bool HashTableT<Key_t, Val_t>::serialize(SafeBuf& sb) {
sb += m_numSlots;
sb += m_numSlotsUsed;
if(m_numSlots == 0) return true;
bool x = true;
x &= sb.safeMemcpy((char*)m_keys, sizeof(Key_t) * m_numSlots);
x &= sb.safeMemcpy((char*)m_vals, sizeof(Val_t) * m_numSlots);
return x;
}
void doneInjectingLinksWrapper ( void *state ) {
Msg7 *msg7 = (Msg7 *)state;
SafeBuf *sb = &msg7->m_sb;
// copy the serps into ou rbuf
if ( ! g_errno ) {
// print header
if ( sb->length() == 0 ) {
// print header of page
sb->safePrintf("<?xml version=\"1.0\" "
"encoding=\"UTF-8\" ?>\n"
"<response>\n" );
}
// serp header
if ( msg7->m_round == 1 )
sb->safePrintf("\t<googleResults>\n");
else
sb->safePrintf("\t<bingResults>\n");
// print results
sb->safeMemcpy(&msg7->m_xd.m_serpBuf);
// end that
if ( msg7->m_round == 1 )
sb->safePrintf("\t</googleResults>\n");
else
sb->safePrintf("\t</bingResults>\n");
}
// do bing now
if ( msg7->m_round == 1 ) {
// return if it blocks
if ( ! msg7->scrapeQuery() ) return;
}
// otherwise, parse out the search results so steve can display them
if ( g_errno )
sb->safePrintf("<error><![CDATA[%s]]></error>\n",
mstrerror(g_errno));
// print header of page
sb->safePrintf("</response>\n");
// page is not more than 32k
//char buf[1024*32];
//char *p = buf;
// return docid and hostid
//p += sprintf ( p , "scraping status ");
// print error msg out, too or "Success"
//p += sprintf ( p , "%s", mstrerror(g_errno));
TcpSocket *sock = msg7->m_socket;
g_httpServer.sendDynamicPage ( sock,
sb->getBufStart(),
sb->length(),
-1/*cachetime*/);
// hopefully sb buffer is copied becaues this will free it:
mdelete ( msg7, sizeof(Msg7) , "PageInject" );
delete (msg7);
}
bool getWords() {
FILE *fd = fopen ( "/usr/share/dict/words" , "r" );
if ( ! fd ) {
log("blaster:: failed to open /usr/share/dict/words %s",
mstrerror(errno));
return 1;
}
char tmp[1024];
while ( fgets ( tmp , 1024 , fd ) ) {
long len = gbstrlen(tmp);
if(len > 2 && tmp[len-2] == 's' && tmp[len-3] == '\'') continue;
s_windices += s_words.length();
s_words.safeMemcpy(tmp, len-1); //copy in data minus the newline
s_words += '\0';
}
fclose ( fd );
log("blaster: read %li words, %li bytes in from dictionary.",
s_windices.length() / sizeof(long), s_words.length());
return true;
}
// returns false if blocked, true otherwise
bool processLoop ( void *state ) {
// get it
State2 *st = (State2 *)state;
// get the tcp socket from the state
TcpSocket *s = st->m_socket;
// get it
XmlDoc *xd = &st->m_xd;
if ( ! xd->m_loaded ) {
// setting just the docid. niceness is 0.
//xd->set3 ( st->m_docId , st->m_coll , 0 );
// callback
xd->setCallback ( state , processLoop );
// . and tell it to load from the old title rec
// . this sets xd->m_oldTitleRec/m_oldTitleRecSize
// . this sets xd->ptr_* and all other member vars from
// the old title rec if found in titledb.
if ( ! xd->loadFromOldTitleRec ( ) ) return false;
}
if ( g_errno ) return sendErrorReply ( st , g_errno );
// now force it to load old title rec
//char **tr = xd->getTitleRec();
SafeBuf *tr = xd->getTitleRecBuf();
// blocked? return false if so. it will call processLoop() when it rets
if ( tr == (void *)-1 ) return false;
// we did not block. check for error? this will free "st" too.
if ( ! tr ) return sendErrorReply ( st , g_errno );
// if title rec was empty, that is a problem
if ( xd->m_titleRecBuf.length() == 0 )
return sendErrorReply ( st , ENOTFOUND);
// set callback
char *na = xd->getIsNoArchive();
// wait if blocked
if ( na == (void *)-1 ) return false;
// error?
if ( ! na ) return sendErrorReply ( st , g_errno );
// forbidden? allow turkeys through though...
if ( ! st->m_isAdmin && *na )
return sendErrorReply ( st , ENOCACHE );
SafeBuf *sb = &st->m_sb;
// &page=4 will print rainbow sections
if ( ! st->m_printed && st->m_r.getLong("page",0) ) {
// do not repeat this call
st->m_printed = true;
// this will call us again since we called
// xd->setCallback() above to us
if ( ! xd->printDocForProCog ( sb , &st->m_r ) )
return false;
}
char *contentType = "text/html";
char format = st->m_format;
if ( format == FORMAT_XML ) contentType = "text/xml";
if ( format == FORMAT_JSON ) contentType = "application/json";
// if we printed a special page (like rainbow sections) then return now
if ( st->m_printed ) {
bool status = g_httpServer.sendDynamicPage (s,
//buf,bufLen,
sb->getBufStart(),
sb->getLength(),
-1,false,
//"text/html",
contentType,
-1, NULL, "utf8" );
// nuke state2
mdelete ( st , sizeof(State2) , "PageGet1" );
delete (st);
return status;
}
/*
// this was calling XmlDoc and setting sections, etc. to
// get the SpiderReply junk... no no no
// is it banned or filtered? this ignores the TagRec in the titleRec
// and uses msg8a to get it fresh instead
char *vi = xd->getIsFiltered();//Visible( );
// wait if blocked
if ( vi == (void *)-1 ) return false;
// error?
if ( ! vi ) return sendErrorReply ( st , g_errno );
// banned?
if ( ! st->m_isAdmin && ! *vi ) return sendErrorReply (st,EDOCBANNED);
*/
// get the utf8 content
char **utf8 = xd->getUtf8Content();
//long len = xd->size_utf8Content - 1;
// wait if blocked???
if ( utf8 == (void *)-1 ) return false;
// strange
if ( xd->size_utf8Content<=0) {
log("pageget: utf8 content <= 0");
return sendErrorReply(st,EBADENGINEER );
}
// alloc error?
if ( ! utf8 ) return sendErrorReply ( st , g_errno );
// get this host
Host *h = g_hostdb.getHost ( g_hostdb.m_hostId );
if ( ! h ) {
log("pageget: hostid %li is bad",g_hostdb.m_hostId);
return sendErrorReply(st,EBADENGINEER );
}
char *content = xd->ptr_utf8Content;
long contentLen = xd->size_utf8Content - 1;
// shortcut
char strip = st->m_strip;
// alloc buffer now
//char *buf = NULL;
//long bufMaxSize = 0;
//bufMaxSize = len + ( 32 * 1024 ) ;
//bufMaxSize = contentLen + ( 32 * 1024 ) ;
//buf = (char *)mmalloc ( bufMaxSize , "PageGet2" );
//char *p = buf;
//char *bufEnd = buf + bufMaxSize;
//if ( ! buf ) {
// return sendErrorReply ( st , g_errno );
//}
// for undoing the header
//char *start1 = p;
long startLen1 = sb->length();
// we are always utfu
if ( strip != 2 )
sb->safePrintf( "<meta http-equiv=\"Content-Type\" "
"content=\"text/html;charset=utf8\">\n");
// base href
//Url *base = &xd->m_firstUrl;
//if ( xd->ptr_redirUrl.m_url[0] )
// base = &xd->m_redirUrl;
char *base = xd->ptr_firstUrl;
if ( xd->ptr_redirUrl ) base = xd->ptr_redirUrl;
//Url *redir = *xd->getRedirUrl();
if ( strip != 2 ) {
sb->safePrintf ( "<BASE HREF=\"%s\">" , base );
//p += gbstrlen ( p );
}
// default colors in case css files missing
if ( strip != 2 ) {
sb->safePrintf( "\n<style type=\"text/css\">\n"
"body{background-color:white;color:black;}\n"
"</style>\n");
//p += gbstrlen ( p );
}
//char format = st->m_format;
if ( format == FORMAT_XML ) sb->reset();
if ( format == FORMAT_JSON ) sb->reset();
// for undoing the stuff below
long startLen2 = sb->length();//p;
// query should be NULL terminated
char *q = st->m_q;
long qlen = st->m_qlen;
char styleTitle[128] = "font-size:14px;font-weight:600;"
"color:#000000;";
char styleText[128] = "font-size:14px;font-weight:400;"
"color:#000000;";
char styleLink[128] = "font-size:14px;font-weight:400;"
"color:#0000ff;";
char styleTell[128] = "font-size:14px;font-weight:600;"
"color:#cc0000;";
// get the url of the title rec
Url *f = xd->getFirstUrl();
bool printDisclaimer = st->m_printDisclaimer;
if ( xd->m_contentType == CT_JSON )
printDisclaimer = false;
if ( format == FORMAT_XML ) printDisclaimer = false;
if ( format == FORMAT_JSON ) printDisclaimer = false;
char tbuf[100];
tbuf[0] = 0;
time_t lastSpiderDate = xd->m_spideredTime;
if ( printDisclaimer ||
format == FORMAT_XML ||
format == FORMAT_JSON ) {
struct tm *timeStruct = gmtime ( &lastSpiderDate );
strftime ( tbuf, 100,"%b %d, %Y UTC", timeStruct);
}
// We should always be displaying this disclaimer.
// - May eventually want to display this at a different location
// on the page, or on the click 'n' scroll browser page itself
// when this page is not being viewed solo.
// CNS: if ( ! st->m_clickNScroll ) {
if ( printDisclaimer ) {
sb->safePrintf(//sprintf ( p ,
//"<BASE HREF=\"%s\">"
//"<table border=1 width=100%%>"
//"<tr><td>"
"<table border=\"1\" bgcolor=\"#"
BGCOLOR
"\" cellpadding=\"10\" "
//"id=\"gbcnsdisctable\" class=\"gbcnsdisctable_v\""
"cellspacing=\"0\" width=\"100%%\" color=\"#ffffff\">"
"<tr"
//" id=\"gbcnsdisctr\" class=\"gbcnsdisctr_v\""
"><td>"
//"<font face=times,sans-serif color=black size=-1>"
"<span style=\"%s\">"
"This is Gigablast's cached page of </span>"
"<a href=\"%s\" style=\"%s\">%s</a>"
"" , styleTitle, f->getUrl(), styleLink,
f->getUrl() );
//p += gbstrlen ( p );
// then the rest
//sprintf(p ,
sb->safePrintf(
"<span style=\"%s\">. "
"Gigablast is not responsible for the content of "
"this page.</span>", styleTitle );
//p += gbstrlen ( p );
sb->safePrintf ( "<br/><span style=\"%s\">"
"Cached: </span>"
"<span style=\"%s\">",
styleTitle, styleText );
//p += gbstrlen ( p );
// then the spider date in GMT
// time_t lastSpiderDate = xd->m_spideredTime;
// struct tm *timeStruct = gmtime ( &lastSpiderDate );
// char tbuf[100];
// strftime ( tbuf, 100,"%b %d, %Y UTC", timeStruct);
//p += gbstrlen ( p );
sb->safeStrcpy(tbuf);
// Moved over from PageResults.cpp
sb->safePrintf( "</span> - <a href=\""
"/get?"
"q=%s&c=%s&rtq=%li&"
"d=%lli&strip=1\""
" style=\"%s\">"
"[stripped]</a>",
q , st->m_coll ,
(long)st->m_rtq,
st->m_docId, styleLink );
// a link to alexa
if ( f->getUrlLen() > 5 ) {
sb->safePrintf( " - <a href=\"http:"
"//web.archive.org/web/*/%s\""
" style=\"%s\">"
"[older copies]</a>" ,
f->getUrl(), styleLink );
}
if (st->m_noArchive){
sb->safePrintf( " - <span style=\"%s\"><b>"
"[NOARCHIVE]</b></span>",
styleTell );
}
if (st->m_isBanned){
sb->safePrintf(" - <span style=\"%s\"><b>"
"[BANNED]</b></span>",
styleTell );
}
// only print this if we got a query
if ( qlen > 0 ) {
sb->safePrintf("<br/><br/><span style=\"%s\"> "
"These search terms have been "
"highlighted: ",
styleText );
//p += gbstrlen ( p );
}
}
// how much space left in p?
