) {
// skip field name, referer:
s += 8;
// skip e to beginning of the host name after ':'
while ( *s==' ' || *s=='\t' ) s++;
// find end of the host name
char *end = s;
while ( *end && !is_wspace_a(*end) ) end++;
// . now *end should be \0, \n, \r, ' ', ...
// . get len
m_refLen = end - s;
// truncate if too big
if ( m_refLen >= 255 ) m_refLen = 254;
// copy into m_ref
memcpy ( m_ref , s , m_refLen );
}
// NULL terminate it
m_ref [ m_refLen ] = '\0';
// get User-Agent: field
s = strstr ( req ,"User-Agent:" );
// find another
if ( ! s ) s = strstr ( req ,"user-agent:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) !='\n' ) s = NULL;
// assume empty
long len = 0;
// parse out the referer if we got one
if ( s ) {
// skip field name, referer:
s += 11;
// skip e to beginning of the host name after ':'
while ( *s==' ' || *s=='\t' ) s++;
// find end of the agent name
char *end = s;
while ( *end && *end!='\n' && *end!='\r' ) end++;
// . now *end should be \0, \n, \r, ' ', ...
// . get agent len
len = end - s;
// truncate if too big
if ( len > 127 ) len = 127;
// copy into m_userAgent
memcpy ( m_userAgent , s , len );
}
// NULL terminate it
m_userAgent [ len ] = '\0';
m_isMSIE = false;
if ( strstr ( m_userAgent , "MSIE" ) )
m_isMSIE = true;
// get Cookie: field
s = strstr ( req, "Cookie:" );
// find another
if ( !s ) s = strstr ( req, "cookie:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) != '\n' ) s = NULL;
// assume empty
// m_cookieBufLen = 0;
m_cookiePtr = s;
// parse out the cookie if we got one
if ( s ) {
// skip field name, Cookie:
s += 7;
// skip s to beginning of cookie after ':'
while ( *s == ' ' || *s == '\t' ) s++;
// find end of the cookie
char *end = s;
while ( *end && *end != '\n' && *end != '\r' ) end++;
// save length
m_cookieLen = end - m_cookiePtr;
// get cookie len
//m_cookieBufLen = end - s;
// trunc if too big
//if (m_cookieBufLen > 1023) m_cookieBufLen = 1023;
// copy into m_cookieBuf
//memcpy(m_cookieBuf, s, m_cookieBufLen);
}
// NULL terminate it
if ( m_cookiePtr ) m_cookiePtr[m_cookieLen] = '\0';
//m_cookieBuf[m_cookieBufLen] = '\0';
// convert every '&' in cookie to a \0 for parsing the fields
// for ( long j = 0 ; j < m_cookieBufLen ; j++ )
// if ( m_cookieBuf[j] == '&' ) m_cookieBuf[j] = '\0';
// mark it as cgi if it has a ?
bool isCgi = ( req [ i ] == '?' ) ;
// reset m_filename length to exclude the ?* stuff
if ( isCgi ) {
// skip over the '?'
i++;
// find a space the delmits end of cgi
long j;
for ( j = i; j < reqLen; j++) if (is_wspace_a(req[j])) break;
// now add it
if ( ! addCgi ( &req[i] , j-i ) ) return false;
// update i
i = j;
}
// . set path ptrs
// . the whole /cgi/14.cgi?coll=xxx&..... thang
m_path = req + filenameStart;
m_plen = i - filenameStart;
// we're local if hostname is 192.168.[0|1].y
//if ( strncmp(iptoa(sock->m_ip),"192.168.1.",10) == 0) {
// m_isAdmin = true; m_isLocal = true; }
//if ( strncmp(iptoa(sock->m_ip),"192.168.0.",10) == 0) {
// m_isAdmin = true; m_isLocal = true; }
//if(strncmp(iptoa(sock->m_ip),"192.168.1.",10) == 0) m_isLocal = true;
//if(strncmp(iptoa(sock->m_ip),"192.168.0.",10) == 0) m_isLocal = true;
if ( sock && strncmp(iptoa(sock->m_ip),"192.168.",8) == 0)
m_isLocal = true;
if ( sock && strncmp(iptoa(sock->m_ip),"10.",3) == 0)
m_isLocal = true;
// steve cook's comcast at home:
// if ( sock && strncmp(iptoa(sock->m_ip),"68.35.100.143",13) == 0)
// m_isLocal = true;
// procog's ip
// if ( sock && strncmp(iptoa(sock->m_ip),"216.168.36.21",13) == 0)
// m_isLocal = true;
// roadrunner ip
// if ( sock && strncmp(iptoa(sock->m_ip),"66.162.42.131",13) == 0)
// m_isLocal = true;
// cnsp ip
//if ( sock && strncmp(iptoa(sock->m_ip),"67.130.216.27",13) == 0)
// m_isLocal = true;
// emily parker
//if ( sock && strncmp(iptoa(sock->m_ip),"69.92.68.202",12) == 0)
//m_isLocal = true;
// 127.0.0.1
if ( sock && sock->m_ip == 16777343 )
m_isLocal = true;
// steve cook's webserver
//if ( sock && strncmp(iptoa(sock->m_ip),"216.168.36.21",13) == 0)
// m_isLocal = true;
// . also if we're coming from lenny at my house consider it local
// . this is a security risk, however... TODO: FIX!!!
//if ( sock->m_ip == atoip ("68.35.105.199" , 13 ) ) m_isAdmin = true;
// . TODO: now add any cgi data from a POST.....
// . look after the mime
//char *d = NULL;
// check for body if it was a POST request
//if ( m_requestType == 2 ) d = strstr ( req , "\r\n\r\n" );
// now put d's char back, just in case... does it really matter?
if ( d ) *d = dc;
// return true now if no cgi stuff to parse
if ( d ) {
char *post = d + 4;
long postLen = reqLen-(d+4-req) ;
// post sometimes has a \r or\n after it
while ( postLen > 0 && post[postLen-1]=='\r' ) postLen--;
// add it to m_cgiBuf, filter and everything
if ( ! addCgi ( post , postLen ) ) return false;
}
// sometimes i don't want to be admin
//if ( getLong ( "admin" , 1 ) == 0 ) m_isAdmin = false;
// success
/////
// Handle Extra parms...
char *ep = g_conf.m_extraParms;
char *epend = g_conf.m_extraParms + g_conf.m_extraParmsLen;
char *qstr = m_cgiBuf;
long qlen = m_cgiBufLen;
while (ep < epend){
char buf[AUTOBAN_TEXT_SIZE];
long bufLen = 0;
// get next substring
while (*ep && ep < epend && *ep != ' ' && *ep != '\n'){
buf[bufLen++] = *ep++;
}
// skip whitespace
while (*ep && ep < epend && *ep == ' '){
ep++;
}
// null terminate
buf[bufLen] = '\0';
// No match
if (!bufLen ||
!strnstr(qstr, qlen, buf)){
// skip to end of line
while (*ep && ep < epend && *ep != '\n') ep++;
// skip newline
while (*ep && ep < epend && *ep == '\n') ep++;
// try next substr
continue;
}
// found a match...
// get parm string
bufLen = 0;
while (*ep && ep < epend && *ep != '\n'){
buf[bufLen++] = *ep++;
}
buf[bufLen] = '\0';
// skip newline
while (*ep && ep < epend && *ep == '\n') ep++;
logf(LOG_DEBUG, "query: appending \"%s\" to query", buf);
long newSize = m_cgiBuf2Size + bufLen+1;
char *newBuf = (char*)mmalloc(newSize, "extraParms");
if (!newBuf){
return log("query: unable to allocate %ld bytes "
"for extraParms", newSize);
}
char *p = newBuf;
if (m_cgiBuf2Size) {
memcpy(newBuf, m_cgiBuf2, m_cgiBuf2Size);
p += m_cgiBuf2Size-1;
mfree(m_cgiBuf2, m_cgiBuf2Size, "extraParms");
m_cgiBuf2 = NULL;
m_cgiBuf2Size = 0;
}
memcpy(p, buf, bufLen);
m_cgiBuf2 = newBuf;
m_cgiBuf2Size = newSize;
p += bufLen;
*p = '\0';
}
// Put '\0' back into the HttpRequest buffer...
if (m_cgiBuf){
// do not mangle the "ucontent"!
long cgiBufLen = m_cgiBufLen;
cgiBufLen -= m_ucontentLen;
char *buf = m_cgiBuf;
for (long i = 0; i < cgiBufLen ; i++)
if (buf[i] == '&') buf[i] = '\0';
// don't decode the ucontent= field!
long decodeLen = m_cgiBufLen;
// so subtract that
if ( m_ucontent ) decodeLen -= m_ucontentLen;
// decode everything
long len = urlDecode ( m_cgiBuf , m_cgiBuf , decodeLen );
// we're parsing crap after the null if the last parm
// has no value
//memset(m_cgiBuf+len, '\0', m_cgiBufLen-len);
m_cgiBufLen = len;
// ensure that is null i guess
if ( ! m_ucontent ) m_cgiBuf[len] = '\0';
}
if (m_cgiBuf2){
char *buf = m_cgiBuf2;
for (long i = 0; i < m_cgiBuf2Size-1 ; i++)
if (buf[i] == '&') buf[i] = '\0';
long len = urlDecode ( m_cgiBuf2 , m_cgiBuf2 , m_cgiBuf2Size);
memset(m_cgiBuf2+len, '\0', m_cgiBuf2Size-len);
}
// . parse the fields after the ? in a cgi filename
// . or fields in the content if it's a POST
// . m_cgiBuf must be and is NULL terminated for this
parseFields ( m_cgiBuf , m_cgiBufLen );
// Add extra parms to the request.
if (m_cgiBuf2Size){
parseFields(m_cgiBuf2, m_cgiBuf2Size);
}
// urldecode the cookie buf too!!
if ( m_cookiePtr ) {
char *p = m_cookiePtr;
for (long i = 0; i < m_cookieLen ; i++) {
//if (p[i] == '&') p[i] = '\0';
// cookies are separated with ';' in the request only
if (p[i] == ';') p[i] = '\0';
// a hack for the metacookie=....
// which uses &'s to separate its subcookies
// this is a hack for msie's limit of 50 cookies
if ( p[i] == '&' ) p[i] = '\0';
// set m_metaCookie to start of meta cookie
if ( p[i] == 'm' && p[i+1] == 'e' &&
strncmp(p,"metacookie",10) == 0 )
m_metaCookie = p;
}
long len = urlDecode ( m_cookiePtr ,
m_cookiePtr,
m_cookieLen );
// we're parsing crap after the null if the last parm
// has no value
memset(m_cookiePtr+len, '\0', m_cookieLen-len);
m_cookieLen = len;
}
return true;
}
bool Log::logR ( long long now , long type , char *msg , bool asterisk ,
bool forced ) {
// filter if we should
//if ( forced ) goto skipfilter;
// return true if we should not log this
if ( ! forced && ! shouldLog ( type , msg ) ) return true;
// skipfilter:
// can we log if we're a sig handler? don't take changes
if ( g_inSigHandler )
return logLater ( now , type , msg , NULL );
//if ( g_inSigHandler ) return false;
// get "msg"'s length
long msgLen = gbstrlen ( msg );
#ifdef PTHREADS
// lock for threads
pthread_mutex_lock ( &s_lock );
#endif
// do a timestamp, too. use the time synced with host #0 because
// it is easier to debug because all log timestamps are in sync.
if ( now == 0 ) now = gettimeofdayInMillisecondsGlobalNoCore();
// . skip all logging if power out, we do not want to screw things up
// . allow logging for 10 seconds after power out though
if ( ! g_process.m_powerIsOn && now - g_process.m_powerOffTime >10000){
#ifdef PTHREADS
pthread_mutex_unlock ( &s_lock );
#endif
return false;
}
//if ( now == 0 ) now = g_nowApprox;
// chop off any spaces at the end of the msg.
while ( is_wspace_a ( msg [ msgLen - 1 ] ) && msgLen > 0 ) msgLen--;
// get this pid
pid_t pid = getpidtid();
// a tmp buffer
char tt [ MAX_LINE_LEN ];
char *p = tt;
char *pend = tt + MAX_LINE_LEN;
/*
// print timestamp, hostid, type
if ( g_hostdb.m_numHosts <= 999 )
sprintf ( p , "%llu %03li %s ",
now , g_hostdb.m_hostId , getTypeString(type) );
else if ( g_hostdb.m_numHosts <= 9999 )
sprintf ( p , "%llu %04li %s ",
now , g_hostdb.m_hostId , getTypeString(type) );
else if ( g_hostdb.m_numHosts <= 99999 )
sprintf ( p , "%llu %05li %s ",
now , g_hostdb.m_hostId , getTypeString(type) );
*/
// print timestamp, hostid, type
if ( m_logTimestamps ) {
if ( g_hostdb.m_numHosts <= 999 )
sprintf ( p , "%llu %03li ",
now , g_hostdb.m_hostId );
else if ( g_hostdb.m_numHosts <= 9999 )
sprintf ( p , "%llu %04li ",
now , g_hostdb.m_hostId );
else if ( g_hostdb.m_numHosts <= 99999 )
sprintf ( p , "%llu %05li ",
now , g_hostdb.m_hostId );
p += gbstrlen ( p );
}
// msg resource
char *x = msg;
long cc = 7;
// the first 7 bytes or up to the : must be ascii
//while ( p < pend && *x && is_alnum_a(*x) ) { *p++ = *x++; cc--; }
// space pad
//while ( cc-- > 0 ) *p++ = ' ';
// ignore the label for now...
while ( p < pend && *x && is_alnum_a(*x) ) { x++; cc--; }
// thread id if in "thread"
if ( pid != s_pid && s_pid != -1 ) {
//sprintf ( p , "[%li] " , (long)getpid() );
sprintf ( p , "[%lu] " , (unsigned long)pid );
p += gbstrlen ( p );
}
// then message itself
long avail = (MAX_LINE_LEN) - (p - tt) - 1;
if ( msgLen > avail ) msgLen = avail;
if ( *x == ':' ) x++;
if ( *x == ' ' ) x++;
strncpy ( p , x , avail );
// capitalize for consistency. no, makes grepping log msgs harder.
