| 56 |
|
inline int unlink(const char* filename) { return DeleteFile(filename); } |
| 57 |
|
#endif // _WIN32 |
| 58 |
|
|
| 59 |
– |
Indexer::Indexer(string& indexFile, set<string>& domains, set<string>& |
| 60 |
– |
restrictions) |
| 61 |
– |
{ |
| 62 |
– |
this->indexFile = indexFile; |
| 63 |
– |
this->domains = domains; |
| 64 |
– |
this->restrictions = restrictions; |
| 65 |
– |
} |
| 66 |
– |
|
| 59 |
|
void Indexer::index(string& begin) |
| 60 |
|
{ |
| 61 |
< |
unsigned separator = indexFile.rfind(slash); |
| 62 |
< |
string dtd = separator != string::npos ? indexFile.substr(0, separator) + |
| 63 |
< |
slash + "index.dtd" : "index.dtd"; |
| 72 |
< |
|
| 61 |
> |
unsigned separator(indexFile.rfind(slash)); |
| 62 |
> |
string dtd(separator != string::npos ? indexFile.substr(0, separator) + |
| 63 |
> |
slash + "index.dtd" : "index.dtd"); |
| 64 |
|
ifstream fin(dtd.c_str()); |
| 65 |
|
|
| 66 |
|
if (!fin.is_open()) |
| 85 |
|
|
| 86 |
|
fin.close(); |
| 87 |
|
|
| 88 |
< |
string lock = indexFile + ".lock"; |
| 98 |
< |
|
| 88 |
> |
string lock(indexFile + ".lock"); |
| 89 |
|
ofstream fout(lock.c_str()); |
| 90 |
+ |
|
| 91 |
|
fout.close(); |
| 92 |
|
fout.open(indexFile.c_str()); |
| 93 |
|
|
| 98 |
|
URL first(begin); |
| 99 |
|
|
| 100 |
|
index(first, fout); |
| 101 |
+ |
|
| 102 |
+ |
while (!links.empty()) |
| 103 |
+ |
{ |
| 104 |
+ |
URL next(links.front()); |
| 105 |
+ |
string referer(referers.front()); |
| 106 |
+ |
|
| 107 |
+ |
links.pop(); |
| 108 |
+ |
referers.pop(); |
| 109 |
+ |
|
| 110 |
+ |
if (debug) cerr << "next = " << next << "\n"; |
| 111 |
+ |
|
| 112 |
+ |
index(next, fout, referer); |
| 113 |
+ |
} |
| 114 |
|
|
| 115 |
|
fout << "</index>\n"; |
| 116 |
|
|
| 119 |
|
unlink(lock.c_str()); |
| 120 |
|
} |
| 121 |
|
|
| 122 |
< |
void Indexer::index(URL& url, ofstream& fout, const string referer) |
| 122 |
> |
void Indexer::index(URL& url, ofstream& fout, const string& referer) |
| 123 |
|
{ |
| 124 |
|
if (domains.find(url.getAddress()) != domains.end() && |
| 125 |
|
pages.find(url.getURL()) == pages.end()) |
| 138 |
|
http.contentType().find("text/html") == 0) |
| 139 |
|
{ |
| 140 |
|
http.clear(); |
| 141 |
+ |
|
| 142 |
|
if (!http.handle(url, referer)) exit(1); |
| 143 |
|
|
| 144 |
|
cout << "Indexing " << url << " ... " << flush; |
| 145 |
|
|
| 146 |
|
if (processor.process(http, url)) |
| 147 |
|
{ |
| 148 |
< |
Page page = processor.getPage(); |
| 148 |
> |
Page page(processor.getPage()); |
| 149 |
> |
|
| 150 |
|
fout << page << "\n"; |
| 151 |
|
|
| 152 |
|
cout << "done.\n"; |
| 157 |
|
} |
| 158 |
|
|
| 159 |
|
pages.insert(url.getURL()); |
| 160 |
< |
Set pageLinks = processor.getLinks(); |
| 160 |
> |
|
| 161 |
> |
Set pageLinks(processor.getLinks()); |
| 162 |
> |
|
| 163 |
|
processor.reset(); |
| 164 |
|
|
| 165 |
< |
for (SetIterator link = pageLinks.begin(); link != |
| 165 |
> |
for (SetIterator link(pageLinks.begin()); link != |
| 166 |
|
pageLinks.end(); link++) |
| 167 |
|
{ |
| 168 |
|
if (pages.find(*link) == pages.end()) |
| 169 |
|
{ |
| 170 |
< |
links.push(URL(*link)); |
| 170 |
> |
links.push(*link); |
| 171 |
|
referers.push(url.getURL()); |
| 172 |
|
} |
| 173 |
|
} |
| 181 |
|
{ |
| 182 |
|
if (pages.find(http.redirect()) == pages.end()) |
| 183 |
|
{ |
| 184 |
< |
links.push(URL(http.redirect())); |
| 184 |
> |
links.push(http.redirect()); |
| 185 |
|
referers.push(url.getURL()); |
| 186 |
|
} |
| 187 |
|
} |
| 189 |
|
http.clear(); |
| 190 |
|
} |
| 191 |
|
} |
| 184 |
– |
|
| 185 |
– |
if (!links.empty()) |