//long avail = bufEnd - p;
// . make the url that we're outputting for (like in PageResults.cpp)
// . "thisUrl" is the baseUrl for click & scroll
char thisUrl[MAX_URL_LEN];
char *thisUrlEnd = thisUrl + MAX_URL_LEN;
char *x = thisUrl;
// . use the external ip of our gateway
// . construct the NAT mapped port
// . you should have used iptables to map port to the correct
// internal ip:port
//unsigned long ip =g_conf.m_mainExternalIp ; // h->m_externalIp;
//unsigned short port=g_conf.m_mainExternalPort;//h->m_externalHttpPort
// local check
//if ( st->m_isLocal ) {
unsigned long ip = h->m_ip;
unsigned short port = h->m_httpPort;
//}
//sprintf ( x , "http://%s:%li/get?q=" , iptoa ( ip ) , port );
// . we no longer put the port in here
// . but still need http:// since we use <base href=>
if (port == 80) sprintf(x,"http://%s/get?q=",iptoa(ip));
else sprintf(x,"http://%s:%hu/get?q=",iptoa(ip),port);
x += gbstrlen ( x );
// the query url encoded
long elen = urlEncode ( x , thisUrlEnd - x , q , qlen );
x += elen;
// separate cgi vars with a &
//sprintf ( x, "&seq=%li&rtq=%lid=%lli",
// (long)st->m_seq,(long)st->m_rtq,st->m_msg22.getDocId());
sprintf ( x, "&d=%lli",st->m_docId );
x += gbstrlen(x);
// set our query for highlighting
Query qq;
qq.set2 ( q, st->m_langId , true );
// print the query terms into our highlight buffer
Highlight hi;
// make words so we can set the scores to ignore fielded terms
Words qw;
qw.set ( q , // content being highlighted, utf8
qlen , // content being highlighted, utf8
TITLEREC_CURRENT_VERSION,
true , // computeIds
false ); // hasHtmlEntities?
// . assign scores of 0 to query words that should be ignored
// . TRICKY: loop over words in qq.m_qwords, but they should be 1-1
// with words in qw.
// . sanity check
//if ( qw.getNumWords() != qq.m_numWords ) { char *xx = NULL; *xx = 0;}
// declare up here
Matches m;
// do the loop
//Scores ss;
//ss.set ( &qw , NULL );
//for ( long i = 0 ; i < qq.m_numWords ; i++ )
// if ( ! m.matchWord ( &qq.m_qwords[i],i ) ) ss.m_scores[i] = 0;
// now set m.m_matches[] to those words in qw that match a query word
// or phrase in qq.
m.setQuery ( &qq );
//m.addMatches ( &qw , &ss , true );
m.addMatches ( &qw );
long hilen = 0;
// CNS: if ( ! st->m_clickNScroll ) {
// and highlight the matches
if ( printDisclaimer ) {
hilen = hi.set ( //p ,
//avail ,
sb ,
&qw , // words to highlight
&m , // matches relative to qw
false , // doSteming
false , // st->m_clickAndScroll ,
(char *)thisUrl );// base url for ClcknScrll
//p += hilen;
// now an hr
//memcpy ( p , "</span></table></table>\n" , 24 ); p += 24;
sb->safeStrcpy("</span></table></table>\n");
}
bool includeHeader = st->m_includeHeader;
// do not show header for json object display
if ( xd->m_contentType == CT_JSON )
includeHeader = false;
if ( format == FORMAT_XML ) includeHeader = false;
if ( format == FORMAT_JSON ) includeHeader = false;
//mfree(uq, uqCapacity, "PageGet");
// undo the header writes if we should
if ( ! includeHeader ) {
// including base href is off by default when not including
// the header, so the caller must explicitly turn it back on
if ( st->m_includeBaseHref ) sb->m_length=startLen2;//p=start2;
else sb->m_length=startLen1;//p=start1;
}
//sb->safeStrcpy(tbuf);
if ( format == FORMAT_XML ) {
sb->safePrintf("<response>\n");
sb->safePrintf("<statusCode>0</statusCode>\n");
sb->safePrintf("<statusMsg>Success</statusMsg>\n");
sb->safePrintf("<url><![CDATA[");
sb->cdataEncode(xd->m_firstUrl.m_url);
sb->safePrintf("]]></url>\n");
sb->safePrintf("<docId>%llu</docId>\n",xd->m_docId);
sb->safePrintf("\t<cachedTimeUTC>%lu</cachedTimeUTC>\n",
lastSpiderDate);
sb->safePrintf("\t<cachedTimeStr>%s</cachedTimeStr>\n",tbuf);
}
if ( format == FORMAT_JSON ) {
sb->safePrintf("{\"response\":{\n");
sb->safePrintf("\t\"statusCode\":0,\n");
sb->safePrintf("\t\"statusMsg\":\"Success\",\n");
sb->safePrintf("\t\"url\":\"");
sb->jsonEncode(xd->m_firstUrl.m_url);
sb->safePrintf("\",\n");
sb->safePrintf("\t\"docId\":%llu,\n",xd->m_docId);
sb->safePrintf("\t\"cachedTimeUTC\":%lu,\n",lastSpiderDate);
sb->safePrintf("\t\"cachedTimeStr\":\"%s\",\n",tbuf);
}
// identify start of <title> tag we wrote out
char *sbstart = sb->getBufStart();
char *sbend = sb->getBufEnd();
char *titleStart = NULL;
char *titleEnd = NULL;
for ( char *t = sbstart ; t < sbend ; t++ ) {
// title tag?
if ( t[0]!='<' ) continue;
if ( to_lower_a(t[1])!='t' ) continue;
if ( to_lower_a(t[2])!='i' ) continue;
if ( to_lower_a(t[3])!='t' ) continue;
if ( to_lower_a(t[4])!='l' ) continue;
if ( to_lower_a(t[5])!='e' ) continue;
// point to it
char *x = t + 5;
// max - to keep things fast
char *max = x + 500;
for ( ; *x && *x != '>' && x < max ; x++ );
x++;
// find end
char *e = x;
for ( ; *e && e < max ; e++ ) {
if ( e[0]=='<' &&
to_lower_a(e[1])=='/' &&
to_lower_a(e[2])=='t' &&
to_lower_a(e[3])=='i' &&
to_lower_a(e[4])=='t' &&
to_lower_a(e[5])=='l' &&
to_lower_a(e[6])=='e' )
break;
}
if ( e < max ) {
titleStart = x;
titleEnd = e;
}
break;
}
// . print title at top!
// . consider moving
if ( titleStart ) {
char *ebuf = st->m_r.getString("eb");
if ( ! ebuf ) ebuf = "";
//p += sprintf ( p ,
sb->safePrintf(
"<table border=1 "
"cellpadding=10 "
"cellspacing=0 "
"width=100%% "
"color=#ffffff>" );
long printLinks = st->m_r.getLong("links",0);
if ( ! printDisclaimer && printLinks )
sb->safePrintf(//p += sprintf ( p ,
// first put cached and live link
"<tr>"
"<td bgcolor=lightyellow>"
// print cached link
//"<center>"
" "
"<b>"
"<a "
"style=\"font-size:18px;font-weight:600;"
"color:#000000;\" "
"href=\""
"/get?"
"c=%s&d=%lli&qh=0&cnsp=1&eb=%s\">"
"cached link</a>"
" "
"<a "
"style=\"font-size:18px;font-weight:600;"
"color:#000000;\" "
"href=%s>live link</a>"
"</b>"
//"</center>"
"</td>"
"</tr>\n"
,st->m_coll
,st->m_docId
,ebuf
,thisUrl // st->ptr_ubuf
);
if ( printLinks ) {
sb->safePrintf(//p += sprintf ( p ,
"<tr><td bgcolor=pink>"
"<span style=\"font-size:18px;"
"font-weight:600;"
"color:#000000;\">"
" "
"<b>PAGE TITLE:</b> "
);
long tlen = titleEnd - titleStart;
sb->safeMemcpy ( titleStart , tlen );
sb->safePrintf ( "</span></td></tr>" );
}
sb->safePrintf( "</table><br>\n" );
}
// is the content preformatted?
bool pre = false;
char ctype = (char)xd->m_contentType;
if ( ctype == CT_TEXT ) pre = true ; // text/plain
if ( ctype == CT_DOC ) pre = true ; // filtered msword
if ( ctype == CT_PS ) pre = true ; // filtered postscript
if ( format == FORMAT_XML ) pre = false;
if ( format == FORMAT_JSON ) pre = false;
// if it is content-type text, add a <pre>
if ( pre ) {//p + 5 < bufEnd && pre ) {
sb->safePrintf("<pre>");
//p += 5;
}
if ( st->m_strip == 1 )
contentLen = stripHtml( content, contentLen,
(long)xd->m_version, st->m_strip );
// it returns -1 and sets g_errno on error, line OOM
if ( contentLen == -1 ) {
//if ( buf ) mfree ( buf , bufMaxSize , "PageGet2" );
return sendErrorReply ( st , g_errno );
}
Xml xml;
Words ww;
// if no highlighting, skip it
bool queryHighlighting = st->m_queryHighlighting;
if ( st->m_strip == 2 ) queryHighlighting = false;
// do not do term highlighting if json
if ( xd->m_contentType == CT_JSON )
queryHighlighting = false;
SafeBuf tmp;
SafeBuf *xb = sb;
if ( format == FORMAT_XML ) xb = &tmp;
if ( format == FORMAT_JSON ) xb = &tmp;
if ( ! queryHighlighting ) {
xb->safeMemcpy ( content , contentLen );
//p += contentLen ;
}
else {
// get the content as xhtml (should be NULL terminated)
//Words *ww = xd->getWords();
if ( ! xml.set ( content , contentLen , false ,
0 , false , TITLEREC_CURRENT_VERSION ,
false , 0 , CT_HTML ) ) { // niceness is 0
//if ( buf ) mfree ( buf , bufMaxSize , "PageGet2" );
return sendErrorReply ( st , g_errno );
}
if ( ! ww.set ( &xml , true , 0 ) ) { // niceness is 0
//if ( buf ) mfree ( buf , bufMaxSize , "PageGet2" );
return sendErrorReply ( st , g_errno );
}
// sanity check
//if ( ! xd->m_wordsValid ) { char *xx=NULL;*xx=0; }
// how much space left in p?