//if ( is_alpha_a(*p) ) *p = to_upper_a(*p);
p += gbstrlen(p);
// back up over spaces
while ( p[-1] == ' ' ) p--;
// end in period or ? or !
//if ( p[-1] != '?' && p[-1] != '.' && p[-1] != '!' )
// *p++ = '.';
*p ='\0';
// the total length, not including the \0
long tlen = p - tt;
// call sprintf, but first make sure we have room in m_buf and in
// the arrays. who know how much room the sprintf is going to need???
// NOTE: TODO: this is shaky -- fix it!
if ( m_bufPtr + tlen >= 1024 * 32 || m_numErrors >= MAX_LOG_MSGS){
// this sets m_bufPtr to 0
if ( ! dumpLog ( ) ) {
fprintf(stderr,"Log::log: could not dump to file!\n");
#ifdef PTHREADS
pthread_mutex_unlock ( &s_lock );
#endif
return false;
}
}
// . filter out nasty chars from the message
// . replace with ~'s
char cs;
char *ttp = tt;
char *ttpend = tt + tlen;
for ( ; ttp < ttpend ; ttp += cs ) {
cs = getUtf8CharSize ( ttp );
if ( is_binary_utf8 ( ttp ) ) {
for ( long k = 0 ; k < cs ; k++ ) *ttp++ = '.';
// careful not to skip the already skipped bytes
cs = 0;
continue;
}
// convert \n's and \r's to spaces
if ( *ttp == '\n' ) *ttp = ' ';
if ( *ttp == '\r' ) *ttp = ' ';
if ( *ttp == '\t' ) *ttp = ' ';
}
if ( m_fd >= 0 ) {
write ( m_fd , tt , tlen );
write ( m_fd , "\n", 1 );
}
else {
// print it out for now
fprintf ( stderr, "%s\n", tt );
}
// set the stuff in the array
m_errorMsg [m_numErrors] = msg;
m_errorMsgLen [m_numErrors] = msgLen;
m_errorTime [m_numErrors] = now;
m_errorType [m_numErrors] = type;
// increase the # of errors
m_numErrors++;
#ifdef PTHREADS
// unlock for threads
pthread_mutex_unlock ( &s_lock );
#endif
return false;
}
// . parse an incoming request
// . return false and set g_errno on error
// . CAUTION: we destroy "req" by replacing it's last char with a \0
// . last char must be \n or \r for it to be a proper request anyway
bool HttpRequest::set ( char *origReq , int32_t origReqLen , TcpSocket *sock ) {
// reset number of cgi field terms
reset();
if ( ! m_reqBuf.reserve ( origReqLen + 1 ) ) {
log("http: failed to copy request: %s",mstrerror(g_errno));
return false;
}
// copy it to avoid mangling it
m_reqBuf.safeMemcpy ( origReq , origReqLen );
// NULL term
m_reqBuf.pushChar('\0');
m_reqBufValid = true;
// and point to that
char *req = m_reqBuf.getBufStart();
if( !req ) {
log(LOG_ERROR, "http: req is NULL");
g_errno = EBADREQUEST;
return false;
}
int32_t reqLen = m_reqBuf.length() - 1;
// save this
m_userIP = sock ? sock->m_ip : 0;
m_isSSL = sock ? (sock->m_ssl!=NULL) : false;
// TcpServer should always give us a NULL terminated request
if ( req[reqLen] != '\0' ) { g_process.shutdownAbort(true); }
// how long is the first line, the primary request
// int32_t i;
// for ( i = 0 ; i<reqLen && i<MAX_REQ_LEN &&
// req[i]!='\n' && req[i]!='\r'; i++);
// . now fill up m_buf, used to log the request
// . make sure the url was encoded correctly
// . we don't want assholes encoding every char so we can't see what
// url they are submitting to be spidered/indexed
// . also, don't de-code encoded ' ' '+' '?' '=' '&' because that would
// change the meaning of the url
// . and finally, non-ascii chars that don't display correctly
// . this should NULL terminate m_buf, too
// . turn this off for now, just try to log a different way
// m_bufLen = urlNormCode ( m_buf , MAX_REQ_LEN - 1 , req , i );
// ensure it's big enough to be a valid request
if ( reqLen < 5 ) {
log(LOG_WARN, "http: got reqlen %" PRId32"<5 = %s",reqLen,req);
g_errno = EBADREQUEST;
return false;
}
int32_t cmdLen = 0;
// or if first line too long
//if ( i >= 1024 ) { g_errno = EBADREQUEST; return false; }
// get the type, must be GET or HEAD
if ( strncmp ( req , "GET " , 4 ) == 0 ) {
m_requestType = RT_GET;
cmdLen = 3;
}
// these means a compressed reply was requested. use by query
// compression proxies.
else if ( strncmp ( req , "ZET " , 4 ) == 0 ) {
m_requestType = RT_GET;
cmdLen = 3;
}
else if ( strncmp ( req , "HEAD " , 5 ) == 0 ) {
m_requestType = RT_HEAD;
cmdLen = 4;
}
else if ( strncmp ( req , "POST " , 5 ) == 0 ) {
m_requestType = RT_POST;
cmdLen = 4;
}
else if ( strncmp ( req , "CONNECT " , 8 ) == 0 ) {
// take this out until it stops losing descriptors and works
//m_requestType = RT_CONNECT;
//cmdLen = 7;
// we no longer insert section info. emmanuel gets section
// info when injecting a doc now i think in PageInject.cpp.
// we do not proxy https requests because we can't
// decrypt the page contents to cache them or to insert
// the sectiondb voting markup, so it's kinda pointless...
// and i'm not aiming to be a full-fledge squid proxy.
log("http: CONNECT request not supported because we "
"can't insert section markup and we can't cache: %s",req);
g_errno = EBADREQUEST;
return false;
}
else {
log("http: got bad request cmd: %s",req);
g_errno = EBADREQUEST;
return false;
}
// . NULL terminate the request (a destructive operation!)
// . this removes the last \n in the trailing \r\n
// . shit, but it f***s up POST requests
if ( m_requestType != RT_POST ) {
req [ reqLen - 1 ] = '\0';
reqLen--;
}
// POST requests can be absolutely huge if you are injecting a 100MB
// file, so limit our strstrs to the end of the mime
char *d = NULL;
char dc;
// check for body if it was a POST request
if ( m_requestType == RT_POST ) {
d = strstr ( req , "\r\n\r\n" );
if ( d ) { dc = *d; *d = '\0'; }
else log("http: Got POST request without \\r\\n\\r\\n.");
}
// is it a proxy request?
m_isSquidProxyRequest = false;
if ( strncmp ( req + cmdLen + 1, "http://" ,7) == 0 ||
strncmp ( req + cmdLen + 1, "https://",8) == 0 ) {
m_isSquidProxyRequest = true;
// set url parms for it
m_squidProxiedUrl = req + cmdLen + 1;
char *p = m_squidProxiedUrl + 7;
if ( *p == '/' ) p++; // https:// ?
// stop at whitespace or \0
for ( ; *p && ! is_wspace_a(*p) ; p++ );
// that's the length of it
m_squidProxiedUrlLen = p - m_squidProxiedUrl;
}
else if ( m_requestType == RT_CONNECT ) {
m_isSquidProxyRequest = true;
// set url parms for it
m_squidProxiedUrl = req + cmdLen + 1;
// usually its like CONNECT diffbot.com:443
char *p = m_squidProxiedUrl;
// stop at whitespace or \0
for ( ; *p && ! is_wspace_a(*p) ; p++ );
// that's the length of it
m_squidProxiedUrlLen = p - m_squidProxiedUrl;
}
// check authentication
char *auth = NULL;
if ( m_isSquidProxyRequest && req )
auth = strstr(req,"Proxy-authorization: Basic ");
//if ( m_isSquidProxyRequest && ! auth ) {
// log("http: no auth in proxy request %s",req);
// g_errno = EBADREQUEST;
// return false;
//}
SafeBuf tmp;
if ( auth ) {
// find end of it
char *p = auth;
for ( ; *p && *p != '\r' && *p != '\n' ; p++ );
tmp.base64Decode ( auth , p - auth );
}
// assume incorrect username/password
bool matched = false;
if ( m_isSquidProxyRequest ) {
// now try to match in g_conf.m_proxyAuth safebuf of
// username:password space-separated list
char *p = g_conf.m_proxyAuth.getBufStart();
// loop over those
for ( ; p && *p ; ) {
// skip initial white space
for ( ; *p && is_wspace_a(*p); p++ );
// skip to end of username:password thing
char *end = p;
for ( ; *end && !is_wspace_a(*end); end++);
// save
char *start = p;
// advance
p = end;
// this is always a match
if ( end-start == 3 && strncmp(start,"*:*",3) == 0 ) {
matched = true;
break;
}
// compare now
if ( tmp.length() != end-start )
continue;
if ( strncmp(tmp.getBufStart(),start,end-start) != 0 )
continue;
// we got a match
matched = true;
break;
}
}
// incorrect username:passwrod?
if ( m_isSquidProxyRequest && ! matched ) {
log("http: bad username:password in proxy request %s",req);
g_errno = EPERMDENIED;
return false;
}
// if proxy request to download a url through us, we are done
if ( m_isSquidProxyRequest ) return true;
bool multipart = false;
if ( m_requestType == 2 ) { // is POST?
char *cd ;
cd = gb_strcasestr(req,"Content-Type: multipart/form-data");
if ( cd ) multipart = true;
}
// . point to the file path
// . skip over the "GET "
int32_t filenameStart = 4 ;
// skip over extra char if it's a "HEAD " request
if ( m_requestType == RT_HEAD || m_requestType == RT_POST )
filenameStart++;
// are we a redirect?
int32_t i = filenameStart;
m_redirLen = 0;
if ( strncmp ( &req[i] , "/?redir=" , 8 ) == 0 ) {
for ( int32_t k = i+8; k<reqLen && m_redirLen<126 ; k++) {
if ( req[k] == '\r' ) break;
if ( req[k] == '\n' ) break;
if ( req[k] == '\t' ) break;
if ( req[k] == ' ' ) break;
m_redir[m_redirLen++] = req[k];
}
}
m_redir[m_redirLen] = '\0';
// find a \n space \r or ? that delimits the filename
for ( i = filenameStart ; i < reqLen ; i++ ) {
if ( is_wspace_a ( req [ i ] ) ) break;
if ( req [ i ] == '?' ) break;
}
// now calc the filename length
m_filenameLen = i - filenameStart;
// return false and set g_errno if it's 0
if ( m_filenameLen <= 0 ) {
log("http: got filenameLen<=0: %s",req);
g_errno = EBADREQUEST;
return false;
}
// . bitch if too big
// . leave room for strcatting "index.html" below
if ( m_filenameLen >= MAX_HTTP_FILENAME_LEN - 10 ) {
log("http: got filenameLen>=max");
g_errno = EBADREQUEST;
return false;
}
// . decode the filename into m_filename and reassign it's length
// . decode %2F to / , etc...
m_filenameLen = urlDecode(m_filename,req+filenameStart,m_filenameLen);
// NULL terminate m_filename
m_filename [ m_filenameLen ] = '\0';
// does it have a file extension AFTER the last / in the filename?
bool hasExtension = false;
for ( int32_t j = m_filenameLen-1 ; j >= 0 ; j-- ) {
if ( m_filename[j] == '.' ) { hasExtension = true; break; }
if ( m_filename[j] == '/' ) break;
}
// if it has no file extension append a /index.html
if ( ! hasExtension && m_filename [ m_filenameLen - 1 ] == '/' ) {
strcat ( m_filename , "index.html" );
m_filenameLen = strlen ( m_filename );
}
// . uses the TcpSocket::m_readBuf
// . if *p was ? then keep going
m_origUrlRequest = origReq + filenameStart;
char *p = origReq + m_filenameLen;
for ( ; *p && ! is_wspace_a(*p) ; p++ );
m_origUrlRequestLen = p - m_origUrlRequest;
// set file offset/size defaults
m_fileOffset = 0;
// -1 means ALL the file from m_fileOffset onwards
m_fileSize = -1;
// "e" points to where the range actually starts, if any
//char *e;
// . TODO: speed up by doing one strstr for Range: and maybe range:
// . do they have a Range: 0-100\n in the mime denoting a partial get?