| 186 |
– |
{ |
| 187 |
– |
URL next = links.front(); |
| 188 |
– |
links.pop(); |
| 189 |
– |
|
| 190 |
– |
string referer = referers.front(); |
| 191 |
– |
referers.pop(); |
| 192 |
– |
|
| 193 |
– |
if (debug) cerr << "next = " << next << "\n"; |
| 194 |
– |
|
| 195 |
– |
index(next, fout, referer); |
| 196 |
– |
} |
| 192 |
|
} |
| 193 |
|
|
| 194 |
|
bool Indexer::restricted(URL& url) |
| 195 |
|
{ |
| 196 |
< |
bool answer = false; |
| 196 |
> |
bool answer(false); |
| 197 |
|
|
| 198 |
< |
for (SetIterator itor = restrictions.begin(); itor != restrictions.end(); |
| 198 |
> |
for (SetIterator itor(restrictions.begin()); itor != restrictions.end(); |
| 199 |
|
itor++) |
| 200 |
|
{ |
| 201 |
< |
URL checker = *itor; |
| 201 |
> |
URL checker(*itor); |
| 202 |
|
|
| 203 |
|
if (url.getAddress() == checker.getAddress() && url.getPort() == |
| 204 |
|
checker.getPort()) |
| 206 |
|
if (url.getPath().find(checker.getPath()) == 0) |
| 207 |
|
{ |
| 208 |
|
answer = true; |
| 209 |
+ |
|
| 210 |
|
break; |
| 211 |
|
} |
| 212 |
|
} |
| 217 |
|
|
| 218 |
|
void Indexer::robots(URL& url) |
| 219 |
|
{ |
| 220 |
< |
URL robots = url; |
| 220 |
> |
URL robots(url); |
| 221 |
> |
|
| 222 |
|
robots.setPath("/robots.txt"); |
| 223 |
|
|
| 224 |
|
if (http.handle(robots)) |
| 226 |
|
cout << "Checking " << robots << " ... " << flush; |
| 227 |
|
|
| 228 |
|
string line; |
| 229 |
< |
|
| 230 |
< |
bool record = false, hasVersion = false, hasName = false, hasAll = |
| 234 |
< |
false; |
| 235 |
< |
robot state = none; |
| 229 |
> |
bool record(false), hasVersion(false), hasName(false), hasAll(false); |
| 230 |
> |
robot state(none); |
| 231 |
|
Set restrictionsVersion, restrictionsName, restrictionsAll; |
| 232 |
|
|
| 233 |
|
while (http.good()) |
| 234 |
|
{ |
| 235 |
|
http.getline(line); |
| 236 |
|
|
| 237 |
< |
unsigned comment = line.find('#'); |
| 237 |
> |
unsigned comment(line.find('#')); |
| 238 |
> |
|
| 239 |
|
if (comment != string::npos) line.erase(comment); |
| 240 |
|
|
| 241 |
|
if (line == "" && comment == string::npos) record = false; |
| 242 |
|
if (line == "") continue; |
| 243 |
|
|
| 244 |
< |
unsigned colon = line.find(':'); |
| 245 |
< |
|
| 246 |
< |
string field = line.substr(0, colon); |
| 251 |
< |
string value = line.substr(colon + 1); |
| 244 |
> |
unsigned colon(line.find(':')); |
| 245 |
> |
string field(line.substr(0, colon)); |
| 246 |
> |
string value(line.substr(colon + 1)); |
| 247 |
|
|
| 248 |
|
normalize(value); |
| 249 |
|
|
| 271 |
|
} |
| 272 |
|
else if (field == "Disallow" && record) |
| 273 |
|
{ |
| 274 |
< |
URL restriction = robots; |
| 274 |
> |
URL restriction(robots); |
| 275 |
> |
|
| 276 |
|
restriction.setPath(value); |
| 277 |
|
|
| 278 |
|
switch (state) |
| 286 |
|
case all: |
| 287 |
|
restrictionsAll.insert(restriction.getURL()); |
| 288 |
|
break; |
| 293 |
– |
default: |
| 294 |
– |
break; |
| 289 |
|
} |
| 290 |
|
} |
| 291 |
|
} |
| 308 |
|
} |
| 309 |
|
|
| 310 |
|
SetIterator itor; |
| 311 |
+ |
|
| 312 |
|
switch (state) |
| 313 |
|
{ |
| 314 |
|
case version: |
| 326 |
|
} |
| 327 |
|
break; |
| 328 |
|
case all: |
| 329 |
< |
for (itor = restrictionsAll.begin(); itor != |
| 330 |
< |
restrictionsAll.end(); itor++) |
| 329 |
> |
for (itor = restrictionsAll.begin(); itor != restrictionsAll.end(); |
| 330 |
> |
itor++) |
| 331 |
|
{ |
| 332 |
|
restrictions.insert(*itor); |
| 333 |
|
} |
| 334 |
|
break; |
| 340 |
– |
default: |
| 341 |
– |
break; |
| 335 |
|
} |
| 336 |
|
|
| 337 |
|
cout << "done.\n"; |