//avail = bufEnd - p;
Matches m;
m.setQuery ( &qq );
m.addMatches ( &ww );
hilen = hi.set ( xb , // p , avail ,
&ww , &m ,
false /*doStemming?*/ ,
st->m_clickAndScroll ,
thisUrl /*base url for click & scroll*/);
//p += hilen;
log(LOG_DEBUG, "query: Done highlighting cached page content");
}
if ( format == FORMAT_XML ) {
sb->safePrintf("\t<content><![CDATA[");
sb->cdataEncode ( xb->getBufStart() );
sb->safePrintf("]]></content>\n");
sb->safePrintf("</response>\n");
}
if ( format == FORMAT_JSON ) {
sb->safePrintf("\t\"content\":\"\n");
sb->jsonEncode ( xb->getBufStart() );
sb->safePrintf("\"\n}\n}\n");
}
// if it is content-type text, add a </pre>
if ( pre ) { // p + 6 < bufEnd && pre ) {
sb->safeMemcpy ( "</pre>" , 6 );
//p += 6;
}
// calculate bufLen
//long bufLen = p - buf;
long ct = xd->m_contentType;
// now filter the entire buffer to escape out the xml tags
// so it is displayed nice
SafeBuf newbuf;
if ( ct == CT_XML ) {
// encode the xml tags into <tagname> sequences
if ( !newbuf.htmlEncodeXmlTags ( sb->getBufStart() ,
sb->getLength(),
0)){// niceness=0
//if ( buf ) mfree ( buf , bufMaxSize , "PageGet2" );
return sendErrorReply ( st , g_errno );
}
// free out buffer that we alloc'd before returning since this
// should have copied it into another buffer
//if ( buf ) mfree ( buf , bufMaxSize , "PageGet2" );
// reassign
//buf = newbuf.getBufStart();
//bufLen = newbuf.length();
sb->stealBuf ( &newbuf );
}
// now encapsulate it in html head/tail and send it off
// sendErr:
contentType = "text/html";
if ( strip == 2 ) contentType = "text/xml";
// xml is usually buggy and this throws browser off
//if ( ctype == CT_XML ) contentType = "text/xml";
if ( xd->m_contentType == CT_JSON )
contentType = "application/json";
if ( format == FORMAT_XML ) contentType = "text/xml";
if ( format == FORMAT_JSON ) contentType = "application/json";
// safebuf, sb, is a member of "st" so this should copy the buffer
// when it constructs the http reply, and we gotta call delete(st)
// AFTER this so sb is still valid.
bool status = g_httpServer.sendDynamicPage (s,
//buf,bufLen,
sb->getBufStart(),
sb->getLength(),
-1,false,
contentType,
-1, NULL, "utf8" );
// nuke state2
mdelete ( st , sizeof(State2) , "PageGet1" );
delete (st);
// free out buffer that we alloc'd before returning since this
// should have copied it into another buffer
//if ( ct == CT_XML ) newbuf.purge();
//else if ( buf ) mfree ( buf , bufMaxSize , "PageGet2" );
// and convey the status
return status;
}
bool sendTurkPageReply ( State60 *st ) {
XmlDoc *xd = &st->m_xd;
//char *content = xd->ptr_utf8Content;
//int32_t contentLen = xd->size_utf8Content - 1;
// count the total number of EventDesc classes for all evids
//char *evd = xd->ptr_eventData;
//EventDisplay *ed = (EventDisplay *)evd;
//char *addr = evd + (int32_t)ed->m_addr;
//char timeZoneOffset = getTimeZoneFromAddr ( addr );
// in case getSections() block come right back in
xd->setCallback ( st , xdcallback );
// . set niceness to 1 so all this processing doesn't slow queries down
// . however, g_niceness should still be zero... hmmm...
xd->m_niceness = 1;
// default to 1 niceness
st->m_niceness = 1;
// now set the sections class
Sections *ss = xd->getSections();
// now for each section with alnum text, telescope up as far as
// possible without containing anymore alnum text than what it
// contained. set SEC_CONTROL bit. such sections will have the
// 2 green/blue dots, that are used for turning on/off title/desc.
// but really the indians will only turn off sections that should
// not have a title/desc.
for ( Section *si = ss->m_rootSection ; si ; si = si->m_next ) {
// breathe
QUICKPOLL(st->m_niceness);
// skip if does not have text
if ( si->m_firstWordPos < 0 ) continue;
// otherwise, find biggest parent that contains just that text
Section *p = si->m_parent;
Section *last = si;
for ( ; p ; p = p->m_parent ) {
if ( p->m_firstWordPos != si->m_firstWordPos ) break;
if ( p->m_lastWordPos != si->m_lastWordPos ) break;
last = p;
}
// set that bit then
last->m_flags |= SEC_CONTROL;
// and speed up the loop
si = last;
}
// * now each SEC_CONTROL sections have a fence activated by a turker
// * an event title or description can not span a fence. it must be
// confined within a fence. however, it is allowed to include
// title or description from a "title section".
// * hold shift down to designate as title section when clicking it
// * show the raw text of each event changing as you fence
// sections in or out. show in a right frame.
// * show list of events on page in the top frame. can toggle them
// all individually.
// * and remove no-display from all tags so we can see everything.
// * highlight addresses, not just dates.
// * each section hash has its own unique bg color when activated
// * with a single click, completely reject an event because:
// contains bad time, address, title or desc. specify which so
// we can improve our algo.
// * when selecting an individual event, scroll to its tod...
// * remove all color from webpage that we can so our colors show up
// * remove all imgs. just src them to dev null.
// * allow for entering a custom title for an event or all events
// that are or will ever appear on the page.
// * when displaying the text of the events, use hyphens to
// delineate the section topology. strike out text as a section
// fence is activated.
// * when a section is activated is it easier to just redownload
// the whole text of the page? maybe just the text frame?
// * clicking on an individual sentence section should just remove
// that sentence. that is kinda a special content hash removal
// tag. like "Click here for video."
// * when an event id is selected i guess activate its bgcolor to
// be light blue for all sentences currently in the event that
// are not in activated sections. (make exception for designated
// title sections). so we need multiple tags for each events
// sentence div section. if sentence is split use multiple div tags
// then to keep the order. so each event sentence would have
// <div ev1=1 ev2=1 ev10=1>...</div> if it is in event ids 1,2 and
// 10. that way we can activate it when one of those event ids is
// activated.
SafeBuf sb;
// int16_tcuts
if ( ! xd->m_wordsValid ) { char *xx=NULL;*xx=0; }
Words *words = &xd->m_words;
int32_t nw = words->getNumWords();
char **wptrs = words->getWords();
int32_t *wlens = words->getWordLens();
nodeid_t *tids = words->getTagIds();
// a special array for printing </div> tags
char *endCounts = (char *)mcalloc ( nw ,"endcounts");
if ( ! endCounts ) return sendErrorReply ( st , g_errno );
//
// now loop over all the words. if word starts a section that has
// SEC_CONTROL bit set, and print out the section hash and a color
// tag to be activated if the turkey activates us.
// CAUTION: word may start multiple sections.
//
for ( int32_t i = 0 ; i < nw ; i++ ) {
// get section ptr
Section *sj = ss->m_sectionPtrs[i];
// sanity check. sj must be first section ptr that starts @ a
if ( sj && sj->m_a==i && sj->m_prev && sj->m_prev->m_a==i ) {
char *xx=NULL;*xx=0; }
// . does word #i start a section?
// . if section is control, print out the control
while ( sj && sj->m_a == i ) {
// print this section's hash
if ( sj->m_flags & SEC_CONTROL) {
// after the turkeys have made all the edits
// they need to submit the changes they made.
// how can we get that data sent back to the
// back end? we need to send back the colors
// of the sections that have been activated
// i guess. just do a loop over them.
sb.safePrintf("<div nobreak gbsecid=%"UINT32" "
"bgcolor=#%"XINT32" "
"onclick=gbtogglecolor()>",
(uint32_t)sj->m_tagHash,
(uint32_t)sj->m_tagHash);
// sanity check
if ( sj->m_b < 0 ) { char *xx=NULL;*xx=0; }
if ( sj->m_b > nw ) { char *xx=NULL;*xx=0; }
// and inc the /div count for that word
endCounts[sj->m_b-1]++;
}
// try next section too
sj = sj->m_next;
}
// if this is a tag, remove any coloring
if ( tids[i] ) {
}
// print the word, be it a tag, alnum, punct
sb.safeMemcpy ( wptrs[i] , wlens[i] );
// end a div tag?
if ( ! endCounts[i] ) continue;
// might be many so loop it
for ( int32_t j = 0 ; j < endCounts[i] ; j++ )
sb.safePrintf("</div>");
}
return false;
}
bool sendReply ( void *state ) {
StateCatdb *st = (StateCatdb*)state;
// check for error
if (g_errno) {
if (st->m_catLookup)
log("PageCatdb: Msg8b had error getting Site Rec: %s",
mstrerror(g_errno));
else
log("PageCatdb: Msg2a had error generating Catdb: %s",
mstrerror(g_errno));
st->m_catLookup = false;
g_errno = 0;
}
long long endTime = gettimeofdayInMilliseconds();
// page buffer
SafeBuf sb;
sb.reserve(64*1024);
// . print standard header
// . do not print big links if only an assassin, just print host ids
g_pages.printAdminTop ( &sb, st->m_socket , &st->m_r );
sb.safePrintf(
"<style>"
".poo { background-color:#%s;}\n"
"</style>\n" ,
LIGHT_BLUE );
sb.safePrintf ( "<table %s>"
"<tr><td colspan=2>"
"<center><font size=+1><b>Catdb</b></font></center>"
"</td></tr>", TABLE_STYLE );
// instructions
sb.safePrintf("<tr bgcolor=#%s>"
"<td colspan=3>"
"<font size=-2>"
"<center>"
"Don't just start using this, you need to follow the "
"instructions in the <i>admin guide</i> for adding "
"DMOZ support."
"</center>"
"</font>"
"</td>"
"</tr>"
,DARK_BLUE
);
// print the generate Catdb link
sb.safePrintf ( "<tr class=poo><td>Update Catdb from DMOZ data.</td>"
"<td><center>"
"<a href=\"/master/catdb?c=%s&gencatdb=2\">"
"Update Catdb</a> "
"</center></td></tr>",
st->m_coll );
sb.safePrintf ( "<tr class=poo>"
"<td>Generate New Catdb from DMOZ data.</td>"
"<td><center>"
"<a href=\"/master/catdb?c=%s&gencatdb=1\">"
"Generate Catdb</a> "
"</center></td></tr>",
st->m_coll );
if (st->m_genCatdb)
sb.safePrintf ( "<tr class=poo>"
"<td> Catdb Generation took %lli ms."
"</td></tr>",
endTime - st->m_startTime );
// print Url Catgory Lookup
sb.safePrintf ( "<tr class=poo><td>Lookup Category of Url.</td>"
"<td><input type=text name=caturl size=80"
" value=\"");
if (st->m_catLookup) {
sb.safeMemcpy(st->m_url.getUrl(), st->m_url.getUrlLen());
}
sb.safePrintf("\"></center></td></tr>" );
// print Url Info if Lookup was done
if (st->m_catLookup) {
sb.safePrintf("<tr><td>");
// print the url
sb.safeMemcpy(st->m_url.getUrl(), st->m_url.getUrlLen());
sb.safePrintf(" (%lli ms)</td><td>",
endTime - st->m_startTime );
// print each category id and path
for (long i = 0; i < st->m_catRec.m_numCatids; i++) {
sb.safePrintf("<b>[%li] ",
st->m_catRec.m_catids[i]);
g_categories->printPathFromId(&sb,
st->m_catRec.m_catids[i]);
sb.safePrintf("</b><br>");
// lookup title and summary
char title[1024];
long titleLen = 0;
char summ[4096];
long summLen = 0;
char anchor[256];
unsigned char anchorLen = 0;
g_categories->getTitleAndSummary(
st->m_url.getUrl(),
st->m_url.getUrlLen(),
st->m_catRec.m_catids[i],
title,
&titleLen,
1023,
summ,
&summLen,
4098,
anchor,
&anchorLen,
255 );
title[titleLen] = '\0';
summ[summLen] = '\0';
anchor[anchorLen] = '\0';
// print title and summary
sb.safePrintf("<b>Title:</b> %s<br>"
"<b>Summary:</b> %s<br>",
title, summ);
if (anchorLen > 0)
sb.safePrintf("<b>Anchor:</b> %s<br>",
anchor);
sb.safePrintf("<br>");
}
sb.safePrintf("<b>Filenum:</b> %li<br>",
st->m_catRec.m_filenum);
// print indirect catids
if (st->m_catRec.m_numIndCatids > 0) {
sb.safePrintf("<hr><b>Indirect Catids [%li]:"
"</b><br>\n",
st->m_catRec.m_numIndCatids );
for (long i = 0;
i < st->m_catRec.m_numIndCatids; i++) {
sb.safePrintf("%lu<br>",
st->m_catRec.m_indCatids[i]);
}
}
sb.safePrintf("</td></tr>");
}
// end it
sb.safePrintf ( "</center></td></tr></table>" );
// print submit button
sb.safePrintf ( "<br><center>"
"<input type=submit value=\"Submit\" border=0>"
"</form></center>" );
// print the final tail
//p += g_httpServer.printTail ( p , pend - p );
// clear g_errno, if any, so our reply send goes through
g_errno = 0;
// extract the socket
TcpSocket *s = st->m_socket;
// clear the state
mdelete ( st, sizeof(StateCatdb), "PageCatdb" );
delete st;
// . send this page
// . encapsulates in html header and tail
// . make a Mime
return g_httpServer.sendDynamicPage(s , sb.getBufStart(), sb.length());
}
bool Msg3a::gotAllSplitReplies ( ) {
// if any of the split requests had an error, give up and set m_errno
// but don't set if for non critical errors like query truncation
if ( m_errno ) {
g_errno = m_errno;
return true;
}
// also reset the finalbuf and the oldNumTopDocIds
if ( m_finalBuf ) {
mfree ( m_finalBuf, m_finalBufSize, "Msg3aF" );
m_finalBuf = NULL;
m_finalBufSize = 0;
}
// update our estimated total hits
m_numTotalEstimatedHits = 0;
for ( long i = 0; i < m_numHosts ; i++ ) {
// get that host that gave us the reply
//Host *h = g_hostdb.getHost(i);
// . get the reply from multicast
// . multicast should have destroyed all slots, but saved reply
// . we are responsible for freeing the reply
// . we need to call this even if g_errno or m_errno is
// set so we can free the replies in Msg3a::reset()
// . if we don't call getBestReply() on it multicast should
// free it, because Multicast::m_ownReadBuf is still true
Multicast *m = &m_mcast[i];
bool freeit = false;
long replySize = 0;
long replyMaxSize;
char *rbuf;
Msg39Reply *mr;
// . only get it if the reply not already full
// . if reply already processed, skip
// . perhaps it had no more docids to give us or all termlists
// were exhausted on its disk and this is a re-call
// . we have to re-process it for count m_numTotalEstHits, etc.
rbuf = m->getBestReply ( &replySize ,
&replyMaxSize ,
&freeit ,
true ); //stealIt?