//char *s = strstr ( req ,"Range:bytes=" );
//e = s + 12;
// try alternate formats
//if ( ! s ) { s = strstr ( req ,"Range: bytes=" ); e = s + 13; }
//if ( ! s ) { s = strstr ( req ,"Range: " ); e = s + 7; }
// parse out the range if we got one
//if ( s ) {
// int32_t x = 0;
// sscanf ( e ,"%" PRId32"-%" PRId32 , &m_fileOffset , &x );
// // get all file if range's 2nd number is non-existant
// if ( x == 0 ) m_fileSize = -1;
// else m_fileSize = x - m_fileOffset;
// // ensure legitimacy
// if ( m_fileOffset < 0 ) m_fileOffset = 0;
//}
// reset our hostname
m_hostLen = 0;
// assume request is NOT from local network
//m_isMasterAdmin = false;
m_isLocal = false;
// get the virtual hostname they want to use
char *s = strstr ( req ,"Host:" );
// try alternate formats
if ( ! s ) s = strstr ( req , "host:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) !='\n' ) s = NULL;
// parse out the host if we got one
if ( s ) {
// skip field name, host:
s += 5;
// skip e to beginning of the host name after "host:"
while ( *s==' ' || *s=='\t' ) s++;
// find end of the host name
char *end = s;
while ( *end && !is_wspace_a(*end) ) end++;
// . now *end should be \0, \n, \r, ' ', ...
// . get host len
m_hostLen = end - s;
// truncate if too big
if ( m_hostLen >= 255 ) m_hostLen = 254;
// copy into hostname
gbmemcpy ( m_host , s , m_hostLen );
}
// NULL terminate it
m_host [ m_hostLen ] = '\0';
// get Referer: field
s = strstr ( req ,"Referer:" );
// find another
if ( ! s ) s = strstr ( req ,"referer:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) !='\n' ) s = NULL;
// assume no referer
m_refLen = 0;
// parse out the referer if we got one
if ( s ) {
// skip field name, referer:
s += 8;
// skip e to beginning of the host name after ':'
while ( *s==' ' || *s=='\t' ) s++;
// find end of the host name
char *end = s;
while ( *end && !is_wspace_a(*end) ) end++;
// . now *end should be \0, \n, \r, ' ', ...
// . get len
m_refLen = end - s;
// truncate if too big
if ( m_refLen >= 255 ) m_refLen = 254;
// copy into m_ref
gbmemcpy ( m_ref , s , m_refLen );
}
// NULL terminate it
m_ref [ m_refLen ] = '\0';
// get User-Agent: field
s = strstr ( req ,"User-Agent:" );
// find another
if ( ! s ) s = strstr ( req ,"user-agent:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) !='\n' ) s = NULL;
// assume empty
int32_t len = 0;
// parse out the referer if we got one
if ( s ) {
// skip field name, referer:
s += 11;
// skip e to beginning of the host name after ':'
while ( *s==' ' || *s=='\t' ) s++;
// find end of the agent name
char *end = s;
while ( *end && *end!='\n' && *end!='\r' ) end++;
// . now *end should be \0, \n, \r, ' ', ...
// . get agent len
len = end - s;
// truncate if too big
if ( len > 127 ) len = 127;
// copy into m_userAgent
gbmemcpy ( m_userAgent , s , len );
}
// NULL terminate it
m_userAgent [ len ] = '\0';
// get Cookie: field
s = strstr ( req, "Cookie:" );
// find another
if ( !s ) s = strstr ( req, "cookie:" );
// must be on its own line, otherwise it's not valid
if ( s && s > req && *(s-1) != '\n' ) s = NULL;
// assume empty
// m_cookieBufLen = 0;
m_cookiePtr = s;
// parse out the cookie if we got one
if ( s ) {
// skip field name, Cookie:
s += 7;
// skip s to beginning of cookie after ':'
while ( *s == ' ' || *s == '\t' ) s++;
// find end of the cookie
char *end = s;
while ( *end && *end != '\n' && *end != '\r' ) end++;
// save length
m_cookieLen = end - m_cookiePtr;
// get cookie len
//m_cookieBufLen = end - s;
// trunc if too big
//if (m_cookieBufLen > 1023) m_cookieBufLen = 1023;
// copy into m_cookieBuf
//gbmemcpy(m_cookieBuf, s, m_cookieBufLen);
}
// NULL terminate it
if ( m_cookiePtr ) m_cookiePtr[m_cookieLen] = '\0';
//m_cookieBuf[m_cookieBufLen] = '\0';
// convert every '&' in cookie to a \0 for parsing the fields
// for ( int32_t j = 0 ; j < m_cookieBufLen ; j++ )
// if ( m_cookieBuf[j] == '&' ) m_cookieBuf[j] = '\0';
// mark it as cgi if it has a ?
bool isCgi = ( req [ i ] == '?' ) ;
// reset m_filename length to exclude the ?* stuff
if ( isCgi ) {
// skip over the '?'
i++;
// find a space the delmits end of cgi
int32_t j;
for ( j = i; j < reqLen; j++) if (is_wspace_a(req[j])) break;
// now add it
if ( ! addCgi ( &req[i] , j-i ) ) return false;
// update i
i = j;
}
// . set path ptrs
// . the whole /cgi/14.cgi?coll=xxx&..... thang
m_path = req + filenameStart;
m_plen = i - filenameStart;
// we're local if hostname is 192.168.[0|1].y
//if ( strncmp(iptoa(sock->m_ip),"192.168.1.",10) == 0) {
// m_isMasterAdmin = true; m_isLocal = true; }
//if ( strncmp(iptoa(sock->m_ip),"192.168.0.",10) == 0) {
// m_isMasterAdmin = true; m_isLocal = true; }
//if(strncmp(iptoa(sock->m_ip),"192.168.1.",10) == 0) m_isLocal = true;
//if(strncmp(iptoa(sock->m_ip),"192.168.0.",10) == 0) m_isLocal = true;
if ( sock && strncmp(iptoa(sock->m_ip),"192.168.",8) == 0)
m_isLocal = true;
if ( sock && strncmp(iptoa(sock->m_ip),"10.",3) == 0)
m_isLocal = true;
// gotta scan all ips in hosts.conf as well...
// if we are coming from any of our own hosts.conf c blocks
// consider ourselves local
uint32_t last = 0;
for ( int32_t i = 0 ; i < g_hostdb.getNumHosts() ; i++ ) {
Host *h = g_hostdb.getHost(i);
// save time with this check
if ( h->m_ip == last ) continue;
// update it
last = h->m_ip;
// returns number of top bytes in comon
int32_t nt = sock ? ipCmp ( sock->m_ip , h->m_ip ) : 0;
// at least be in the same c-block as a host in hosts.conf
if ( nt < 3 ) continue;
m_isLocal = true;
break;
}
// connectips/adminips
// for ( int32_t i = 0 ; i < g_conf.m_numConnectIps ; i++ ) {
// if ( sock->m_ip != g_conf.m_connectIps[i] ) continue;
// m_isLocal = true;
// break;
// }
// 127.0.0.1
if ( sock && sock->m_ip == 16777343 )
m_isLocal = true;
// . TODO: now add any cgi data from a POST.....
// . look after the mime
//char *d = NULL;
// check for body if it was a POST request
//if ( m_requestType == RT_POST ) d = strstr ( req , "\r\n\r\n" );
// return true now if no cgi stuff to parse
if ( d ) {
// now put d's char back, just in case... does it really matter?
*d = dc;
char *post = d + 4;
int32_t postLen = reqLen-(d+4-req) ;
// post sometimes has a \r or\n after it
while ( postLen > 0 && post[postLen-1]=='\r' ) postLen--;
// add it to m_cgiBuf, filter and everything
if ( ! addCgi ( post , postLen ) ) return false;
}
// Put '\0' back into the HttpRequest buffer...
// crap, not if we are multi-part unencoded stuff...
if ( m_cgiBuf && ! multipart ) {
// do not mangle the "ucontent"!
int32_t cgiBufLen = m_cgiBufLen;
cgiBufLen -= m_ucontentLen;
char *buf = m_cgiBuf;
for (int32_t i = 0; i < cgiBufLen ; i++)
if (buf[i] == '&') buf[i] = '\0';
// don't decode the ucontent= field!
int32_t decodeLen = m_cgiBufLen;
// so subtract that
if ( m_ucontent ) decodeLen -= m_ucontentLen;
// decode everything. fixed for %00 in &content= so it
// doesn't set our parms when injecting.
int32_t len = urlDecodeNoZeroes(m_cgiBuf,m_cgiBuf,decodeLen);
// we're parsing crap after the null if the last parm
// has no value
//memset(m_cgiBuf+len, '\0', m_cgiBufLen-len);
m_cgiBufLen = len;
// ensure that is null i guess
if ( ! m_ucontent ) m_cgiBuf[len] = '\0';
}
if (m_cgiBuf2){
char *buf = m_cgiBuf2;
for (int32_t i = 0; i < m_cgiBuf2Size-1 ; i++)
if (buf[i] == '&') buf[i] = '\0';
// decode everything. fixed for %00 in &content= so it
// doesn't set our parms when injecting.
int32_t len = urlDecodeNoZeroes ( m_cgiBuf2 ,
m_cgiBuf2 ,
m_cgiBuf2Size);
memset(m_cgiBuf2+len, '\0', m_cgiBuf2Size-len);
}
// . parse the fields after the ? in a cgi filename
// . or fields in the content if it's a POST
// . m_cgiBuf must be and is NULL terminated for this
parseFields ( m_cgiBuf , m_cgiBufLen );
// Add extra parms to the request.
if (m_cgiBuf2Size){
parseFields(m_cgiBuf2, m_cgiBuf2Size);
}
// urldecode the cookie buf too!!
if ( m_cookiePtr ) {
char *p = m_cookiePtr;
for (int32_t i = 0; i < m_cookieLen ; i++) {
//if (p[i] == '&') p[i] = '\0';
// cookies are separated with ';' in the request only
if (p[i] == ';') p[i] = '\0';
// a hack for the metacookie=....
// which uses &'s to separate its subcookies
// this is a hack for msie's limit of 50 cookies
if ( p[i] == '&' ) p[i] = '\0';
// set m_metaCookie to start of meta cookie
if ( p[i] == 'm' && p[i+1] == 'e' &&
strncmp(p,"metacookie",10) == 0 )
m_metaCookie = p;
}
int32_t len = urlDecode ( m_cookiePtr ,
m_cookiePtr,
m_cookieLen );
// we're parsing crap after the null if the last parm
// has no value
memset(m_cookiePtr+len, '\0', m_cookieLen-len);
m_cookieLen = len;
}
return true;
}
// returns false on bad mime
bool HttpMime::parse ( char *mime , long mimeLen , Url *url ) {
// reset locUrl to 0
m_locUrl.reset();
// return if we have no valid complete mime
if ( mimeLen == 0 ) return false;
// status is on first line
m_status = -1;
// skip HTTP/x.x till we hit a space
char *p = mime;
char *pend = mime + mimeLen;
while ( p < pend && !is_wspace_a(*p) ) p++;
// then skip over spaces
while ( p < pend && is_wspace_a(*p) ) p++;
// return false on a problem
if ( p == pend ) return false;
// then read in the http status
m_status = atol2 ( p , pend - p );
// if no Content-Type: mime field was provided, assume html
m_contentType = CT_HTML;
// assume default charset
m_charset = NULL;
m_charsetLen = 0;
// set contentLen, lastModifiedDate, m_cookie
p = mime;
while ( p < pend ) {
// compute the length of the string starting at p and ending
// at a \n or \r
long len = 0;
while ( &p[len] < pend && p[len]!='\n' && p[len]!='\r' ) len++;
// . if we could not find a \n or \r there was an error
// . MIMEs must always end in \n or \r
if ( &p[len] >= pend ) return false;
// . stick a NULL at the end of the line
// . overwrites \n or \r TEMPORARILY
char c = p [ len ];
p [ len ] = '\0';
// parse out some meaningful data
if ( strncasecmp ( p , "Content-Length:" ,15) == 0 ) {
m_contentLengthPos = p + 15;
m_contentLen = atol( m_contentLengthPos);
}
else if ( strncasecmp ( p , "Last-Modified:" ,14) == 0 ) {
m_lastModifiedDate=atotime(p+14);
// do not let them exceed current time for purposes
// of sorting by date using datedb (see Msg16.cpp)
time_t now = time(NULL);
if (m_lastModifiedDate > now) m_lastModifiedDate = now;
}
else if ( strncasecmp ( p , "Content-Type:" ,13) == 0 )
m_contentType = getContentTypePrivate ( p + 13 );
else if ( strncasecmp ( p , "Set-Cookie: " ,11) == 0 ) {
m_cookie = p + 11;
m_cookieLen = gbstrlen ( p + 11 );
}
else if ( strncasecmp ( p , "Location:" , 9) == 0 ) {
// point to it
char *tt = p + 9;
// skip if space
if ( *tt == ' ' ) tt++;
if ( *tt == ' ' ) tt++;
// at least set this for Msg13.cpp to use
m_locationField = tt;
m_locationFieldLen = gbstrlen(tt);
// . we don't add the "www." because of slashdot.com
// . we skip initial spaces in this Url::set() routine
if(url)
m_locUrl.set ( url, p + 9, len - 9,
false/*addWWW?*/);
}
else if ( strncasecmp ( p , "Content-Encoding:", 17) == 0 ) {
//only support gzip now, it doesn't seem like servers
//implement the other types much
m_contentEncodingPos = p+17;
if(strstr(m_contentEncodingPos, "gzip")) {
m_contentEncoding = ET_GZIP;
}
else if(strstr(m_contentEncodingPos, "deflate")) {
//zlib's compression
m_contentEncoding = ET_DEFLATE;
}
}
//else if ( strncasecmp ( p, "Cookie:", 7) == 0 )
// log (LOG_INFO, "mime: Got Cookie = %s", (p+7));
// re-insert the character that we replaced with a '\0'
p [ len ] = c;
// go to next line
p += len;
// skip over the cruft at the end of this line
while ( p < pend && ( *p=='\r' || *p=='\n' ) ) p++;
}
return true;
}
// returns false on bad mime
bool HttpMime::parse(char *mime, int32_t mimeLen, Url *url) {
#ifdef _VALGRIND_
VALGRIND_CHECK_MEM_IS_DEFINED(mime,mimeLen);
#endif
// reset locUrl to 0
m_locUrl.reset();
// return if we have no valid complete mime
if (mimeLen == 0) {
return false;
}
// status is on first line
m_status = -1;
// skip HTTP/x.x till we hit a space
char *p = mime;
char *pend = mime + mimeLen;
while (p < pend && !is_wspace_a(*p)) p++;
// then skip over spaces
while (p < pend && is_wspace_a(*p)) p++;
// return false on a problem
if (p == pend) return false;
// then read in the http status
m_status = atol2(p, pend - p);
// if no Content-Type: mime field was provided, assume html
m_contentType = CT_HTML;
// assume default charset
m_charset = NULL;
m_charsetLen = 0;
// skip over first line
getNextLine();
while (getNextLine()) {
const char *field = NULL;
size_t fieldLen = 0;
if (getField(&field, &fieldLen)) {
if (parseContentEncoding(field, fieldLen)) {
continue;
}
if (parseContentLength(field, fieldLen)) {
continue;
}
if (parseContentType(field, fieldLen)) {
continue;
}
if (parseLocation(field, fieldLen, url)) {
continue;
}
if (parseSetCookie(field, fieldLen)) {
continue;
}
// add parsing of other header here
}
}
return true;
}
char *getMatches2 ( Needle *needles ,
int32_t numNeedles ,
char *haystack ,
int32_t haystackSize ,
char *linkPos ,
int32_t *needleNum ,
bool stopAtFirstMatch ,
bool *hadPreMatch ,
bool saveQuickTables ,
int32_t niceness ) {
// assume not
if ( hadPreMatch ) *hadPreMatch = false;
// empty haystack? then no matches
if ( ! haystack || haystackSize <= 0 ) return NULL;
// JAB: no needles? then no matches
if ( ! needles || numNeedles <= 0 ) return NULL;
//char tmp[8192];
//char *t = tmp;
//char *tend = tmp + 8192;
// reset counts to 0
//if ( ! stopAtFirstMatch )
// for ( int32_t i=0 ; i < numNeedles ; i++ )
// needles[i].m_count = 0;
// are we responsible for init'ing string lengths? this is much
// faster than having to specify lengths manually.
for ( int32_t i=0 ; i < numNeedles; i++ ) {
// breathe
QUICKPOLL(niceness);
// clear
needles[i].m_count = 0;
needles[i].m_firstMatch = NULL;
// set the string size in bytes if not provided
if ( needles[i].m_stringSize == 0 )
needles[i].m_stringSize = gbstrlen(needles[i].m_string);
}
// . set up the quick tables.
// . utf16 is not as effective here because half the bytes are zeroes!
// . TODO: use a static cache of like 4 of these tables where the key
// is the Needles ptr ... done
int32_t numNeedlesToInit = numNeedles;
char space[256 * 4 * sizeof(BITVEC)];
char *buf = NULL;
BITVEC *s0;
BITVEC *s1;
BITVEC *s2;
BITVEC *s3;
/*
static bool s_quickTableInit = false;
static char s_qtbuf[128*(12+1)*2];
int32_t slot = -1;
if(saveQuickTables) {
if ( ! s_quickTableInit ) {
s_quickTableInit = true;
s_quickTables.set(8,4,128,s_qtbuf,256*13,false,0,"qx");
}
uint64_t key = (uint32_t)needles;
slot = s_quickTables.getSlot(&key);
if ( slot >= 0 ) {
buf = s_quickTables.getValueFromSlot(slot);
numNeedlesToInit = 0;
}
}
*/
if(!buf) {
buf = space;
memset ( buf , 0 , sizeof(BITVEC)*256*4);
}
/*
if( useQuickTables && slot == -1 ) {
//buf = (char*)mcalloc(sizeof(uint32_t)*256*5,
// "matches");
if(buf) s_quickTables.addKey(&key, &buf);
//sanity check, no reason why there needs to be a
//limit, I just don't expect there to be this many
//static needles at this point.
if(s_quickTables.getNumSlotsUsed() > 32){
char *xx=NULL; *xx = 0;
}
}
*/
// try 64 bit bit vectors now since we doubled # of needles
int32_t offset = 0;
s0 = (BITVEC *)(buf + offset);
offset += sizeof(BITVEC)*256;
s1 = (BITVEC *)(buf + offset);
offset += sizeof(BITVEC)*256;
s2 = (BITVEC *)(buf + offset);
offset += sizeof(BITVEC)*256;
s3 = (BITVEC *)(buf + offset);
offset += sizeof(BITVEC)*256;
BITVEC mask;
// set the letter tables, s0[] through sN[], for each needle
for ( int32_t i = 0 ; i < numNeedlesToInit ; i++ ) {
// breathe
QUICKPOLL(niceness);
unsigned char *w = (unsigned char *)needles[i].m_string;
unsigned char *wend = w + needles[i].m_stringSize;
// BITVEC is now 64 bits
mask = (1<<(i&0x3f)); // (1<<(i%64));
// if the needle is small, fill up the remaining letter tables
// with its mask... so it matches any character in haystack.
s0[(unsigned char)to_lower_a(*w)] |= mask;
s0[(unsigned char)to_upper_a(*w)] |= mask;
w += 1;//step;
if ( w >= wend ) {
for ( int32_t j = 0 ; j < 256 ; j++ ) {
s1[j] |= mask;
s2[j] |= mask;
s3[j] |= mask;
}
continue;
}
s1[(unsigned char)to_lower_a(*w)] |= mask;
s1[(unsigned char)to_upper_a(*w)] |= mask;
w += 1;//step;
if ( w >= wend ) {
for ( int32_t j = 0 ; j < 256 ; j++ ) {
s2[j] |= mask;
s3[j] |= mask;
}
continue;
}
s2[(unsigned char)to_lower_a(*w)] |= mask;
s2[(unsigned char)to_upper_a(*w)] |= mask;
w += 1;//step;
if ( w >= wend ) {
for ( int32_t j = 0 ; j < 256 ; j++ ) {
s3[j] |= mask;
}
continue;
}
s3[(unsigned char)to_lower_a(*w)] |= mask;
s3[(unsigned char)to_upper_a(*w)] |= mask;
w += 1;//step;
}
// return a ptr to the first match if we should, this is it
char *retVal = NULL;
// debug vars
//int32_t debugCount = 0;
//int32_t pp = 0;
// now find the first needle in the haystack
unsigned char *p = (unsigned char *)haystack;
unsigned char *pend = (unsigned char *)haystack + haystackSize;
char *dend = (char *)pend;
// do not breach!
pend -= 4;
for ( ; p < pend ; p++ ) {
// breathe
QUICKPOLL(niceness);
//if ( (char *)p - (char *)haystack >= 12508 )
// log("hey");
// analytics...
// is this a possible match? (this should be VERY fast)
mask = s0[*(p+0)];
if ( ! mask ) continue;
mask &= s1[*(p+1)];
if ( ! mask ) continue;
mask &= s2[*(p+2)];
if ( ! mask ) continue;
mask &= s3[*(p+3)];
if ( ! mask ) continue;
//debugCount++;
/*
// display
char oo[148];
char *xx ;
xx = oo;
//gbmemcpy ( xx , p , 8 );
for ( int32_t k = 0 ; k < 5 ; k++ ) {
*xx++ = p[k];
}
gbmemcpy ( xx , "..." , 3 );
xx += 3;
*/
//
// XXX: do a hashtable lookup here so we have the candidate
// matches in a chain...
// XXX: for small needles which match frequently let's have
// a single char hash table, a 2 byte char hash table,
// etc. so if we have small needles we check the hash
// in those tables first, but only if mask & SMALL_NEEDLE
// is true! the single byte needle hash table can just
// be a lookup table. just XOR the bytes together for
// the hash.
// XXX: just hash the mask into a table to get candidate
// matches in a chain? but there's 4B hashes!!