// cast it
mr = (Msg39Reply *)rbuf;
// in case of mem leak, re-label from "mcast" to this so we
// can determine where it came from, "Msg3a-GBR"
relabel( rbuf, replyMaxSize , "Msg3a-GBR" );
// . we must be able to free it... we must own it
// . this is true if we should free it, but we should not have
// to free it since it is owned by the slot?
if ( freeit ) {
log(LOG_LOGIC,"query: msg3a: Steal failed.");
char *xx = NULL;
*xx=0;
}
// bad reply?
if ( ! mr ) {
log(LOG_LOGIC,"query: msg3a: Bad NULL reply.");
m_reply [i] = NULL;
m_replyMaxSize[i] = 0;
// it might have been timd out, just ignore it!!
continue;
// if size is 0 it can be Msg39 giving us an error!
g_errno = EBADREPLYSIZE;
m_errno = EBADREPLYSIZE;
// all reply buffers should be freed on reset()
return true;
}
// how did this happen?
if ( replySize < 29 && ! mr->m_errno ) {
// if size is 0 it can be Msg39 giving us an error!
g_errno = EBADREPLYSIZE;
m_errno = EBADREPLYSIZE;
log(LOG_LOGIC,"query: msg3a: Bad reply size of %li.",
replySize);
// all reply buffers should be freed on reset()
return true;
}
// can this be non-null? we shouldn't be overwriting one
// without freeing it...
if ( m_reply[i] )
// note the mem leak now
log("query: mem leaking a 0x39 reply");
// cast it and set it
m_reply [i] = mr;
m_replyMaxSize[i] = replyMaxSize;
// deserialize it (just sets the ptr_ and size_ member vars)
//mr->deserialize ( );
deserializeMsg ( sizeof(Msg39Reply) ,
&mr->size_docIds,
&mr->size_clusterRecs,
&mr->ptr_docIds,
mr->m_buf );
// sanity check
if ( mr->m_nqt != m_q->getNumTerms() ) {
g_errno = EBADREPLY;
m_errno = EBADREPLY;
log("query: msg3a: Split reply qterms=%li != %li.",
(long)mr->m_nqt,(long)m_q->getNumTerms() );
return true;
}
// return if split had an error, but not for a non-critical
// error like query truncation
if ( mr->m_errno && mr->m_errno != EQUERYTRUNCATED ) {
g_errno = mr->m_errno;
m_errno = mr->m_errno;
log("query: msg3a: Split had error: %s",
mstrerror(g_errno));
return true;
}
// skip down here if reply was already set
//skip:
// add of the total hits from each split, this is how many
// total results the lastest split is estimated to be able to
// return
// . THIS should now be exact since we read all termlists
// of posdb...
m_numTotalEstimatedHits += mr->m_estimatedHits;
// debug log stuff
if ( ! m_debug ) continue;
// cast these for printing out
long long *docIds = (long long *)mr->ptr_docIds;
score_t *scores = (score_t *)mr->ptr_scores;
// print out every docid in this split reply
for ( long j = 0; j < mr->m_numDocIds ; j++ ) {
// print out score_t
logf( LOG_DEBUG,
"query: msg3a: [%lu] %03li) "
"split=%li docId=%012llu domHash=0x%02lx "
"score=%lu" ,
(unsigned long)this ,
j ,
i ,
docIds [j] ,
(long)g_titledb.getDomHash8FromDocId(docIds[j]),
(long)scores[j] );
}
}
// this seems to always return true!
mergeLists ( );
if ( ! m_r->m_useSeoResultsCache ) return true;
// now cache the reply
SafeBuf cr;
long dataSize = 4 + 4 + 4 + m_numDocIds * (8+4+4);
long need = sizeof(key_t) + 4 + dataSize;
bool status = cr.reserve ( need );
// sanity
if ( ( m_ckey.n0 & 0x01 ) == 0x00 ) {
char *xx=NULL;
*xx=0;
}
// ignore errors
g_errno = 0;
// return on error with g_errno cleared if cache add failed
if ( ! status ) return true;
// add to buf otherwise
cr.safeMemcpy ( &m_ckey , sizeof(key_t) );
cr.safeMemcpy ( &dataSize , 4 );
long now = getTimeGlobal();
cr.pushLong ( now );
cr.pushLong ( m_numDocIds );
cr.pushLong ( m_numTotalEstimatedHits );//Results );
long max = m_numDocIds;
// then the docids
for ( long i = 0 ; i < max ; i++ )
cr.pushLongLong(m_docIds[i] );
for ( long i = 0 ; i < max ; i++ )
cr.pushFloat(m_scores[i]);
for ( long i = 0 ; i < max ; i++ )
cr.pushLong(getSiteHash26(i));
// sanity
if ( cr.length() != need ) {
char *xx=NULL;
*xx=0;
}
// make these
key_t startKey;
key_t endKey;
startKey = m_ckey;
// clear delbit
startKey.n0 &= 0xfffffffffffffffeLL;
// end key is us
endKey = m_ckey;
// that is the single record
m_seoCacheList.set ( cr.getBufStart() ,
cr.length(),
cr.getBufStart(), // alloc
cr.getCapacity(), // alloc size
(char *)&startKey,
(char *)&endKey,
-1, // fixeddatasize
true, // owndata?
false,// use half keys?
sizeof(key_t) );
// do not allow cr to free it, msg1 will
cr.detachBuf();
// note it
//log("seopipe: storing ckey=%s q=%s"
// ,KEYSTR(&m_ckey,12)
// ,m_r->ptr_query
// );
//log("msg1: sending niceness=%li",(long)m_r->m_niceness);
// this will often block, but who cares!? it just sends a request off
if ( ! m_msg1.addList ( &m_seoCacheList ,
RDB_SERPDB,//RDB_CACHEDB,
m_r->ptr_coll,
this, // state
gotSerpdbReplyWrapper, // callback
false, // forcelocal?
m_r->m_niceness ) ) {
//log("blocked");
return false;
}
// we can safely delete m_msg17... just return true
return true;
}
// . return the score of the highest-scoring window containing match #m
// . window is defined by the half-open interval [a,b) where a and b are
// word #'s in the Words array indicated by match #m
// . return -1 and set g_errno on error
int64_t Summary::getBestWindow ( Matches *matches, int32_t mm, int32_t *lasta,
int32_t *besta, int32_t *bestb, char *gotIt,
char *retired, int32_t maxExcerptLen ) {
// get the window around match #mm
Match *m = &matches->m_matches[mm];
// what is the word # of match #mm?
int32_t matchWordNum = m->m_wordNum;
// what Words/Pos/Bits classes is this match in?
Words *words = m->m_words;
Section **sp = NULL;
int32_t *pos = m->m_pos->m_pos;
// use "m_swbits" not "m_bits", that is what Bits::setForSummary() uses
const swbit_t *bb = m->m_bits->m_swbits;
// shortcut
if ( m->m_sections ) {
sp = m->m_sections->m_sectionPtrs;
}
int32_t nw = words->getNumWords();
int64_t *wids = words->getWordIds();
nodeid_t *tids = words->getTagIds();
// . sanity check
// . this prevents a core i've seen
if ( matchWordNum >= nw ) {
log("summary: got overflow condition for q=%s",m_q->m_orig);
// assume no best window
*besta = -1;
*bestb = -1;
*lasta = matchWordNum;
return 0;
}
// . we NULLify the section ptrs if we already used the word in another summary.
int32_t badFlags = SEC_SCRIPT|SEC_STYLE|SEC_SELECT|SEC_IN_TITLE;
if ( (bb[matchWordNum] & D_USED) || ( sp && (sp[matchWordNum]->m_flags & badFlags) ) ) {
// assume no best window
*besta = -1;
*bestb = -1;
*lasta = matchWordNum;
return 0;
}
// . "a" is the left fence post of the window (it is a word # in Words)
// . go to the left as far as we can
// . thus we decrement "a"
int32_t a = matchWordNum;
// "posa" is the character position of the END of word #a
int32_t posa = pos[a+1];
int32_t firstFrag = -1;
bool startOnQuote = false;
bool goodStart = false;
int32_t wordCount = 0;
// . decrease "a" as int32_t as we stay within maxNumCharsPerLine
// . avoid duplicating windows by using "lasta", the last "a" of the
// previous call to getBestWindow(). This can happen if our last
// central query term was close to this one.
for ( ; a > 0 && posa - pos[a-1] < maxExcerptLen && a > *lasta; a-- ) {
// . don't include any "dead zone",
// . dead zones have already been used for the summary, and
// we are getting a second/third/... excerpt here now then
// stop if its the start of a sentence, too
// stop before title word
if ( (bb[a-1] & D_USED) || (bb[a] & D_STARTS_SENTENCE) || ( bb[a-1] & D_IN_TITLE )) {
goodStart = true;
break;
}
// don't go beyond an LI, TR, P tag
if ( tids && ( tids[a-1] == TAG_LI ||
tids[a-1] == TAG_TR ||
tids[a-1] == TAG_P ||
tids[a-1] == TAG_DIV ) ) {
goodStart = true;
break;
}
// stop if its the start of a quoted sentence
if ( a+1<nw && (bb[a+1] & D_IN_QUOTES) &&
words->getWord(a)[0] == '\"' ){
startOnQuote = true;
goodStart = true;
break;
}
// find out the first instance of a fragment (comma, etc)
// watch out! because frag also means 's' in there's
if ( ( bb[a] & D_STARTS_FRAG ) && !(bb[a-1] & D_IS_STRONG_CONNECTOR) && firstFrag == -1 ) {
firstFrag = a;
}
if ( wids[a] ) {
wordCount++;
}
}
// if didn't find a good start, then start at the start of the frag
if ( !goodStart && firstFrag != -1 ) {
a = firstFrag;
}
// don't let punct or tag word start a line, unless a quote
if ( a < matchWordNum && !wids[a] && words->getWord(a)[0] != '\"' ){
while ( a < matchWordNum && !wids[a] ) a++;
// do not break right after a "strong connector", like
// apostrophe
while ( a < matchWordNum && a > 0 &&
( bb[a-1] & D_IS_STRONG_CONNECTOR ) )
a++;
// don't let punct or tag word start a line
while ( a < matchWordNum && !wids[a] ) a++;
}
// remember, b is not included in the summary, the summary is [a,b-1]
// remember to include all words in a matched phrase
int32_t b = matchWordNum + m->m_numWords ;
int32_t endQuoteWordNum = -1;
int32_t numTagsCrossed = 0;
for ( ; b <= nw; b++ ) {
if ( b == nw ) {
break;
}
if ( pos[b+1] - pos[a] >= maxExcerptLen ) {
break;
}
if ( startOnQuote && words->getWord(b)[0] == '\"' ) {
endQuoteWordNum = b;
}
// don't include any dead zone, those are already-used samples
if ( bb[b] & D_USED ) {
break;
}
// stop on a title word
if ( bb[b] & D_IN_TITLE ) {
break;
}
if ( wids[b] ) {
wordCount++;
}
// don't go beyond an LI or TR backtag
if ( tids && ( tids[b] == (BACKBIT|TAG_LI) ||
tids[b] == (BACKBIT|TAG_TR) ) ) {
numTagsCrossed++;
// try to have atleast 10 words in the summary
if ( wordCount > 10 ) {
break;
}
}
// go beyond a P or DIV backtag in case the earlier char is a
// ':'. This came from a special case for wikipedia pages
// eg. http://en.wikipedia.org/wiki/Flyover
if ( tids && ( tids[b] == (BACKBIT|TAG_P) ||
tids[b] == (BACKBIT|TAG_DIV) )) {
numTagsCrossed++;
// try to have atleast 10 words in the summary
if ( wordCount > 10 && words->getWord(b-1)[0] != ':' ) {
break;
}
}
}
// don't end on a lot of punct words
if ( b > matchWordNum && !wids[b-1]){
// remove more than one punct words. if we're ending on a quote
// keep it
while ( b > matchWordNum && !wids[b-2] && endQuoteWordNum != -1 && b > endQuoteWordNum ) {
b--;
}
// do not break right after a "strong connector", like apostrophe
while ( b > matchWordNum && (bb[b-2] & D_IS_STRONG_CONNECTOR) ) {
b--;
}
}
Match *ms = matches->m_matches;
// make m_matches.m_matches[mi] the first match in our [a,b) window
int32_t mi ;
// . the match at the center of the window is match #"mm", so that
// matches->m_matches[mm] is the Match class
// . set "mi" to it and back up "mi" as int32_t as >= a
for ( mi = mm ; mi > 0 && ms[mi-1].m_wordNum >=a ; mi-- )
;
// now get the score of this excerpt. Also mark all the represented
// query words. Mark the represented query words in the array that
// comes to us. also mark how many times the same word is repeated in
// this summary.