// we got a good candidate, loop through all the needles
for ( int32_t j = 0 ; j < numNeedles ; j++ ) {
// skip if does not match mask, will save time
if ( ! ((1<<(j&0x3f)) & mask) ) continue;
if( needles[j].m_stringSize > 3) {
// ensure first 4 bytes matches this needle's
if (needles[j].m_string[0]!=to_lower_a(*(p+0)))
continue;
if (needles[j].m_string[1]!=to_lower_a(*(p+1)))
continue;
if (needles[j].m_string[2]!=to_lower_a(*(p+2)))
continue;
if (needles[j].m_string[3]!=to_lower_a(*(p+3)))
continue;
}
// get needle size
int32_t msize = needles[j].m_stringSize;
// can p possibly be big enough?
if ( pend - p < msize ) continue;
// needle is "m" now
char *m = needles[j].m_string;
char *mend = needles[j].m_stringSize + m;
// use a tmp ptr for ptr into haystack
char *d = (char *)p;
// skip first 4 bytes since we know they match
if(msize > 3) {
d += 4;
m += 4;
}
// loop over each char in "m"
//for ( ; *m ; m++ ) {
for ( ; m < mend ; m++ ) {
//while ( ! *d && d < dend ) d++;
//while ( ! *m && m < mend ) m++;
// if we are a non alnum, that will match
// any string of non-alnums, like a space
// for instance. the 0 byte does not count
// because it is used in utf16 a lot. this
// may trigger some false matches in utf16
// but, oh well... this way "link partner"
// will match "link - partner" in the haystk
if ( is_wspace_a(*m) && m < mend ) {
// skip all in "d" then.
while (d<dend&&is_wspace_a(*d)) d++;
// advance m then
continue;
}
// make sure we match otherwise
if ( *m != to_lower_a(*d) ) break;
// ok, we matched, go to next
d++;
}
// if not null, keep going
if ( m < mend ) continue;
// if this needle is "special" AND it occurs AFTER
// linkPos, then do not consider it a match. this is
// if we have a comment section indicator, like
// "div id=\"comment" AND it occurs AFTER linkPos
// (the char ptr to our link in the haystack) then
// the match does not count.
if ( linkPos && needles[j].m_isSection &&
(char *)p>linkPos ) {
// record this for LinkText.cpp
if ( hadPreMatch ) *hadPreMatch = true;
continue;
}
// store ptr if NULL
if ( ! needles[j].m_firstMatch )
needles[j].m_firstMatch = (char *)p;
// return ptr to needle in "haystack"
if ( stopAtFirstMatch ) {
// ok, we got a match
if ( needleNum ) *needleNum = j;
//return (char *)p;
retVal = (char *)p;
p = pend;
break;
}
// otherwise, just count it
needles[j].m_count++;
// see if we match another needle, fixes bug
// of matching "anal" but not "analy[tics]"
continue;
// advance to next char in the haystack
break;
}
// ok, we did not match any needles, advance p and try again
}
//
// HACK:
//
// repeat above loop but for the last 4 characters in haystack!!
// this fixes a electric fence mem breach core
//
// it is slower because we check for \0
//
pend += 4;
for ( ; p < pend ; p++ ) {
// breathe
QUICKPOLL(niceness);
//if ( (char *)p - (char *)haystack >= 12508 )
// log("hey");
// is this a possible match? (this should be VERY fast)
mask = s0[*(p+0)];
if ( ! mask ) continue;
if ( p+1 < pend ) {
mask &= s1[*(p+1)];
if ( ! mask ) continue;
}
if ( p+2 < pend ) {
mask &= s2[*(p+2)];
if ( ! mask ) continue;
}
if ( p+3 < pend ) {
mask &= s3[*(p+3)];
if ( ! mask ) continue;
}
//debugCount++;
/*
// display
char oo[148];
char *xx ;
xx = oo;
//gbmemcpy ( xx , p , 8 );
for ( int32_t k = 0 ; k < 5 ; k++ ) {
*xx++ = p[k];
}
gbmemcpy ( xx , "..." , 3 );
xx += 3;
*/
//
// XXX: do a hashtable lookup here so we have the candidate
// matches in a chain...
// XXX: for small needles which match frequently let's have
// a single char hash table, a 2 byte char hash table,
// etc. so if we have small needles we check the hash
// in those tables first, but only if mask & SMALL_NEEDLE
// is true! the single byte needle hash table can just
// be a lookup table. just XOR the bytes together for
// the hash.
// XXX: just hash the mask into a table to get candidate
// matches in a chain? but there's 4B hashes!!
// we got a good candidate, loop through all the needles
for ( int32_t j = 0 ; j < numNeedles ; j++ ) {
// skip if does not match mask, will save time
if ( ! ((1<<(j&0x3f)) & mask) ) continue;
if( needles[j].m_stringSize > 3) {
// ensure first 4 bytes matches this needle's
if (needles[j].m_string[0]!=to_lower_a(*(p+0)))
continue;
if (!p[1] ||
needles[j].m_string[1]!=to_lower_a(*(p+1)))
continue;
if (!p[2] ||
needles[j].m_string[2]!=to_lower_a(*(p+2)))
continue;
if (!p[3] ||
needles[j].m_string[3]!=to_lower_a(*(p+3)))
continue;
}
// get needle size
int32_t msize = needles[j].m_stringSize;
// can p possibly be big enough?
if ( pend - p < msize ) continue;
// needle is "m" now
char *m = needles[j].m_string;
char *mend = needles[j].m_stringSize + m;
// use a tmp ptr for ptr into haystack
char *d = (char *)p;
// skip first 4 bytes since we know they match
if(msize > 3) {
d += 4;
m += 4;
}
// loop over each char in "m"
//for ( ; *m ; m++ ) {
for ( ; m < mend ; m++ ) {
//while ( ! *d && d < dend ) d++;
//while ( ! *m && m < mend ) m++;
// if we are a non alnum, that will match
// any string of non-alnums, like a space
// for instance. the 0 byte does not count
// because it is used in utf16 a lot. this
// may trigger some false matches in utf16
// but, oh well... this way "link partner"
// will match "link - partner" in the haystk
if ( is_wspace_a(*m) && m < mend ) {
// skip all in "d" then.
while (d<dend&&is_wspace_a(*d)) d++;
// advance m then
continue;
}
// make sure we match otherwise
if ( *m != to_lower_a(*d) ) break;
// ok, we matched, go to next
d++;
}
// if not null, keep going
if ( m < mend ) continue;
// if this needle is "special" AND it occurs AFTER
// linkPos, then do not consider it a match. this is
// if we have a comment section indicator, like
// "div id=\"comment" AND it occurs AFTER linkPos
// (the char ptr to our link in the haystack) then
// the match does not count.
if ( linkPos && needles[j].m_isSection &&
(char *)p>linkPos ) {
// record this for LinkText.cpp
if ( hadPreMatch ) *hadPreMatch = true;
continue;
}
// store ptr if NULL
if ( ! needles[j].m_firstMatch )
needles[j].m_firstMatch = (char *)p;
// return ptr to needle in "haystack"
if ( stopAtFirstMatch ) {
// ok, we got a match
if ( needleNum ) *needleNum = j;
//return (char *)p;
retVal = (char *)p;
p = pend;
break;
}
// otherwise, just count it
needles[j].m_count++;
// advance to next char in the haystack
break;
}
// ok, we did not match any needles, advance p and try again
}
//if ( debugCount > 0 ) pp = haystackSize / debugCount;
//log("build: debug count = %"INT32" uc=%"INT32" hsize=%"INT32" "
// "1 in %"INT32" chars matches.",
// debugCount,(int32_t)isHaystackUtf16,haystackSize,pp);
// before we exit, clean up
return retVal;
}
bool HttpMime::getAttribute(const char **attribute, size_t *attributeLen, const char **attributeValue, size_t *attributeValueLen) {
// initialize value
*attribute = NULL;
*attributeLen = 0;
*attributeValue = NULL;
*attributeValueLen = 0;
// no attribute
if (m_attributeStartPos == 0) {
return false;
}
// strip starting whitespaces
while (is_wspace_a(m_currentLine[m_attributeStartPos]) && (m_attributeStartPos < m_currentLineLen)) {
++m_attributeStartPos;
}
*attribute = m_currentLine + m_attributeStartPos;
*attributeLen = m_currentLineLen - m_attributeStartPos;
// next attribute
const char *semicolonPos = (const char *)memchr(*attribute, ';', *attributeLen);
if (semicolonPos) {
*attributeLen = semicolonPos - *attribute;
m_attributeStartPos = semicolonPos - m_currentLine + 1;
} else {
m_attributeStartPos = 0;
}
// attribute value
const char *equalPos = (const char *)memchr(*attribute, '=', *attributeLen);
if (equalPos) {
*attributeValueLen = *attributeLen;
*attributeLen = equalPos - *attribute;
*attributeValueLen -= *attributeLen + 1;
*attributeValue = equalPos + 1;
// strip ending attribute whitespace
while (is_wspace_a((*attribute)[*attributeLen - 1])) {
--(*attributeLen);
}
// strip starting attribute value whitespace/quote
while (is_wspace_a((*attributeValue)[0]) || (*attributeValue)[0] == '"' || (*attributeValue)[0] == '\'') {
++(*attributeValue);
--(*attributeValueLen);
}
// strip ending attribute value whitespace/quote
while (is_wspace_a((*attributeValue)[*attributeValueLen - 1]) || (*attributeValue)[*attributeValueLen - 1] == '"' || (*attributeValue)[*attributeValueLen - 1] == '\'') {
--(*attributeValueLen);
}
}
// cater for empty values between semicolon
// eg: Set-Cookie: name=value; Path=/; ;SECURE; HttpOnly;
if (*attributeLen == 0 && m_attributeStartPos) {
return getAttribute(attribute, attributeLen, attributeValue, attributeValueLen);
}
logTrace(g_conf.m_logTraceHttpMime, "attribute='%.*s' value='%.*s'", static_cast<int>(*attributeLen), *attribute, static_cast<int>(*attributeValueLen), *attributeValue);
return (*attributeLen > 0);
}
JsonItem *Json::parseJsonStringIntoJsonItems ( char *json , int32_t niceness ) {
m_prev = NULL;
m_stackPtr = 0;
m_sb.purge();
JsonItem *ji = NULL;
if ( ! json ) return NULL;
// how much space will we need to avoid any reallocs?
char *p = json;
bool inQuote = false;
int32_t need = 0;
for ( ; *p ; p++ ) {
// ignore any escaped char. also \x1234
if ( *p == '\\' ) {
if ( p[1] ) p++;
continue;
}
if ( *p == '\"' )
inQuote = ! inQuote;
if ( inQuote )
continue;
if ( *p == '{' ||
*p == ',' ||
*p == '[' ||
*p == ':' )
// +1 for null terminating string of each item
need += sizeof(JsonItem) +1;
}
// plus the length of the string to store it decoded etc.
need += p - json;
// plus a \0 for the value and a \0 for the name of each jsonitem
need += 2;
// prevent cores for now
need += 10;
// . to prevent safebuf from reallocating do this
// . safeMemcpy() calls reserve(m_length+len) and reserves
// tries to alloc m_length + (m_length+len) so since,
// m_length+len should never be more than "need" we need to
// double up here
need *= 2;
// this should be enough
if ( ! m_sb.reserve ( need ) ) return NULL;
// for testing if we realloc
char *mem = m_sb.getBufStart();
int32_t size;
char *NAME = NULL;
int32_t NAMELEN = 0;
// reset p
p = json;
// json maybe bad utf8 causing us to miss the \0 char, so use "pend"
char *pend = json + gbstrlen(json);
// scan
for ( ; p < pend ; p += size ) {
// get size
size = getUtf8CharSize ( p );
// skip spaces
if ( is_wspace_a (*p) )
continue;
// skip commas
if ( *p == ',' ) continue;
// did we hit a '{'? that means the existing json item
// is a parent of the item(s) inside the {}'s
if ( *p == '{' ) {
// if ji is non-null it must be a name like in
// \"stats\":{\"fetchTime\":2069,....}
// . this indicates the start of a json object
// . addNewItem() will push the current item on stack
ji = addNewItem();
if ( ! ji ) return NULL;
// current ji is an object type then
ji->m_type = JT_OBJECT;
// set the name
ji->m_name = NAME;
ji->m_nameLen = NAMELEN;
// this goes on the stack
if ( m_stackPtr >= MAXJSONPARENTS ) return NULL;
m_stack[m_stackPtr++] = ji;
// and null this
ji = NULL;
continue;
}
// pop the stack?
if ( *p == '}' ) {
// just pop it and restore name cursor
if ( m_stackPtr > 0 ) {
JsonItem *px = m_stack[m_stackPtr-1];
NAME = px->m_name;
NAMELEN = px->m_nameLen;
m_stackPtr--;
}
continue;
}
// array of things?
if ( *p == '[' ) {
// make a newitem to put on stack
ji = addNewItem();
if ( ! ji ) return NULL;
// current ji is an object type then
ji->m_type = JT_ARRAY;
// start of array hack. HACK!
//ji->m_valueLong = (int32_t)p;
ji->m_valueArray = p;
// set the name
ji->m_name = NAME;
ji->m_nameLen = NAMELEN;
// init to a bogus value. should be set below.