int64_t score = 0LL;
// is a url contained in the summary, that looks bad! punish!
bool hasUrl = false;
// the word count we did above was just an approximate. count it right
wordCount = 0;
// for debug
//char buf[5000];
//char *xp = buf;
SafeBuf xp;
// wtf?
if ( b > nw ) {
b = nw;
}
// first score from the starting match down to a, including match
for ( int32_t i = a ; i < b ; i++ ) {
// debug print out
if ( g_conf.m_logDebugSummary ) {
int32_t len = words->getWordLen(i);
char cs;
for (int32_t k=0;k<len; k+=cs ) {
const char *c = words->getWord(i)+k;
cs = getUtf8CharSize(c);
if ( is_binary_utf8 ( c ) ) {
continue;
}
xp.safeMemcpy ( c , cs );
xp.nullTerm();
}
}
// skip if in bad section, marquee, select, script, style
if ( sp && (sp[i]->m_flags & badFlags) ) {
continue;
}
// don't count just numeric words
if ( words->isNum(i) ) {
continue;
}
// check if there is a url. best way to check for '://'
if ( wids && !wids[i] ) {
const char *wrd = words->getWord(i);
int32_t wrdLen = words->getWordLen(i);
if ( wrdLen == 3 && wrd[0] == ':' && wrd[1] == '/' && wrd[2] == '/' ) {
hasUrl = true;
}
}
// skip if not wid
if ( ! wids[i] ) {
continue;
}
// just make every word 100 pts
int32_t t = 100;
// penalize it if in one of these sections
if ( bb[i] & ( D_IN_PARENS | D_IN_SUP | D_IN_LIST ) ) {
t /= 2;
}
// boost it if in bold or italics
if ( bb[i] & D_IN_BOLDORITALICS ) {
t *= 2;
}
// add the score for this word
score += t;
// print the score, "t"
if ( g_conf.m_logDebugSummary ) {
xp.safePrintf("(%" PRId32")",t);
}
// count the alpha words we got
wordCount++;
// if no matches left, skip
if ( mi >= matches->m_numMatches ) {
continue;
}
// get the match
Match *next = &ms[mi];
// skip if not a match
if ( i != next->m_wordNum ) {
continue;
}
// must be a match in this class
if ( next->m_words != words ) {
continue;
}
// advance it
mi++;
// which query word # does it match
int32_t qwn = next->m_qwordNum;
if ( qwn < 0 || qwn >= m_q->m_numWords ){g_process.shutdownAbort(true);}
// undo old score
score -= t;
// add 100000 per match
t = 100000;
// weight based on tf, goes from 0.1 to 1.0
t = (int32_t)((float)t * m_wordWeights [ qwn ]);
// if it is a query stop word, make it 10000 pts
if ( m_q->m_qwords[qwn].m_isQueryStopWord ) {
t = 0;//10000;
}
// penalize it if in one of these sections
if ( bb[i] & ( D_IN_PARENS | D_IN_SUP | D_IN_LIST ) ) {
t /= 2;
}
if ( gotIt[qwn] > 0 ) {
// have we matched it in this [a,b) already?
if ( gotIt[qwn] == 1 ) {
t /= 15;
} else {
// if we have more than 2 matches in the same window,
// it may not give a good summary. give a heavy penalty
t -= 200000;
}
} else if ( retired [qwn] > 0 ) {
// have we matched it already in a winning window?
t /= 12;
}
// add it back
score += t;
if ( g_conf.m_logDebugSummary ) {
xp.safePrintf ("[%" PRId32"]{qwn=%" PRId32",ww=%f}",t,qwn,
m_wordWeights[qwn]);
}
// inc the query word count for this window
if ( gotIt[qwn] < 100 ) {
gotIt[qwn]++;
}
}
int32_t oldScore = score;
// apply the bonus if it starts or a sentence
// only apply if the score is positive and if the wordcount is decent
if ( score > 0 && wordCount > 7 ){
// a match can give us 10k to 100k pts based on the tf weights
// so we don't want to overwhelm that too much, so let's make
// this a 20k bonus if it starts a sentence
if ( bb[a] & D_STARTS_SENTENCE ) {
score += 8000;
} else if ( bb[a] & D_STARTS_FRAG ) {
// likewise, a fragment, like after a comma
score += 4000;
}
// 1k if the match word is very close to the
// start of a sentence, lets say 3 alphawords
if ( matchWordNum - a < 7 ) {
score += 1000;
}
}
// a summary isn't really a summary if its less than 7 words.
// reduce the score, but still give it a decent score.
// minus 5M.
if ( wordCount < 7 ) {
score -= 20000;
}
// summaries that cross a lot of tags are usually bad, penalize them
if ( numTagsCrossed > 1 ) {
score -= (numTagsCrossed * 20000);
}
if ( hasUrl ) {
score -= 8000;
}
// show it
if ( g_conf.m_logDebugSummary ) {
log(LOG_DEBUG, "sum: score=%08" PRId32" prescore=%08" PRId32" a=%05" PRId32" b=%05" PRId32" %s",
(int32_t)score,oldScore,(int32_t)a,(int32_t)b,
xp.getBufStart());
}
// set lasta, besta, bestb
*lasta = a;
*besta = a;
*bestb = b;
return score;
}
//
// . ENTRY POINT FOR IMPORTING TITLEDB RECS FROM ANOTHER CLUSTER
// . when user clicks 'begin' in import page we come here..
// . so when that parm changes in Parms.cpp we sense that and call
// beginImport(CollectionRec *cr)
// . or on startup we call resumeImports to check each coll for
// an import in progress.
// . search for files named titledb*.dat
// . if none found just return
// . when msg7 inject competes it calls this
// . call this from sleep wrapper in Process.cpp
// . returns false if would block (outstanding injects), true otherwise
// . sets g_errno on error
bool ImportState::importLoop ( ) {
CollectionRec *cr = g_collectiondb.getRec ( m_collnum );
if ( ! cr || g_hostdb.m_hostId != 0 ) {
// if coll was deleted!
log("import: collnum %li deleted while importing into",
(long)m_collnum);
//if ( m_numOut > m_numIn ) return true;
// delete the entire import state i guess
// what happens if we have a msg7 reply come back in?
// it should see the collrec is NULL and just fail.
mdelete ( this, sizeof(ImportState) , "impstate");
delete (this);
return true;
}
INJECTLOOP:
// stop if waiting on outstanding injects
long long out = m_numOut - m_numIn;
if ( out >= cr->m_numImportInjects ) {
g_errno = 0;
return false;
}
if ( ! cr->m_importEnabled ) {
// wait for all to return
if ( out > 0 ) return false;
// then delete it
log("import: collnum %li import loop disabled",
(long)m_collnum);
mdelete ( this, sizeof(ImportState) , "impstate");
delete (this);
return true;
}
// scan each titledb file scanning titledb0001.dat first,
// titledb0003.dat second etc.
//long long offset = -1;
// . when offset is too big for current m_bigFile file then
// we go to the next and set offset to 0.
// . sets m_bf and m_fileOffset
if ( ! setCurrentTitleFileAndOffset ( ) ) {//cr , -1 );
log("import: import: no files to read");
//goto INJECTLOOP;
return true;
}
// this is -1 if none remain!
if ( m_fileOffset == -1 ) {
log("import: import fileoffset is -1. done.");
return true;
}
long long saved = m_fileOffset;
//Msg7 *msg7;
//GigablastRequest *gr;
//SafeBuf *sbuf = NULL;
long need = 12;
long dataSize = -1;
//XmlDoc xd;
key_t tkey;
bool status;
SafeBuf tmp;
SafeBuf *sbuf = &tmp;
long long docId;
long shardNum;
long key;
Multicast *mcast;
char *req;
long reqSize;
if ( m_fileOffset >= m_bfFileSize ) {
log("inject: import: done processing file %li %s",
m_bfFileId,m_bf.getFilename());
goto nextFile;
}
// read in title rec key and data size
status = m_bf.read ( &tkey, sizeof(key_t) , m_fileOffset );
//if ( n != 12 ) goto nextFile;
if ( g_errno ) {
log("inject: import: reading file error: %s. advancing "
"to next file",mstrerror(g_errno));
goto nextFile;
}
m_fileOffset += 12;
// if negative key, skip
if ( (tkey.n0 & 0x01) == 0 ) {
goto INJECTLOOP;
}
// if non-negative then read in size
status = m_bf.read ( &dataSize , 4 , m_fileOffset );
if ( g_errno ) {
log("main: failed to read in title rec "
"file. %s. Skipping file %s",
mstrerror(g_errno),m_bf.getFilename());
goto nextFile;
}
m_fileOffset += 4;
need += 4;
need += dataSize;
need += 4; // collnum, first 4 bytes
if ( dataSize < 0 || dataSize > 500000000 ) {
log("main: could not scan in titledb rec of "
"corrupt dataSize of %li. BAILING ENTIRE "
"SCAN of file %s",dataSize,m_bf.getFilename());
goto nextFile;
}
//gr = &msg7->m_gr;
//XmlDoc *xd = getAvailXmlDoc();
//msg7 = getAvailMsg7();
mcast = getAvailMulticast();
// if none, must have to wait for some to come back to us
if ( ! mcast ) {
// restore file offset
//m_fileOffset = saved;
// no, must have been a oom or something
log("import: import no mcast available");
return true;//false;
}
// this is for holding a compressed titlerec
//sbuf = &mcast->m_sbuf;//&gr->m_sbuf;
// point to start of buf
sbuf->reset();
// ensure we have enough room
sbuf->reserve ( need );
// collnum first 4 bytes
sbuf->pushLong( (long)m_collnum );
// store title key
sbuf->safeMemcpy ( &tkey , sizeof(key_t) );
// then datasize if any. neg rec will have -1 datasize
if ( dataSize >= 0 )
sbuf->pushLong ( dataSize );
// then read data rec itself into it, compressed titlerec part
if ( dataSize > 0 ) {
// read in the titlerec after the key/datasize
status = m_bf.read ( sbuf->getBuf() ,
dataSize ,
m_fileOffset );
if ( g_errno ) { // n != dataSize ) {
log("main: failed to read in title rec "
"file. %s. Skipping file %s",
mstrerror(g_errno),m_bf.getFilename());
// essentially free up this msg7 now
//msg7->m_inUse = false;
//msg7->reset();
goto nextFile;
}
// advance
m_fileOffset += dataSize;
// it's good, count it
sbuf->m_length += dataSize;
}
// set xmldoc from the title rec
//xd->set ( sbuf.getBufStart() );
//xd->m_masterState = NULL;
//xd->m_masterCallback ( titledbInjectLoop );
// we use this so we know where the doc we are injecting
// was in the foregien titledb file. so we can update our bookmark
// code.
mcast->m_hackFileOff = saved;//m_fileOffset;
mcast->m_hackFileId = m_bfFileId;
//
// inject a title rec buf this time, we are doing an import
// FROM A TITLEDB FILE!!!
//
//gr->m_titleRecBuf = &sbuf;
// break it down into gw
// xd.set2 ( sbuf.getBufStart() ,
// sbuf.length() , // max size
// cr->m_coll, // use our coll
// NULL , // pbuf for page parser
// 1 , // niceness
// NULL ); //sreq );
// // note it
// log("import: importing %s",xd.m_firstUrl.getUrl());
// now we can set gr for the injection
// TODO: inject the whole "sbuf" so we get sitenuminlinks etc
// all exactly the same...
// gr->m_url = xd.getFirstUrl()->getUrl();
// gr->m_queryToScrape = NULL;
// gr->m_contentDelim = 0;
// gr->m_contentTypeStr = g_contentTypeStrings [xd.m_contentType];
// gr->m_contentFile = NULL;
// gr->m_content = xd.ptr_utf8Content;
// gr->m_diffbotReply = NULL;
// gr->m_injectLinks = false;
// gr->m_spiderLinks = true;
// gr->m_shortReply = false;
// gr->m_newOnly = false;
// gr->m_deleteUrl = false;
// gr->m_recycle = true; // recycle content? or sitelinks?
// gr->m_dedup = false;
// gr->m_hasMime = false;
// gr->m_doConsistencyTesting = false;
// gr->m_getSections = false;
// gr->m_gotSections = false;
// gr->m_charset = xd.m_charset;
// gr->m_hopCount = xd.m_hopCount;
//
// point to next doc in the titledb file
//
//m_fileOffset += need;
// get docid from key
docId = g_titledb.getDocIdFromKey ( &tkey );
// get shard that holds the titlerec for it
shardNum = g_hostdb.getShardNumFromDocId ( docId );
// for selecting which host in the shard receives it
key = (long)docId;
m_numOut++;
// then index it. master callback will be called
//if ( ! xd->index() ) return false;
// TODO: make this forward the request to an appropriate host!!
// . gr->m_sbuf is set to the titlerec so this should handle that
// and use XmlDoc::set4() or whatever
// if ( msg7->injectTitleRec ( msg7 , // state
// gotMsg7ReplyWrapper , // callback
// cr )) {
// // it didn't block somehow...
// msg7->m_inUse = false;
// msg7->gotMsg7Reply();
// }
req = sbuf->getBufStart();
reqSize = sbuf->length();
if ( reqSize != need ) { char *xx=NULL;*xx=0 ; }
// do not free it, let multicast free it after sending it
sbuf->detachBuf();
if ( ! mcast->send ( req ,
reqSize ,
0x07 ,
true , // ownmsg?
shardNum,
false, // send to whole shard?
key , // for selecting host in shard
mcast , // state
NULL , // state2
gotMulticastReplyWrapper ,
999999 ) ) { // total timeout in seconds
log("import: import mcast had error: %s",mstrerror(g_errno));
m_numIn++;
}
goto INJECTLOOP;
nextFile:
// invalidate this flag
//m_offIsValid = false;
// . and call this function. we add one to m_bfFileId so we
// do not re-get the file we just injected.
// . sets m_bf and m_fileOffset
// . returns false if nothing to read
if ( ! setCurrentTitleFileAndOffset ( ) ) { //cr , m_bfFileId+1 );
log("import: import: no files left to read");
//goto INJECTLOOP;
return true;
}
// if it returns NULL we are done!
log("main: titledb injection loop completed. waiting for "
"outstanding injects to return.");
if ( m_numOut > m_numIn )
return false;
log("main: all injects have returned. DONE.");
// dummy return
return true;
}
bool sendPageThesaurus( TcpSocket *s, HttpRequest *r ) {
SafeBuf p;
char getBuf[64]; // holds extra values for GET method
char formBuf[256]; // holds extra values for forms
snprintf(getBuf, 64, "c=%s",
r->getString("c", 0, ""));
snprintf(formBuf, 256,
"<input type=hidden name=\"c\" value=\"%s\">",
//"<input type=hidden name=\"pwd\" value=\"%s\">",
r->getString("c", 0, ""));
g_pages.printAdminTop( &p, s, r);
if (r->getLong("cancel", 0) != 0) {
g_thesaurus.cancelRebuild();
p.safePrintf("<br><br>\n");
p.safePrintf(
"<center><b><font color=#ff0000>"
"rebuild canceled"
"</font></b></center>");
}
if (r->getLong("rebuild", 0) != 0) {
bool full = r->getLong("full", 0);
p.safePrintf("<br><br>\n");
if (g_thesaurus.rebuild(0, full)) {
p.safePrintf(
"<center><b><font color=#ff0000>"
"error starting rebuild, check log for details"
"</font></b></center>");
} else {
p.safePrintf(
"<center><b><font color=#ff0000>"
"rebuild started"
"</font></b></center>");
}
}
if (r->getLong("rebuildaff", 0) != 0) {
bool full = r->getLong("full", 0);
p.safePrintf("<br><br>\n");
if (g_thesaurus.rebuildAffinity(0, full)) {
p.safePrintf(
"<center><b><font color=#ff0000>"
"error starting rebuild, check log for details"
"</font></b></center>");
} else {
p.safePrintf(
"<center><b><font color=#ff0000>"
"rebuild started"
"</font></b></center>");
}
}
if (r->getLong("distribute", 0) != 0) {
char cmd[1024];
p.safePrintf("<br><br>\n");
if (g_thesaurus.m_affinityState) {
p.safePrintf(
"<center><b><font color=#ff0000>"
"cannot distribute during rebuild"
"</font></b></center>");
} else {
for ( long i = 0; i < g_hostdb.getNumHosts() ; i++ ) {
Host *h = g_hostdb.getHost(i);
snprintf(cmd, 512,
"rcp -r "
"./dict/thesaurus.* "
"%s:%s/dict/ &",
iptoa(h->m_ip),
h->m_dir);
log(LOG_INFO, "admin: %s", cmd);
system( cmd );
}
p.safePrintf(
"<center><b><font color=#ff0000>"
"data distributed"
"</font></b></center>");
}
}
if (r->getLong("reload", 0) != 0) {
p.safePrintf("<br><br>\n");
if (r->getLong("cast", 0) != 0) {
p.safePrintf(
"<center><b><font color=#ff0000>"
"reload command broadcast"
"</font></b></center>");
} else if (g_thesaurus.init()) {
p.safePrintf(
"<center><b><font color=#ff0000>"
"thesaurus data reloaded"
"</font></b></center>");
} else {
p.safePrintf(
"<center><b><font color=#ff0000>"
"error reloading thesaurus data"
"</font></b></center>");
}
}
long manualAddLen = 0;
char *manualAdd = NULL;
SafeBuf manualAddBuf;
if ((manualAdd = r->getString("manualadd", &manualAddLen))) {
trimWhite(manualAdd);
manualAddLen = gbstrlen(manualAdd);
File manualFile;
manualFile.set(g_hostdb.m_dir, "dict/thesaurus-manual.txt");
if (manualFile.open(O_WRONLY | O_CREAT | O_TRUNC) &&
(manualFile.write(manualAdd, manualAddLen, 0) ==
manualAddLen)) {
char newl = '\n'; // for write()
if (manualAdd[manualAddLen-1] != '\n')
manualFile.write(&newl, 1, manualAddLen);
p.safePrintf(
"<center><b><font color=#ff0000>"
"updated manual add file sucessfully"
"</font></b></center>");
} else {
p.safePrintf(
"<center><b><font color=#ff0000>"
"error writing manual add file"
"</font></b></center>");
}
} else {
char ff[PATH_MAX];
snprintf(ff, PATH_MAX, "%sdict/thesaurus-manual.txt",
g_hostdb.m_dir);
if (manualAddBuf.fillFromFile(ff)) {
if (*(manualAddBuf.getBuf()-1) != '\n')
manualAddBuf.pushChar('\n');
manualAdd = manualAddBuf.getBufStart();
manualAddLen = manualAddBuf.length();
}
}
long affinityAddLen = 0;
char *affinityAdd = NULL;
SafeBuf affinityAddBuf;
if ((affinityAdd = r->getString("affinityadd", &affinityAddLen))) {
trimWhite(affinityAdd);
affinityAddLen = gbstrlen(affinityAdd);
File affinityFile;
affinityFile.set(g_hostdb.m_dir,
"dict/thesaurus-affinity.txt");
if (affinityFile.open(O_WRONLY | O_CREAT | O_TRUNC) &&
(affinityFile.write(affinityAdd, affinityAddLen, 0) ==
affinityAddLen)) {
char newl = '\n'; // for write()
if (affinityAdd[affinityAddLen-1] != '\n')
affinityFile.write(&newl, 1, affinityAddLen);
p.safePrintf(
"<center><b><font color=#ff0000>"
"updated affinity add file sucessfully"
"</font></b></center>");
} else {
p.safePrintf(
"<center><b><font color=#ff0000>"
"error writing affinity add file"
"</font></b></center>");
}
} else {
char ff[PATH_MAX];
snprintf(ff, PATH_MAX, "%sdict/thesaurus-affinity.txt",
g_hostdb.m_dir);
if (affinityAddBuf.fillFromFile(ff)) {
if (*(affinityAddBuf.getBuf()-1) != '\n')
affinityAddBuf.pushChar('\n');
affinityAdd = affinityAddBuf.getBufStart();
affinityAddLen = affinityAddBuf.length();
}
}
char *syn = r->getString("synonym");
long len = 0;
if (syn) len = gbstrlen(syn);
if (len) {
SynonymInfo info;
bool r = g_thesaurus.getAllInfo(syn, &info, len, SYNBIT_ALL);
p.safePrintf("<br><br>\n");
p.safePrintf (
"<table cellpadding=4 width=100%% bgcolor=#%s border=1>"
"<tr>"
"<td colspan=2 bgcolor=#%s>"
"<center><b>Synonym List (%ld)</b></center>"
"</td>"
"</tr>\n",
LIGHT_BLUE, DARK_BLUE, info.m_numSyns);
if (r) {
p.safePrintf("<tr>"
"<td align=right><tt>%s</tt></td>"
"<td align=left>"
"<tt>1.000/%08lX (1.000/%08lX)</tt>"
"</td>"
"</tr>\n", syn, MAX_AFFINITY, MAX_AFFINITY);
for (long i = 0; i < info.m_numSyns; i++) {
// get the reverse affinity as well
long aff = g_thesaurus.getAffinity(
info.m_syn[i], syn,
info.m_len[i], len);
p.safePrintf(
"<tr>"
"<td width=40%% align=right>"
"<tt>");
p.safeMemcpy(info.m_syn[i], info.m_len[i]);
p.safePrintf("</tt>"
"</td>"
"<td width=60%% align=left>"
"<tt>");
if (info.m_affinity[i] >= 0) {
p.safePrintf("%0.3f/%08lX ",
(float)info.m_affinity[i]
/ MAX_AFFINITY,
info.m_affinity[i]);
} else {
p.safePrintf("u ");
}
if (aff >= 0) {
p.safePrintf("(%0.3f/%08lX) ",
(float)aff / MAX_AFFINITY,
aff);
} else {
p.safePrintf("(u) ");
}
p.safePrintf("(%ld) (%ld) (%ld) (%ld) "
"(%lld) (%lld)",
(long)info.m_type[i], (long)info.m_sort[i],
info.m_firstId[i], info.m_lastId[i],
info.m_leftSynHash[i],
info.m_rightSynHash[i]);
for (int j = info.m_firstId[i];
j <= info.m_lastId[i];
j++) {
p.safePrintf(" (%lld)",
info.m_termId[j]);
}
p.safePrintf(
"</tt>"
"</td>"
"</tr>\n");
}
p.safePrintf("</table>");
} else {
p.safePrintf("<tr>"
"<td align=center><font color=#FF0000>"
"synonym not found: %s"
"</font></td>"
"</tr>\n",
syn);
}
}
p.safePrintf ( "<br><br>\n" );
p.safePrintf (
"<table cellpadding=4 width=100%% bgcolor=#%s border=1>"
"<tr>"
"<td colspan=2 bgcolor=#%s>"
"<center><b>Thesaurus Controls"
"</b></center></td>"
"</tr>\n",
LIGHT_BLUE, DARK_BLUE);
p.safePrintf (
"<tr>"
"<td width=37%%><b>rebuild all data</b><br>"
"<font size=1>"
"rebuilds synonyms and then begins the rebuild process for "
"affinity data; this should only be run on one host, as the "
"data is copied when the process is finished; full rebuild "
"does not use existing affinity data"
"</font>"
"</td>"
"<td width=12%% bgcolor=#0000ff>"
"<center><b><a href=\"/master/thesaurus?rebuild=1&%s\">"
"rebuild all data</a> <a href=\"/master/thesaurus?"