// at least this should avoid a core in XmlDoc.cpp
// getTokenizedDiffbotReply()
ji->m_valueLen = 0;
// this goes on the stack
if ( m_stackPtr >= MAXJSONPARENTS ) return NULL;
m_stack[m_stackPtr++] = ji;
ji = NULL;
continue;
}
// pop the stack?
if ( *p == ']' ) {
// just pop it and restore name cursor
if ( m_stackPtr > 0 ) {
JsonItem *px = m_stack[m_stackPtr-1];
NAME = px->m_name;
NAMELEN = px->m_nameLen;
// start of array hack. HACK!
char *start = (char *)px->m_valueArray;//Long;
// include ending ']' in length of array
px->m_valueLen = p - start + 1;
m_stackPtr--;
}
continue;
}
// a quote?
if ( *p == '\"' ) {
// find end of quote
char *end = p + 1;
for ( ; *end ; end++ ) {
// skip two chars if escaped
if ( *end == '\\' && end[1] ) {
end++;
continue;
}
// this quote is unescaped then
if ( *end == '\"' ) break;
}
// field?
char *x = end + 1;
// skip spaces
for ( ; *x && is_wspace_a(*x) ; x++ );
// define the string
char *str = p + 1;
int32_t slen = end - str;
// . if a colon follows, it was a field
if ( *x == ':' ) {
// we can't be the first thing in the safebuf
// json must start with { or [ i guess
// otherwise getFirstItem() won't work!
if ( m_sb.m_length==0 ) {
g_errno = EBADJSONPARSER;
return NULL;
}
// let's push this now so we can \0 term
char *savedStr = m_sb.getBuf();
m_sb.safeMemcpy ( str , slen );
m_sb.pushChar('\0');
// just set the name cursor
NAME = savedStr;//str;
NAMELEN = slen;
}
// . otherwise, it was field value, so index it
// . TODO: later make field names compounded to
// better represent nesting?
// . added 'else if (NAME){' fix for json=\"too small\"
else if ( NAME ) {
// make a new one in safebuf. our
// parent will be the array type item.
ji = addNewItem();
if ( ! ji ) return NULL;
// we are a string
ji->m_type = JT_STRING;
// use name cursor
ji->m_name = NAME;
ji->m_nameLen = NAMELEN;
// get length decoded
int32_t curr = m_sb.length();
// store decoded string right after jsonitem
if ( !m_sb.safeDecodeJSONToUtf8 (str,slen,
niceness ))
return NULL;
// store length decoded json
ji->m_valueLen = m_sb.length() - curr;
// end with a \0
m_sb.pushChar('\0');
// ok, this one is done
ji = NULL;
}
else {
log("json: fieldless name in json");
g_errno = EBADJSONPARSER;
return NULL;
}
// skip over the string
size = 0;
p = x;
continue;
}
// true or false?
if ( (*p == 't' && strncmp(p,"true",4)==0) ||
(*p == 'f' && strncmp(p,"false",5)==0) ) {
// make a new one
ji = addNewItem();
if ( ! ji ) return NULL;
// copy the number as a string as well
int32_t curr = m_sb.length();
// what is the length of it?
int32_t slen = 4;
ji->m_valueLong = 1;
ji->m_valueDouble = 1.0;
if ( *p == 'f' ) {
slen = 5;
ji->m_valueLong = 0;
ji->m_valueDouble = 0;
}
// store decoded string right after jsonitem
if ( !m_sb.safeDecodeJSONToUtf8 (p,slen,niceness))
return NULL;
// store length decoded json
ji->m_valueLen = m_sb.length() - curr;
// end with a \0
m_sb.pushChar('\0');
ji->m_type = JT_NUMBER;
// use name cursor
ji->m_name = NAME;
ji->m_nameLen = NAMELEN;
ji = NULL;
// skip over the string
size = 1;
//p = end;
continue;
}
// if we hit a digit they might not be in quotes like
// "crawled":123
if ( is_digit ( *p ) ||
// like .123 ?
( *p == '.' && is_digit(p[1]) ) ) {
// find end of the number
char *end = p + 1;
// . allow '.' for decimal numbers
// . TODO: allow E for exponent
for ( ; *end && (is_digit(*end) || *end=='.');end++) ;
// define the string
char *str = p;
int32_t slen = end - str;
// make a new one
ji = addNewItem();
if ( ! ji ) return NULL;
// back up over negative sign?
if ( str > json && str[-1] == '-' ) str--;
// decode
//char c = str[slen];
//str[slen] = '\0';
ji->m_valueLong = atol(str);
ji->m_valueDouble = atof(str);
// copy the number as a string as well
int32_t curr = m_sb.length();
// store decoded string right after jsonitem
if ( !m_sb.safeDecodeJSONToUtf8 ( str, slen,niceness))
return NULL;
// store length decoded json
ji->m_valueLen = m_sb.length() - curr;
// end with a \0
m_sb.pushChar('\0');
//str[slen] = c;
ji->m_type = JT_NUMBER;
// use name cursor
ji->m_name = NAME;
ji->m_nameLen = NAMELEN;
ji = NULL;
// skip over the string
size = 0;
p = end;
continue;
}
}
// for testing if we realloc
char *memEnd = m_sb.getBufStart();
if ( mem != memEnd ) { char *xx=NULL;*xx=0; }
return (JsonItem *)m_sb.getBufStart();
}
// *next is set to ptr into m_cgiBuf so that the next successive call to
// getString with the SAME "field" will start at *next. that way you
// can use the same cgi parameter multiple times. (like strstr kind of)
char *HttpRequest::getStringFromCookie ( char *field ,
long *len ,
char *defaultStr ,
long *next ) {
// get field len
long flen = gbstrlen(field);
// assume none
if ( len ) *len = 0;
// if no cookie, forget it
if ( ! m_cookiePtr ) return defaultStr;
// the end of the cookie
//char *pend = m_cookieBuf + m_cookieBufLen;
char *pend = m_cookiePtr + m_cookieLen;
char *p = m_cookiePtr;
// skip over spaces and punct
for ( ; p && p < pend ; p++ )
if ( is_alnum_a(*p) ) break;
// skip "Cookie:"
if ( p + 7 < pend && ! strncasecmp(p,"cookie:",7) ) p += 7;
// skip spaces after that
for ( ; p && p < pend ; p++ )
if ( is_alnum_a(*p) ) break;
// crazy?
if ( p >= pend ) return defaultStr;
char *savedVal = NULL;
// so we do not skip the first cookie, jump right in!
// otherwise we lose the calendar cookie for msie
goto entryPoint;
// . loop over all xxx=yyy\0 thingies in the cookie
// . we converted every '&' to a \0 when the cookiebuf was set above
//for ( char *p = m_cookieBuf ; *p ; p += gbstrlen(p) + 1 ) {
// . no, we just keep them as &'s because seems like cookies use ;'s
// as delimeters not so much &'s. and when we log the cookie in the
// log, i wanted to see the whole cookie, so having \0's in the
// cookie was messing that up.
for ( ; p < pend ; p++ ) {
// need a \0
// fixes "display=0&map=0&calendar=0;" that is only one cookie.
// so do not grap value of map or calendar from that!!
if ( *p ) continue;
// back to back \0's? be careful how we skip over them!
if ( ! p[1] ) continue;
// skip that
if ( ++p >= pend ) break;
// skip whitespace that follows
for ( ; p < pend ; p++ )
if ( ! is_wspace_a(*p) ) break;
// end of cookie?
if ( p >= pend ) break;
entryPoint:
// check first char
if ( *p != *field ) continue;
// does it match? continue if not a match
if ( strncmp ( p , field , flen ) ) continue;
// point to value
char *val = p + flen;
// must be an equal sign
if ( *val != '=' ) continue;
// skip that sign
val++;
// . cookies terminate fields by space or ; or &
// . skip to end of cookie value for this field
char *e = val;
// skip over alnum. might also be \0 if this function
// was already called somewhere else!
// we NULL separated each cookie and then urldecoded each
// cookie above in the m_cookieBuf logic. cookies can contain
// encoded ;'s and &'s so i took this checks out of this while
// loop. like the widgetHeader has semicolons in it and it
// stores in the cookie.
while ( e < pend && *e ) e++;
// that is the length
if ( len ) *len = e - val;
// NULL terminate it, we should have already logged the cookie
// so it should be ok to NULL terminate now. we already
// call urlDecode() now above... and make the &'s into \0's
*e = '\0';
// if we were in the meta cookie, return that...
// otherwise if you visited this site before metacookies
// were used you might have the cookie outside the meta
// cookie AND inside the metacookie, and only the value
// inside the metacookie is legit...
if ( val > m_metaCookie ) return val;
// otherwise, save it and try to get from meta cookie
savedVal = val;
// length
//if ( len ) *len = gbstrlen(val);
// this is the value!
//return val;
}
// did we save something?
if ( savedVal ) return savedVal;
// no match
return defaultStr;
}
// . return the value of the specified "field" within this html tag, "s"
// . the case of "field" does not matter
char *getFieldValue ( char *s ,
long slen ,
char *field ,
long *valueLen ) {
// reset this to 0
*valueLen = 0;
// scan for the field name in our node
long flen = gbstrlen(field);
char inQuotes = '\0';
long i;
// make it sane
if ( slen > 2000 ) slen = 2000;
for ( i = 1; i + flen < slen ; i++ ) {
// skip the field if it's quoted
if ( inQuotes) {
if (s[i] == inQuotes ) inQuotes = 0;
continue;
}
// set inQuotes to the quote if we're in quotes
if ( (s[i]=='\"' || s[i]=='\'')){
inQuotes = s[i];
continue;
}
// if not in quote tag might end
if ( s[i] == '>' && ! inQuotes ) return NULL;
// a field name must be preceeded by non-alnum
if ( is_alnum_a ( s[i-1] ) ) continue;
// the first character of this field shout match field[0]
if ( to_lower_a (s[i]) != to_lower_a(field[0] )) continue;
// field just be immediately followed by an = or space
if (s[i+flen]!='='&&!is_wspace_a(s[i+flen]))continue;
// field names must match
if ( strncasecmp ( &s[i], field, flen ) != 0 ) continue;
// break cuz we got a match for our field name
break;
}
// return NULL if no matching field
if ( i + flen >= slen ) return NULL;
// advance i over the fieldname so it pts to = or space
i += flen;
// advance i over spaces
while ( i < slen && is_wspace_a ( s[i] ) ) i++;
// advance over the equal sign, return NULL if does not exist
if ( i < slen && s[i++] != '=' ) return NULL;
// advance i over spaces after the equal sign
while ( i < slen && is_wspace_a ( s[i] ) ) i++;
// now parse out the value of this field (could be in quotes)
inQuotes = '\0';
// set inQuotes to the quote if we're in quotes
if ( s[i]=='\"' || s[i]=='\'') inQuotes = s[i++];
// mark this as the start of the value
int start=i;
// advance i until we hit a space, or we hit a that quote if inQuotes
if (inQuotes) while (i<slen && s[i] != inQuotes ) i++;
else while ( i<slen &&!is_wspace_a(s[i])&&s[i]!='>')i++;
// set the length of the value
*valueLen = i - start;
// return a ptr to the value
return s + start;
}
// . so now this adds a list of Synonyms to the m_pools[] and returns a ptr
// to the first one.
// . then the parent caller can store that ptr in the m_wordToSyn[] array
// which we pre-alloc upon calling the set() function based on the # of
// words we got
// . returns # of synonyms stored into "tmpBuf"
long Synonyms::getSynonyms ( Words *words ,
long wordNum ,
uint8_t langId ,
char *tmpBuf ,
long niceness ) {
// punct words have no synoyms
if ( ! words->m_wordIds[wordNum] ) return 0;
// store these
m_words = words;
m_docLangId = langId;
m_niceness = niceness;
// sanity check
if ( wordNum > m_words->m_numWords ) { char *xx=NULL;*xx=0; }
// init the dedup table to dedup wordIds
HashTableX dt;
char dbuf[512];
dt.set(8,0,12,dbuf,512,false,m_niceness,"altwrds");
long maxSyns = (long)MAX_SYNS;
char *bufPtr = tmpBuf;
// point into buffer
m_aids = (long long *)bufPtr;
bufPtr += maxSyns * 8;
// then the word ids
m_wids0 = (long long *)bufPtr;
bufPtr += maxSyns * 8;
// second word ids, for multi alnum word synonyms, i.e. "New Jersey"
m_wids1 = (long long *)bufPtr;
bufPtr += maxSyns * 8;
m_termPtrs = (char **)bufPtr;
bufPtr += maxSyns * 4;
m_termLens = (long *)bufPtr;
bufPtr += maxSyns * 4;
m_numAlnumWords = (long *)bufPtr;
bufPtr += maxSyns * 4;
m_numAlnumWordsInBase = (long *)bufPtr;
bufPtr += maxSyns * 4;
// source
m_src = bufPtr;
bufPtr += maxSyns;
// cursors
m_aidsPtr = m_aids;
m_wids0Ptr = m_wids0;
m_wids1Ptr = m_wids1;
m_srcPtr = m_src;
m_termPtrsPtr = m_termPtrs;
m_termLensPtr = m_termLens;
m_numAlnumWordsPtr = m_numAlnumWords;
m_numAlnumWordsInBasePtr = m_numAlnumWordsInBase;
char *w = m_words->m_words [wordNum];
long wlen = m_words->m_wordLens[wordNum];
//
// NOW hit wiktionary
// Trust this less then our s_exceptions above, but more than
// our morph computations below
//
char sourceId = SOURCE_WIKTIONARY;
char *ss = NULL;
long long bwid;
char wikiLangId = m_docLangId;
bool hadSpace ;
long klen ;
long baseNumAlnumWords;
tryOtherLang:
/*
// if word only exists in one language, assume that language for word
// even if m_docLangId is langUnknown (0)
if ( ! ss &&
! m_docLangId &&
! wikiLangId ) {
// get raw word id
bwid = m_words->m_wordIds[wordNum];
// each lang has its own bit
long long bits = g_speller.getLangBits64 ( &bwid );
// skip if not unique
char count = getNumBitsOn64 ( bits ) ;
// if we only got one lang we could be, assume that
if ( count == 1 )
// get it. bit #0 is english, so add 1
wikiLangId = getBitPosLL((uint8_t *)&bits) + 1;
// try setting based on script. greek. russian. etc.