"rebuild=1&full=1&%s\">(full)</a></b></center>"
"</td>"
"</tr>\n", getBuf, getBuf);
p.safePrintf (
"<tr>"
"<td width=37%%><b>distribute data</b><br>"
"<font size=1>"
"distributes all thesaurus data to all hosts, this is "
"normally done automatically but if there was a problem "
"with the copy, this lets you do it manually"
"</font>"
"</td>"
"<td width=12%% bgcolor=#0000ff>"
"<center><b><a href=\"/master/thesaurus?distribute=1&%s\">"
"distribute data</a></b></center>"
"</td>"
"</tr>\n", getBuf);
p.safePrintf (
"<tr>"
"<td width=37%%><b>reload data</b><br>"
"<font size=1>"
"reloads the synonyms and affinity table on this host only"
"</font>"
"</td>"
"<td width=12%% bgcolor=#0000ff>"
"<center><b>"
"<a href=\"/master/thesaurus?reload=1&cast=0&%s\">"
"reload data</a></b></center>"
"</td>"
"</tr>\n", getBuf);
p.safePrintf (
"<tr>"
"<td width=37%%><b>reload data (all hosts)</b><br>"
"<font size=1>"
"reloads the synonyms and affinity table on all hosts"
"</font>"
"</td>"
"<td width=12%% bgcolor=#0000ff>"
"<center><b>"
"<a href=\"/master/thesaurus?reload=1&cast=1&%s\">"
"reload data (all hosts)</a></b></center>"
"</td>"
"</tr>\n", getBuf);
p.safePrintf (
"<tr>"
"<td width=37%%><b>list synonyms</b><br>"
"<font size=1>"
"enter a word here to list all synonym entries and their "
"affinities"
"</font>"
"</td>"
"<td width=12%%>"
"<form action=\"/master/thesaurus>\">"
"<input type=text name=synonym size=20>"
"<input type=submit value=Submit>"
"%s"
"</form></td>"
"</tr>\n", formBuf);
p.safePrintf (
"<tr>"
"<td colspan=2 bgcolor=#%s>"
"<center><b>Affinity Controls"
"</b></center></td>"
"</tr>\n",
DARK_BLUE);
p.safePrintf (
"<tr>"
"<td width=37%%><b>cancel running rebuild</b><br>"
"<font size=1>"
"cancels the rebuild and throws all intermediate data away"
"</font>"
"</td>"
"<td width=12%% bgcolor=#0000ff>"
"<center><b><a href=\"/master/thesaurus?cancel=1&%s\">"
"cancel running rebuild</a></b></center>"
"</td>"
"</tr>\n", getBuf);
p.safePrintf (
"<tr>"
"<td width=37%%><b>rebuild affinity only</b><br>"
"<font size=1>"
"begins the rebuild process for affinity data, has no "
"effect if a rebuild is already in progress; full rebuild "
"does not reuse existing affinity data"
"</font>"
"</td>"
"<td width=12%% bgcolor=#0000ff>"
"<center><b><a href=\"/master/thesaurus?rebuildaff=1&%s\">"
"rebuild affinity</a> <a href=\"/master/thesaurus?"
"rebuildaff=1&full=1&%s\">(full)</a></b></center>"
"</td>"
"</tr>\n", getBuf, getBuf);
p.safePrintf (
"<tr>"
"<td colspan=2 bgcolor=#%s>"
"<center><b>Manual File Controls"
"</b></td>"
"</tr>\n",
DARK_BLUE);
p.safePrintf (
"<tr>"
"<td align=center colspan=2>");
p.safePrintf(
"<b>manually added pairs</b><br>\n"
"<font size=1>place word pairs here that should be linked "
"as synonyms, one pair per line, seperated by a pipe '|' "
"character, optionally followed by another pipe and a type "
"designation; any badly formatted lines will be silently "
"ignored</font><br>\n"
"<form action=\"/master/thesaurus\" method=post>"
"<textarea name=\"manualadd\" rows=20 cols=80>");
if (manualAdd && manualAddLen) {
p.htmlEncode(manualAdd, manualAddLen, true);
}
p.safePrintf (
"</textarea><br>"
"<input type=submit value=Submit>"
"<input type=reset value=Reset>"
"%s"
"</form></td>"
"</tr>\n",
formBuf);
p.safePrintf (
"<tr>"
"<td align=center colspan=2>"
"<b>affinity value overrides</b><br>\n"
"<font size=1>place word/phrase pairs here that should have "
"there affinity values overridden, format is "
"\"word1|word2|value\", where value is a floating point, "
"integer (either decimal or hex), or the word \"max\"; "
"any badly formatted lines will be silently ignored; note "
"that these pairs will only work if the thesaurus otherwise "
"has an entry for them, so add them to the manual add file "
"above if need be</font><br>\n"
"<form action=\"/master/thesaurus\" method=post>"
"<textarea name=\"affinityadd\" rows=20 cols=80>");
if (affinityAdd && affinityAddLen) {
p.htmlEncode(affinityAdd, affinityAddLen, true);
}
p.safePrintf (
"</textarea><br>"
"<input type=submit value=Submit>"
"<input type=reset value=Reset>"
"%s"
"</form></td>"
"</tr>\n",
formBuf);
p.safePrintf ( "</table>\n" );
p.safePrintf ( "<br><br>\n" );
p.safePrintf (
"<table cellpadding=4 width=100%% bgcolor=#%s border=1>"
"<tr>"
"<td colspan=2 bgcolor=#%s>"
"<center><b>Affinity Builder Status"
"</b></td>"
"</tr>\n",
LIGHT_BLUE, DARK_BLUE);
long long a, b, c, d, e, f, g, h, i, j, k;
StateAffinity *aff = g_thesaurus.m_affinityState;
if (!aff) {
p.safePrintf (
"<tr><td colspan=2>"
"<center><b>Not running</b></center>"
"</td></tr>\n");
a = b = c = d = e = f = g = h = i = j = k = 0;
} else {
a = aff->m_oldTable->getNumSlotsUsed();
b = aff->m_oldTable->getNumSlotsUsed() - aff->m_n;
c = aff->m_n;
d = (gettimeofdayInMilliseconds() - aff->m_time) / 1000;
if (!d || !(c / d)) {
e = 0;
} else {
e = b / (c / d);
}
f = aff->m_sent;
g = aff->m_recv;
h = aff->m_errors;
i = aff->m_old;
j = aff->m_cache;
k = aff->m_hitsTable.getNumSlotsUsed();
}
p.safePrintf (
"<tr><td><b># of total pairs</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of pairs remaining</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of pairs processed</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b>elapsed time in seconds</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b>estimated remaining time in seconds</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of requests sent</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of requests received</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of request errors</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of old values reused</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b># of cache hits</b></td>"
"<td>%lli</td></tr>\n"
"<tr><td><b>cache size</b></td>"
"<td>%lli</td></tr>\n",
a, b, c, d, e, f, g, h, i, j, k);
p.safePrintf ( "</table>\n" );
return g_httpServer.sendDynamicPage ( s, p.getBufStart(), p.length() );
}
// this should be called when all docs have finished spidering
void Test::stopIt ( ) {
// sanity
if ( m_isAdding ) { char *xx=NULL;*xx=0; }
// flag that we are done
m_isRunning = false;
// print time
log("test: took %lli ms to complete injections.",
gettimeofdayInMilliseconds() - m_testStartTime );
// get this before setting testParserEnabled to false
char *testDir = g_test.getTestDir();
// turn this off now too
g_conf.m_testParserEnabled = false;
g_conf.m_testSpiderEnabled = false;
// save all!
bool disabled = g_threads.m_disabled;
g_threads.disableThreads();
// save it blocking style
g_process.save();
if ( ! disabled ) g_threads.enableThreads();
// save ips.txt
saveTestBuf ( testDir );
log("test: test completed. making qa.html");
//
//
// NOW MAKE THE qa.html FILE
//
//
// only analyze up to last 7 runs
long start = m_runId - 7;
if ( start < 0 ) start = 0;
SafeBuf sb;
sb.safePrintf("<table border=1>\n");
sb.safePrintf("<tr>"
"<td><b><nobr>run id</nobr></b></td>"
"<td><b><nobr>conf diff</nobr></b></td>"
"<td><b><nobr>coll diff</nobr></b></td>"
"<td><b><nobr>run info</nobr></b></td>"
"</tr>\n");
// take diffs between this run and the last run for confparms
for ( long i = m_runId ; i > start ; i-- ) {
// shortcut
char *dir = g_hostdb.m_dir;
// make diff filename
char diff1[200];
sprintf(diff1,"%s/%s/run.%li.confparms.txt.diff",dir,
testDir,i);
File f1;
f1.set(diff1);
if ( ! f1.doesExist() ) {
char df1[200];
char df2[200];
sprintf(df1,"%s/%s/run.%li.confparms.txt",dir,
testDir,i);
sprintf(df2,"%s/%s/run.%li.confparms.txt",dir,
testDir,i-1);
// do the diff
char cmd[600];
sprintf(cmd,"diff %s %s > %s",df1,df2,diff1);
log("test: system(\"%s\")",cmd);
system (cmd);
}
long fs1 = f1.getFileSize();
sb.safePrintf("<tr><td>%li</td><td>%li</td>", i,fs1);
// make diff filename
char diff2[200];
sprintf(diff2,"%s/%s/run.%li.collparms.txt.diff",dir,
testDir,i);
File f2;
f2.set(diff2);
if ( ! f2.doesExist() ) {
char df1[200];
char df2[200];
sprintf(df1,"%s/%s/run.%li.collparms.txt",dir,
testDir,i);
sprintf(df2,"%s/%s/run.%li.collparms.txt",dir,
testDir,i-1);
// do the diff
char cmd[600];
sprintf(cmd,"diff %s %s > %s",df1,df2,diff2);
log("test: system(\"%s\")",cmd);
system (cmd);
}
long fs2 = f2.getFileSize();
sb.safePrintf("<td>%li</td>", fs2);
// the version
char vf[200];
sprintf(vf,"%s/%s/run.%li.version.txt",dir,testDir,i);
File f3;
f3.set ( vf );
long fs3 = f3.getFileSize();
char vbuf[1000];
vbuf[0] = 0;
if ( fs3 > 0 ) {
f3.open(O_RDONLY);
long rs = f3.read(vbuf,fs3,0);
vbuf[fs3] = '\0';
if ( rs <= 0 ) continue;
f3.close();
}
// show it
sb.safePrintf("<td><pre>%s</pre></td></tr>\n", vbuf);
}
sb.safePrintf("</table>\n");
sb.safePrintf("<br>\n");
//
// now diff each parser output file for each url in urls.txt
//
//
// loop over url buf first so we can print one table per url
//
char *next = NULL;
// reset the url buf ptr
m_urlPtr = m_urlBuf;
// count em
long count = 0;
// ptrs to each url table
long un = 0;
long uptr [5000]; // offsets now, not char ptr since buf gets reallocd
char udiff[5000];
long ulen [5000];
long uhits[5000]; // critical errors! validateOutput() choked!
long uunchecked[5000]; // events/addresses found but were not validatd
long umiss[5000];
long usort[5000];
long uevents[5000];
SafeBuf tmp;
long niceness = MAX_NICENESS;
// advance to next url
for ( ; m_urlPtr < m_urlEnd ; m_urlPtr = next ) {
// breathe
QUICKPOLL(niceness);
// we converted all non-url chars into \0's so skip those!