// if the word was not in the wiktionary.
// this will be langUnknown if not definitive.
else
wikiLangId = getCharacterLanguage(w);
}
*/
// try looking up bigram so "new jersey" gets "nj" as synonym
if ( wikiLangId &&
wordNum+2< m_words->m_numWords &&
m_words->m_wordIds[wordNum+2]) {
// get phrase id bigram then
long conti = 0;
bwid = hash64Lower_utf8_cont(w,wlen,0,&conti);
// then the next word
char *wp2 = m_words->m_words[wordNum+2];
long wlen2 = m_words->m_wordLens[wordNum+2];
bwid = hash64Lower_utf8_cont(wp2,wlen2,bwid,&conti);
baseNumAlnumWords = 2;
ss = g_wiktionary.getSynSet( bwid, wikiLangId );
}
// need a language for wiktionary to work with
if ( wikiLangId && ! ss ) {
// get raw word id
bwid = m_words->m_wordIds[wordNum];
baseNumAlnumWords = 1;
//if ( bwid == 1424622907102375150LL)
// log("a");
ss = g_wiktionary.getSynSet( bwid, wikiLangId );
// if that failed try removing 's from word if there
if ( ! ss &&
wlen >= 3 &&
w[wlen-2]=='\'' &&
w[wlen-1]=='s' ) {
long long cwid = hash64Lower_utf8(w,wlen-2);
ss = g_wiktionary.getSynSet( cwid, wikiLangId );
}
}
// even though a document may be in german it often has some
// english words "pdf download" "copyright" etc. so if the word
// has no synset in german, try it in english
if ( //numPresets == 0 &&
! ss &&
m_docLangId != langEnglish &&
wikiLangId != langEnglish &&
m_docLangId &&
g_speller.getSynsInEnglish(w,wlen,m_docLangId,langEnglish) ) {
// try english
wikiLangId = langEnglish;
sourceId = SOURCE_WIKTIONARY_EN;
goto tryOtherLang;
}
// if it was in wiktionary, just use that synset
if ( ss ) {
// prepare th
HashTableX dedup;
HashTableX *dd = NULL;
char dbuf[512];
long count = 0;
addSynSet:
// do we have another set following this
char *next = g_wiktionary.getNextSynSet(bwid,m_docLangId,ss);
// if so, init the dedup table then
if ( next && ! dd ) {
dd = &dedup;
dd->set ( 8,0,8,dbuf,512,false,m_niceness,"sddbuf");
}
// skip over the pipe i guess
char *pipe = ss + 2;
// zh_ch?
if ( *pipe == '_' ) pipe += 3;
// sanity
if ( *pipe != '|' ) { char *xx=NULL;*xx=0; }
// point to word list
char *p = pipe + 1;
// hash up the list of words, they are in utf8 and
char *e = p + 1;
// save count in case we need to undo
//long saved = m_numAlts[wordNum];
hashLoop:
// skip synonyms that are anagrams because its to ambiguous
// the are mappings like
// "PC" -> "PC,Personal Computer"
// "PC" -> "PC,Probable Cause" ... (lots more!)
//bool isAnagram = true;
for ( ; *e !='\n' && *e != ',' ; e++ ) ;
// if ( ! is_upper_a(*e) ) isAnagram = false;
// get it
long long h = hash64Lower_utf8_nospaces ( p , e - p );
// skip if same as base word
if ( h == bwid ) goto getNextSyn;
// should we check for dups?
if ( dd ) {
// skip dups
if ( dd->isInTable(&h) ) goto getNextSyn;
// dedup. return false with g_errno set on error
if ( ! dd->addKey(&h) ) return m_aidsPtr - m_aids;
}
// store it
*m_aidsPtr++ = h;
// store source
*m_srcPtr++ = sourceId;
hadSpace = false;
klen = e - p;
for ( long k = 0 ; k < klen ; k++ )
if ( is_wspace_a(p[k]) ) hadSpace = true;
*m_termPtrsPtr++ = p;
*m_termLensPtr++ = e-p;
// only for multi-word synonyms like "New Jersey"...
*m_wids0Ptr = 0LL;
*m_wids1Ptr = 0LL;
*m_numAlnumWordsPtr = 1;
// and for multi alnum word synonyms
if ( hadSpace ) {
Words sw;
sw.setx ( p , e - p , m_niceness );
*(long long *)m_wids0Ptr = sw.m_wordIds[0];
*(long long *)m_wids1Ptr = sw.m_wordIds[2];
*(long *)m_numAlnumWordsPtr = sw.getNumAlnumWords();
}
m_wids0Ptr++;
m_wids1Ptr++;
m_numAlnumWordsPtr++;
// how many words did we have to hash to find a synset?
// i.e. "new jersey" would be 2, to get "nj"
*m_numAlnumWordsInBasePtr++ = baseNumAlnumWords;
// do not breach
if ( ++count >= maxSyns ) goto done;
getNextSyn:
// loop for more
if ( *e == ',' ) { e++; p = e; goto hashLoop; }
// add in the next syn set, deduped
if ( next ) { ss = next; goto addSynSet; }
// wrap it up
done:
// all done
return m_aidsPtr - m_aids;
}
// strip marks from THIS word, return -1 w/ g_errno set on error
if ( ! addStripped ( w , wlen,&dt ) ) return m_aidsPtr - m_aids;
// returns false with g_errno set
if ( ! addAmpPhrase ( wordNum, &dt ) ) return m_aidsPtr - m_aids;
// if we end in apostrophe, strip and add
if ( wlen>= 3 &&
w[wlen-1] == 's' &&
w[wlen-2]=='\'' &&
! addWithoutApostrophe ( wordNum, &dt ) )
return m_aidsPtr - m_aids;
return m_aidsPtr - m_aids;
}
// . sets m_qbuf1[] and m_qbuf2[]
// . m_qbuf1[] is the advanced query
// . m_qbuf2[] is the query to be used for spell checking
// . returns false and set g_errno on error
bool SearchInput::setQueryBuffers ( ) {
m_sbuf1.reset();
m_sbuf2.reset();
m_sbuf3.reset();
short qcs = csUTF8;
if (m_queryCharset && m_queryCharsetLen){
// we need to convert the query string to utf-8
qcs = get_iana_charset(m_queryCharset, m_queryCharsetLen);
if (qcs == csUnknown) {
//g_errno = EBADCHARSET;
//g_msg = "(error: unknown query charset)";
//return false;
qcs = csUTF8;
}
}
// prepend sites terms
long numSites = 0;
char *csStr = NULL;
numSites = 0;
csStr = get_charset_str(qcs);
if ( m_sites && m_sites[0] ) {
char *s = m_sites;
char *t;
long len;
m_sbuf1.pushChar('(');//*p++ = '(';
loop:
// skip white space
while ( *s && ! is_alnum_a(*s) ) s++;
// bail if done
if ( ! *s ) goto done;
// get length of it
t = s;
while ( *t && ! is_wspace_a(*t) ) t++;
len = t - s;
// add site: term
//if ( p + 12 + len >= pend ) goto toobig;
if ( numSites > 0 ) m_sbuf1.safeStrcpy ( " UOR " );
m_sbuf1.safeStrcpy ( "site:" );
//p += ucToUtf8(p, pend-p,s, len, csStr, 0,0);
m_sbuf1.safeMemcpy ( s , len );
//memcpy ( p , s , len ); p += len;
//*p++ = ' ';
m_sbuf1.pushChar(' ');
s = t;
numSites++;
goto loop;
done:
m_sbuf1.safePrintf(") | ");
// inc totalLen
m_sitesQueryLen = m_sitesLen + (numSites * 10);
}
// append site: term
if ( m_siteLen > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//memcpy ( p , "+site:" , 6 ); p += 6;
m_sbuf1.safePrintf("+site:");
//memcpy ( p , m_site , m_siteLen ); p += m_siteLen;
m_sbuf1.safeMemcpy(m_site,m_siteLen);
}
// append gblang: term
if( m_gblang > 0 ) {
//if( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//p += sprintf( p, "+gblang:%li |", m_gblang );
m_sbuf1.safePrintf( "+gblang:%li |", m_gblang );
}
// bookmark here so we can copy into st->m_displayQuery below
//long displayQueryOffset = m_sbuf1.length();
// append url: term
if ( m_urlLen > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//memcpy ( p , "+url:" , 5 ); p += 5;
m_sbuf1.safeStrcpy ( "+url:");
//memcpy ( p , m_url , m_urlLen ); p += m_urlLen;
m_sbuf1.safeMemcpy ( m_url , m_urlLen );
}
// append url: term
if ( m_linkLen > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//memcpy ( p , "+link:" , 6 ); p += 6;
m_sbuf1.safeStrcpy ( "+link:");
//memcpy ( p , m_link , m_linkLen ); p += m_linkLen;
m_sbuf1.safeMemcpy ( m_link , m_linkLen );
}
// append the natural query
if ( m_queryLen > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//p += ucToUtf8(p, pend-p, m_query, m_queryLen, csStr, 0,0);
m_sbuf1.safeMemcpy ( m_query , m_queryLen );
//memcpy ( p , m_query , m_queryLen ); p += m_queryLen;
// add to spell checked buf, too
//if ( p2 > pstart2 ) *p2++ = ' ';
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
//p2 +=ucToUtf8(p2, pend2-p2, m_query, m_queryLen, csStr, 0,0);
m_sbuf2.safeMemcpy ( m_query , m_queryLen );
//memcpy ( p2 , m_query , m_queryLen ); p2 += m_queryLen;
}
if ( m_query2Len > 0 ) {
//if ( p3 > pstart3 ) *p3++ = ' ';
if ( m_sbuf3.length() ) m_sbuf3.pushChar(' ');
//p3+=ucToUtf8(p3, pend3-p3, m_query2, m_query2Len, csStr,0,0);
m_sbuf3.safeMemcpy ( m_query2 , m_query2Len );
}
//if (g_errno == EILSEQ){ // illegal character seq
// log("query: bad char set");
// g_errno = 0;
// if (qcs == csUTF8) {qcs = csISOLatin1;goto doOver;}
// if (qcs != csISOLatin1) {qcs = csUTF8;goto doOver;}
//}
// append quoted phrases to query
if ( m_quoteLen1 > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//*p++ = '+';
//*p++ = '\"';
m_sbuf1.safeStrcpy("+\"");
//p += ucToUtf8(p, pend-p, m_quote1, m_quoteLen1, csStr, 0,0);
m_sbuf1.safeMemcpy ( m_quote1 , m_quoteLen1 );
//memcpy ( p , m_quote1 , m_quoteLen1 ); p += m_quoteLen1 ;
//*p++ = '\"';
m_sbuf1.safeStrcpy("\"");
// add to spell checked buf, too
//if ( p2 > pstart2 ) *p2++ = ' ';
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
//*p2++ = '+';
//*p2++ = '\"';
m_sbuf2.safeStrcpy("+\"");
//p2+=ucToUtf8(p2, pend2-p2, m_quote1, m_quoteLen1, csStr,0,0);
m_sbuf2.safeMemcpy ( m_quote1 , m_quoteLen1 );
//memcpy ( p2 , m_quote1 , m_quoteLen1 ); p2 += m_quoteLen1 ;
//*p2++ = '\"';
m_sbuf2.safeStrcpy("\"");
}
//if (g_errno == EILSEQ){ // illegal character seq
// g_errno = 0;
// if (qcs == csUTF8) {qcs = csISOLatin1;goto doOver;}
// if (qcs != csISOLatin1) {qcs = csUTF8;goto doOver;}
//}
if ( m_quoteLen2 > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//*p++ = '+';
//*p++ = '\"';
m_sbuf1.safeStrcpy("+\"");
//p += ucToUtf8(p, pend-p, m_quote2, m_quoteLen2, csStr, 0,0);
m_sbuf1.safeMemcpy ( m_quote2 , m_quoteLen2 );
//memcpy ( p , m_quote2 , m_quoteLen2 ); p += m_quoteLen2 ;
//*p++ = '\"';
m_sbuf1.safeStrcpy("\"");