for ( ; m_urlPtr<m_urlEnd && !*m_urlPtr ; m_urlPtr++ );
// breach check
if ( m_urlPtr >= m_urlEnd ) break;
// set this up
next = m_urlPtr;
// compute next url ptr
for ( ; next < m_urlEnd && *next ; next++ );
// point to this url
char *u = m_urlPtr;
// get hash
long long h = hash64 ( u , gbstrlen(u) );
// shortcut
char *dir = g_hostdb.m_dir;
// print into a secondary safe buf with a ptr to
// it so we can sort that and transfer into the
// primary safebuf later
uptr[un] = tmp.length();
// assume no diff
udiff[un] = 0;
// print number
tmp.safePrintf("%li) ",count++);
// . link to our stored http server reply
// . TODO: link it to our [cached] copy in the test coll!!!
char local[1200];
sprintf(local,"/%s/doc.%llu.html",testDir,h);
tmp.safePrintf("<a href=\"%s\"><b>%s</b></a> ",local,u);
// link to live page
tmp.safePrintf(" <a href=\"%s\">live</a> ",u);
// link to page parser
char ubuf[2000];
urlEncode(ubuf,2000,u,gbstrlen(u),true);
tmp.safePrintf(" <a href=\"/master/parser?c=test&"
"u=%s\">parser</a> ",ubuf);
//tmp.safePrintf(" (%llu)",h);
tmp.safePrintf("<br>\n");
//tmp.safePrintf("<br>\n");
tmp.safePrintf("<table border=1>\n");
tmp.safePrintf("<tr>"
"<td><b><nobr>run id</nobr></b></td>"
"<td><b><nobr>crit hits</nobr></b></td>"
"<td><b><nobr>crit errors</nobr></b></td>"
"<td><b><nobr># e</nobr></b></td>"
"<td><b><nobr>unchecked</nobr></b></td>"
"<td><b><nobr>diff chars</nobr></b></td>"
"<td><b><nobr>diff file</nobr></b></td>"
"<td><b><nobr>full output</nobr></b></td>"
"</tr>\n");
//SafeBuf sd;
// loop over all the runs now, starting with latest run first
for ( long ri = m_runId ; ri >= start ; ri-- ) {
QUICKPOLL(niceness);
// the diff filename
char pdiff[200];
sprintf(pdiff,"%s/%s/parse.%llu.%li.html.diff",dir,
testDir,h,ri);
File f;
f.set(pdiff);
long fs = f.getFileSize();
if ( ! f.doesExist() && ri > 0 ) {
// make the parse filename
char pbuf1[200];
char pbuf2[200];
sprintf(pbuf1,"%s/%s/parse.%llu.%li.html",
dir,testDir,h,ri);
sprintf(pbuf2,"%s/%s/parse.%llu.%li.html",
dir,testDir,h,ri-1);
// sanity check
//File tf; tf.set(pbuf1);
//if ( ! tf.doesExist()) {char *xx=NULL;*xx=0;}
// tmp file name
char tmp1[200];
char tmp2[200];
sprintf(tmp1,"%s/%s/t1.html",dir,testDir);
sprintf(tmp2,"%s/%s/t2.html",dir,testDir);
// filter first
char cmd[600];
sprintf(cmd,
"cat %s | "
"grep -v \"<!--ignore-->\" "
" > %s", pbuf1,tmp1);
system(cmd);
sprintf(cmd,
"cat %s | "
"grep -v \"<!--ignore-->\" "
" > %s", pbuf2,tmp2);
system(cmd);
// make the system cmd to do the diff
sprintf(cmd,
"echo \"<pre>\" > %s ; "
"diff -w --text %s %s "
// ignore this table header row
//" | grep -v \"R#4\""
" >> %s",
pdiff,
tmp1,tmp2,pdiff);
log("test: system(\"%s\")",cmd);
system(cmd);
// try again
f.set(pdiff);
fs = f.getFileSize();
}
QUICKPOLL(niceness);
// this means 0 . it just has the <pre> tag in it!
if ( fs < 0 || fs == 6 ) fs = 0;
// . if no diff and NOT current run, do not print it
// . print it if the run right before the current
// now always too
if ( ri != m_runId && ri != m_runId-1 && fs == 0 )
continue;
// relative filename
char rel[200];
sprintf(rel,"/%s/parse.%llu.%li.html.diff",
testDir,h,ri);
char full[200];
sprintf(full,"/%s/parse.%llu.%li.html",
testDir,h,ri);
char validate[200];
sprintf(validate,
"/%s/parse-shortdisplay.%llu.%li.html",
testDir,h,ri);
// use red font for current run that has a diff!
char *t1 = "";
char *t2 = "";
if ( ri == m_runId && fs != 0 ) {
t1 = "<font color=pink><b>";
t2 = "</b></font>";
// a diff
udiff[un] = 1;
}
// . get critical errors
// . i.e. XmlDoc::validateOutput() could not validate
// a particular event or address that was in the
// url's "validated.uh64.txt" file since the admin
// clicked on the checkbox in the page parser output
// . if we do not find such a tag in the parser output
// any more then Spider.cpp creates this file!
if ( ri == m_runId ) {
char cfile[256];
sprintf(cfile,"%s/%s/critical.%llu.%li.txt",
g_hostdb.m_dir,testDir,h,ri);
SafeBuf ttt;
ttt.fillFromFile(cfile);
// first long is misses, then hits then events
umiss[un] = 0;
uhits[un] = 0;
uevents[un] = 0;
uunchecked[un] = 0;
if ( ttt.length() >= 3 )
sscanf(ttt.getBufStart(),
"%li %li %li %li",
&umiss[un],
&uhits[un],
&uevents[un],
&uunchecked[un]);
usort[un] = umiss[un] + uunchecked[un];
//File cf;
//cf.set(cfile);
//if ( cf.doesExist()) ucrit[un] = 1;
//else ucrit[un] = 0;
}
// more critical?
if ( ri == m_runId && umiss[un] != 0 ) {
t1 = "<font color=red><b>";
t2 = "</b></font>";
}
// . these are good to have
// . if you don't have 1+ critical hits then you
// probably need to be validate by the qa guy
char *uhb1 = "";
char *uhb2 = "";
if ( ri == m_runId && uhits[un] != 0 ) {
uhb1 = "<font color=green><b>**";
uhb2 = "**</b></font>";
}
QUICKPOLL(niceness);
char *e1 = "<td>";
char *e2 = "</td>";
long ne = uevents[un];
if ( ne ) {
e1="<td bgcolor=orange><b><font color=brown>";
e2="</font></b></td>";
}
char *u1 = "<td>";
char *u2 = "</td>";
if ( uunchecked[un] ) {
u1="<td bgcolor=purple><b><font color=white>";
u2="</font></b></td>";
}
// print the row!
tmp.safePrintf("<tr>"
"<td>%s%li%s</td>"
"<td>%s%li%s</td>" // critical hits
"<td>%s%li%s</td>" // critical misses
"%s%li%s" // # events
"%s%li%s" // unchecked
"<td>%s%li%s</td>" // filesize of diff
// diff filename
"<td><a href=\"%s\">%s%s%s</a></td>"
// full parser output
"<td>"
"<a href=\"%s\">full</a> | "
"<a href=\"%s\">validate</a> "
"</td>"
"</tr>\n",
t1,ri,t2,
uhb1,uhits[un],uhb2,
t1,umiss[un],t2,
e1,ne,e2,
u1,uunchecked[un],u2,
t1,fs,t2,
rel,t1,rel,t2,
full,
validate);
// only fill "sd" for the most recent guy
if ( ri != m_runId ) continue;
// now concatenate the parse-shortdisplay file
// to this little table so qa admin can check/uncheck
// validation checkboxes for addresses and events
//sprintf(cfile,
// "%s/test/parse-shortdisplay.%llu.%li.html",
// g_hostdb.m_dir,h,ri);
//sd.fillFromFile ( cfile );
}
// end table
tmp.safePrintf("</table>\n");
// . and a separate little section for the checkboxes
// . should already be in tables, etc.
// . each checkbox should provide its own uh64 when it
// calls senddiv() when clicked now
//tmp.cat ( sd );
tmp.safePrintf("<br>\n");
tmp.safePrintf("<br>\n");
// set this
ulen[un] = tmp.length() - uptr[un] ;
// sanity check
if ( ulen[un] > 10000000 ) { char *xx=NULL;*xx=0; }
// inc it
un++;
// increase the 5000!!
if ( un >= 5000 ) { char *xx=NULL; *xx=0; }
}
char flag ;
bubble:
flag = 0;
// sort the url tables
for ( long i = 0 ; i < un - 1 ; i++ ) {
QUICKPOLL(niceness);
if ( usort[i] > usort[i+1] ) continue;
if ( usort[i] == usort[i+1] )
if ( udiff[i] >= udiff[i+1] ) continue;
// swap em
long tp = uptr[i];
long td = udiff[i];
long um = umiss[i];
long us = usort[i];
long uh = uhits[i];
long tl = ulen [i];
uptr[i] = uptr[i+1];
umiss[i] = umiss[i+1];
usort[i] = usort[i+1];
uhits[i] = uhits[i+1];
udiff[i] = udiff[i+1];
ulen[i] = ulen[i+1];
uptr[i+1] = tp;
umiss[i+1] = um;
usort[i+1] = us;
uhits[i+1] = uh;
udiff[i+1] = td;
ulen [i+1] = tl;
flag = 1;
}
if ( flag ) goto bubble;
// transfer into primary safe buf now
for ( long i = 0 ; i < un ; i++ )
sb.safeMemcpy(tmp.getBufStart() + uptr[i],ulen[i]);
sb.safePrintf("</html>\n");
char dfile[200];
sprintf(dfile,"%s/%s/qa.html",g_hostdb.m_dir,testDir);
sb.dumpToFile ( dfile );
// free the buffer of urls
reset();
// turn off spiders
g_conf.m_spideringEnabled = 0;
// all done
return;
}
bool HttpMime::addCookieHeader(const char *cookieJar, const char *url, SafeBuf *sb) {
Url tmpUrl;
tmpUrl.set(url);
SafeBuf tmpSb;
size_t cookieJarLen = strlen(cookieJar);
const char *lineStartPos = cookieJar;
const char *lineEndPos = NULL;
while ((lineEndPos = (const char*)memchr(lineStartPos, '\n', cookieJarLen - (lineStartPos - cookieJar))) != NULL) {
const char *currentPos = lineStartPos;
const char *tabPos = NULL;
unsigned fieldCount = 0;
bool skipCookie = false;
const char *domain = NULL;
int32_t domainLen = 0;
while (fieldCount < 5 && (tabPos = (const char*)memchr(currentPos, '\t', lineEndPos - currentPos)) != NULL) {
switch (fieldCount) {
case 0:
// domain
domain = currentPos;
domainLen = tabPos - currentPos;
break;
case 1:
// flag
if (memcmp(currentPos, "TRUE", 4) == 0) {
// allow subdomain
if (tmpUrl.getHostLen() >= domainLen) {
if (!endsWith(tmpUrl.getHost(), tmpUrl.getHostLen(), domain, domainLen)) {
// doesn't end with domain - ignore cookie
skipCookie = true;
break;
}
} else {
skipCookie = true;
break;
}
} else {
// only specific domain
if (tmpUrl.getHostLen() != domainLen || strncasecmp(domain, tmpUrl.getHost(), domainLen) != 0) {
// non-matching domain - ignore cookie
skipCookie = true;
break;
}
}
break;
case 2: {
// path
const char *path = currentPos;
int32_t pathLen = tabPos - currentPos;
if (strncasecmp(path, tmpUrl.getPath(), pathLen) == 0) {
if (tmpUrl.getPathLen() != pathLen) {
if (path[pathLen - 1] != '/' && tmpUrl.getPath()[tmpUrl.getPathLen() - 1] != '/') {
// non-matching path - ignore cookie
skipCookie = true;
break;
}
}
} else {
// non-matching path - ignore cookie
skipCookie = true;
break;
}
} break;
case 3:
// secure
break;
case 4:
// expiration
break;
}
currentPos = tabPos + 1;
++fieldCount;
}
if (!skipCookie) {
tmpSb.safeMemcpy(currentPos, lineEndPos - currentPos);
tmpSb.pushChar(';');
}
lineStartPos = lineEndPos + 1;
}
// we don't need to care about the last line (we always end on \n)
if (tmpSb.length() > 0) {
sb->safeStrcpy("Cookie: ");
sb->safeMemcpy(&tmpSb);
sb->safeStrcpy("\r\n");
}
return true;
}