// add to spell checked buf, too
//if ( p2 > pstart2 ) *p2++ = ' ';
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
//*p2++ = '+';
//*p2++ = '\"';
m_sbuf2.safeStrcpy("+\"");
//p2+=ucToUtf8(p2, pend2-p2, m_quote2, m_quoteLen2, csStr,0,0);
m_sbuf2.safeMemcpy ( m_quote2 , m_quoteLen2 );
//memcpy ( p2 , m_quote2 , m_quoteLen2 ); p2 += m_quoteLen2 ;
//*p2++ = '\"';
m_sbuf2.safeStrcpy("\"");
}
//if (g_errno == EILSEQ){ // illegal character seq
// g_errno = 0;
// if (qcs == csUTF8) {qcs = csISOLatin1;goto doOver;}
// if (qcs != csISOLatin1) {qcs = csUTF8;goto doOver;}
//}
// append plus terms
if ( m_plusLen > 0 ) {
char *s = m_plus, *send = m_plus + m_plusLen;
//if ( p > pstart && p < pend ) *p++ = ' ';
//if ( p2 > pstart2 && p2 < pend2) *p2++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
while (s < send) {
while (isspace(*s) && s < send) s++;
char *s2 = s+1;
if (*s == '\"') {
// if there's no closing quote just treat
// the end of the line as such
while (*s2 != '\"' && s2 < send) s2++;
if (s2 < send) s2++;
} else {
while (!isspace(*s2) && s2 < send) s2++;
}
if (s < send) break;
//if (p < pend) *p++ = '+';
//if (p2 < pend2) *p2++ = '+';
m_sbuf1.pushChar('+');
m_sbuf2.pushChar('+');
//p += ucToUtf8(p, pend-p, s, s2-s, csStr, 0,0);
//p2 += ucToUtf8(p2, pend2-p2, s, s2-s, csStr, 0,0);
m_sbuf1.safeMemcpy ( s , s2 - s );
m_sbuf2.safeMemcpy ( s , s2 - s );
/*
if (g_errno == EILSEQ) { // illegal character seq
g_errno = 0;
if (qcs == csUTF8) {
qcs = csISOLatin1;
goto doOver;
}
if (qcs != csISOLatin1) {
qcs = csUTF8;
goto doOver;
}
}
*/
s = s2 + 1;
if (s < send) {
//if (p < pend) *p++ = ' ';
//if (p2 < pend2) *p2++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
}
}
}
// append minus terms
if ( m_minusLen > 0 ) {
char *s = m_minus, *send = m_minus + m_minusLen;
//if ( p > pstart && p < pend ) *p++ = ' ';
//if ( p2 > pstart2 && p2 < pend2) *p2++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
while (s < send) {
while (isspace(*s) && s < send) s++;
char *s2 = s+1;
if (*s == '\"') {
// if there's no closing quote just treat
// the end of the line as such
while (*s2 != '\"' && s2 < send) s2++;
if (s2 < send) s2++;
} else {
while (!isspace(*s2) && s2 < send) s2++;
}
if (s < send) break;
//if (p < pend) *p++ = '-';
//if (p2 < pend2) *p2++ = '-';
m_sbuf1.pushChar('-');
m_sbuf2.pushChar('-');
//p += ucToUtf8(p, pend-p, s, s2-s, csStr, 0,0);
//p2 += ucToUtf8(p2, pend2-p2, s, s2-s, csStr, 0,0);
m_sbuf1.safeMemcpy ( s , s2 - s );
m_sbuf2.safeMemcpy ( s , s2 - s );
/*
if (g_errno == EILSEQ) { // illegal character seq
g_errno = 0;
if (qcs == csUTF8) {
qcs = csISOLatin1;
goto doOver;
}
if (qcs != csISOLatin1) {
qcs = csUTF8;
goto doOver;
}
}
*/
s = s2 + 1;
if (s < send) {
//if (p < pend) *p++ = ' ';
//if (p2 < pend2) *p2++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
if ( m_sbuf2.length() ) m_sbuf2.pushChar(' ');
}
}
}
// append gbkeyword:numinlinks if they have &mininlinks=X, X>0
long minInlinks = m_hr->getLong("mininlinks",0);
if ( minInlinks > 0 ) {
//if ( p > pstart ) *p++ = ' ';
if ( m_sbuf1.length() ) m_sbuf1.pushChar(' ');
//char *str = "gbkeyword:numinlinks";
//long len = gbstrlen(str);
//memcpy ( p , str , len );
//p += len;
m_sbuf1.safePrintf ( "gbkeyword:numinlinks");
}
// null terms
m_sbuf1.pushChar('\0');
m_sbuf2.pushChar('\0');
m_sbuf3.pushChar('\0');
// the natural query
m_displayQuery = m_sbuf2.getBufStart();// + displayQueryOffset;
if ( ! m_displayQuery ) m_displayQuery = "";
while ( *m_displayQuery == ' ' ) m_displayQuery++;
m_displayQueryLen = gbstrlen(m_displayQuery);//p-m_displayQuery
//log("query: got query %s",m_sbuf1.getBufStart());
//log("query: got display query %s",m_displayQuery);
// urlencoded display query
urlEncode(m_qe,
MAX_QUERY_LEN*2,
m_displayQuery,
m_displayQueryLen);
return true;
}
// . when the Conf::m_proxyIps parm is updated we call this to rebuild
// s_iptab, our table of SpiderProxy instances, which has the proxies and
// their performance statistics.
// . we try to maintain stats of ip/ports that did NOT change when rebuilding.
bool buildProxyTable ( ) {
// scan the NEW list of proxy ip/port pairs in g_conf
char *p = g_conf.m_proxyIps.getBufStart();
HashTableX tmptab;
tmptab.set(8,0,16,NULL,0,false,"tmptab");
// scan the user inputted space-separated list of ip:ports
// (optional username:password@ip:port)
for ( ; *p ; ) {
// skip white space
if ( is_wspace_a(*p) ) { p++; continue; }
// skip http://
if ( strncasecmp(p,"http://",7) == 0 ) { p += 7; continue; }
// scan in an ip:port
char *s = p; char *portStr = NULL;
int32_t dc = 0, pc = 0, gc = 0, bc = 0;
const char *msg;
char *usernamePwd = NULL;
int32_t usernamePwdLen = 0;
char *ipStart = p;
// scan all characters until we hit \0 or another whitespace
for ( ; *s && !is_wspace_a(*s); s++) {
if ( *s == '@' ) {
// must be username:pwd
if ( pc != 1 ) {
msg = "bad username:password";
goto hadError;
}
usernamePwd = p;
usernamePwdLen = s - p;
if ( usernamePwdLen >= MAXUSERNAMEPWD-2 ) {
msg = "username:password too long";
goto hadError;
}
dc = 0;
gc = 0;
bc = 0;
pc = 0;
portStr = NULL;
ipStart = s+1;
continue;
}
if ( *s == '.' ) { dc++; continue; }
if ( *s == ':' ) { portStr=s; pc++; continue; }
if ( is_digit(*s) ) { gc++; continue; }
bc++;
continue;
}
// ensure it is a legit ip:port combo
msg = NULL;
if ( gc < 4 )
msg = "not enough digits for an ip";
if ( pc > 1 )
msg = "too many colons";
if ( dc != 3 )
msg = "need 3 dots for an ip address";
if ( bc )
msg = "got illegal char in ip:port listing";
if ( msg ) {
hadError:
char c = *s;
*s = '\0';
log("buf: %s for %s",msg,p);
*s = c;
return false;
}
// convert it
int32_t iplen = s - ipStart;
if ( portStr ) iplen = portStr - ipStart;
int32_t ip = atoip(ipStart,iplen);
// another sanity check
if ( ip == 0 || ip == -1 ) {
log("spider: got bad proxy ip for %s",p);
return false;
}
// and the port default is 80
int32_t port = 80;
if ( portStr ) port = atol2(portStr+1,s-portStr-1);
if ( port < 0 || port > 65535 ) {
log("spider: got bad proxy port for %s",p);
return false;
}
// . we got a legit ip:port
// . see if already in our table
uint64_t ipKey = (uint32_t)ip;
ipKey <<= 16;
ipKey |= (uint16_t)(port & 0xffff);
// also store into tmptable to see what we need to remove
tmptab.addKey(&ipKey);
// see if in table
int32_t islot = s_iptab.getSlot( &ipKey);
// advance p
p = s;
// if in there, keep it as is
if ( islot >= 0 ) continue;
// otherwise add new entry
SpiderProxy newThing;
memset ( &newThing , 0 , sizeof(SpiderProxy));
newThing.m_ip = ip;
newThing.m_port = port;
newThing.m_lastDownloadTookMS = -1;
newThing.m_lastSuccessfulTestMS = -1;
gbmemcpy(newThing.m_usernamePwd,usernamePwd,usernamePwdLen);
// ensure it is NULL terminated
newThing.m_usernamePwd[usernamePwdLen] = '\0';
if ( ! s_iptab.addKey ( &ipKey, &newThing ) )
return false;
}
redo:
int32_t removed = 0;
// scan all SpiderProxies in tmptab
for ( int32_t i = 0 ; i < s_iptab.getNumSlots() ; i++ ) {
// skip empty buckets in hashtable s_iptab
if ( ! s_iptab.m_flags[i] ) continue;
// get the key
int64_t key = *(int64_t *)s_iptab.getKeyFromSlot(i);
// must also exist in tmptab, otherwise it got removed by user
if ( tmptab.isInTable ( &key ) ) continue;
// skip if not in table
if ( s_iptab.getSlot ( &key ) < 0 ) {
log("sproxy: iptable hashing messed up");
continue;
}
// shoot, it got removed. not in the new list of ip:ports
s_iptab.removeKey ( &key );
removed++;
// hashtable is messed up now, start over
//goto redo;
}
if ( removed ) goto redo;
return true;
}
void HttpMime::addCookie(const httpcookie_t &cookie, const Url ¤tUrl, SafeBuf *cookieJar) {
// don't add expired cookie into cookie jar
if (cookie.m_expired) {
return;
}
if (cookie.m_domain) {
cookieJar->safeMemcpy(cookie.m_domain, cookie.m_domainLen);
cookieJar->pushChar('\t');
cookieJar->safeStrcpy(cookie.m_defaultDomain ? "FALSE\t" : "TRUE\t");
} else {
cookieJar->safeMemcpy(currentUrl.getHost(), currentUrl.getHostLen());
cookieJar->pushChar('\t');
cookieJar->safeStrcpy("FALSE\t");
}
if (cookie.m_path) {
cookieJar->safeMemcpy(cookie.m_path, cookie.m_pathLen);
cookieJar->pushChar('\t');
} else {
if (currentUrl.getPathLen()) {
cookieJar->safeMemcpy(currentUrl.getPath(), currentUrl.getPathLen());
} else {
cookieJar->pushChar('/');
}
cookieJar->pushChar('\t');
}
if (cookie.m_secure) {
cookieJar->safeStrcpy("TRUE\t");
} else {
cookieJar->safeStrcpy("FALSE\t");
}
// we're not using expiration field
cookieJar->safeStrcpy("0\t");
int32_t currentLen = cookieJar->length();
cookieJar->safeMemcpy(cookie.m_cookie, cookie.m_cookieLen);
// cater for multiline cookie
const char *currentPos = cookieJar->getBufStart() + currentLen;
const char *delPosStart = NULL;
int32_t delLength = 0;
while (currentPos < cookieJar->getBufPtr() - 1) {
if (delPosStart) {
if (is_wspace_a(*currentPos) || *currentPos == '\n' || *currentPos == '\r') {
++delLength;
} else {
break;
}
} else {
if (*currentPos == '\n' || *currentPos == '\r') {
delPosStart = currentPos;
++delLength;
}
}
++currentPos;
}
cookieJar->removeChunk1(delPosStart, delLength);
/// @todo ALC handle httpOnly attribute
cookieJar->pushChar('\n');
}
