/* WIDE AREA INFORMATION SERVER SOFTWARE: No guarantees or restrictions. See the readme file for the full standard disclaimer. */ /*----------------------------------------------------------------------*/ /* This code implements a simple Z39.50+WAIS server, which consults a local database using Brewster's search engine. The main routine is interpret_buffer() which reads the contents of a receive buffer, and writes results back to a send buffer. The basic structure is: interpret_buffer gets bytes and returns bytes from whatever transport mechanism. It calls either handleInit (which handles init requests) or handleSearch. handleSearch calls either handleRelevanceFeedbackSearch or handleElementRetrieval based on the type of question. handleElementRetrieval calls getData or getDocumentText to answer that question. handleRelevanceFeedbackSearch calls run_search and/or help_header to answer the question. A server must supply: getData, getDocumentText, run_search, and help_header then it should work. To do: - help facilities: on a null query pass back random documents from the pool (?) - Add dates to search responses */ /* change log: * 3/91 fixed db name defaulting for info server * 5/31/91 fixed handleRelevanceFeedbackSearch to do search if * no seedwords but relevant document supplied - JG * 5/31/91 fixed databaseName in handleElementRetrieval - HWM * 7/19/91 fixed handleElementRetrieval prototype -BK */ /*----------------------------------------------------------------------*/ #include "ir.h" #include "wprot.h" #include "irsearch.h" #include "docid.h" #include "cutil.h" #include "irfiles.h" /* for pathname_name */ #include "irretrvl.h" #include "sockets.h" /* for connect_to_server */ #include #include #include #ifdef ANSI_LIKE #include #else #include "ustubs.h" #endif static void handleInit _AP((char** recBuf,char** sendBuf, long* sendBufLen, long* maxBufLen)); static void handleSearch _AP((char** recBuf,char** sendBuf, long* sendBufLen, long waisProtocolVersion, char *index_directory)); static void handleRelevanceFeedbackSearch _AP((SearchAPDU* search, char** sendBuf,long* sendBufLen, long waisProtocolVersion, char *index_directory)); static void handleElementRetrieval _AP((SearchAPDU* search, char** sendBuf, long* sendBufLen, long waisProtocolVersion, char *index_directory)); /*----------------------------------------------------------------------*/ /* Utility */ /*----------------------------------------------------------------------*/ /* note - at present, it is not clear to me what the value of present-status is, and how it is to be interpreted. Furthermore, are our text retrieval queries considered presents, or are they searches? */ /*----------------------------------------------------------------------*/ /* interpret_buffer() char* receiveBuffer - buffer containing data to interpret long receiveBufLen - how much data is there char* sendBuffer - buffer to write results to long sendBufLen - how much space there is to write to long* maxBufferSize - see below long waisProtocolVersion - what version of the wias protocol is in use char *index_directory - the directory to find the indexes on a search maxBufferSize is a pointer to a per connection variable that contains the maximum size send/receive buffer to use. Seems a lot like sendBufLen does't it. Well it usually is, but not always. Here is how it works from a server's point of view. When the client connection is first established, the server spawns a new process to deal with it. The new process contains a global variable (bufferSize in server.c) which is initialized to BUFSZ (defined in server.c = 30000). This is the physical size of the server's internal bufferes. Clearly that is the absolute maximum size of any z3950 message to or from this server. So now *maxBufferSize = sendBufLen. Now, the first thing that a z3950 client is likely to do is send an init APDU. The only useful thing (and it is useful) that the init APDU currently does is allow the client and server to negotiate the maxumum size of the messages that they will send. This takes place somewhere down inside of the interpret_buffer() logic where the APDU's are decoded and response APDU's are recoded. A pointer to bufferSize is passed to interpret_buffer() in the maxBufferSize argument, and if the buffer happens to contain an init message, bufferSize is changed (for the rest of the connection). That is the only function maxBufferSize serves. Note that I could have gotten rid of sendBufLen, and just used *maxBufferSize, but sendBufLen can be and does get modified by the z3950 APDU writting code, and we don't want the overall value being modified. */ long interpret_buffer(receiveBuffer, receiveBufLen, sendBuffer, sendBufLen, maxBufferSize, waisProtocolVersion, index_directory) char* receiveBuffer; long receiveBufLen; char* sendBuffer; long sendBufLen; long* maxBufferSize; long waisProtocolVersion; char *index_directory; /* read & interpret receiveBuffer until receiveBufLen. Write results into send buffer. Return number of bytes written - negative if there was an error */ { char* readPos = receiveBuffer; char* writePos = sendBuffer; while (readPos - receiveBuffer < receiveBufLen && /* there is more to read */ writePos != NULL /* no write error */ ) { pdu_type pdu = peekPDUType(readPos); switch (pdu) { case initAPDU: handleInit(&readPos,&writePos,&sendBufLen,maxBufferSize); break; case searchAPDU: handleSearch(&readPos,&writePos,&sendBufLen, waisProtocolVersion, index_directory); break; default: /* unknown APDU error */ writePos = NULL; break; } } return(writePos - sendBuffer); } /*----------------------------------------------------------------------*/ static void handleInit _AP((char** recBuf,char** sendBuf, long* sendBufLen,long* maxBufferSize)); static void handleInit(recBuf,sendBuf,sendBufLen,maxBufferSize) char** recBuf; char** sendBuf; long* sendBufLen; long* maxBufferSize; /* negotiate functionality and buffer sizes. A session ususally begins with one of these, but is not required to. NOTE - even if the server decides not to accept the client, it does not shut down the connection. It simply declies acceptatance, and waits for the client to shut down. */ { InitAPDU* anInit = NULL; /* read the init - note there is no WAIS protocol extension here */ *recBuf = readInitAPDU(&anInit,*recBuf); if (recBuf == NULL || *recBuf == NULL) { *sendBuf = NULL; /* error in the read */ return; } else /* respond to the init */ { InitResponseAPDU* reply = NULL; WAISInitResponse* wais_response = NULL; boolean connectionAccepted; /* negotiate services */ if (anInit->willPresent == false && anInit->willDelete == false) connectionAccepted = true; else connectionAccepted = false; /* negotiate buffer sizes */ if (*maxBufferSize > anInit->MaximumRecordSize) *maxBufferSize = anInit->MaximumRecordSize; if(anInit->IDAuthentication != NULL) waislog(WLOG_LOW, WLOG_INFO, "Init message: %s", anInit->IDAuthentication); /* not much use huh? */ wais_response = makeWAISInitResponse(0L,0L,NULL,NULL,NULL,NULL); reply = makeInitResponseAPDU(connectionAccepted, true,false,false,false,false,*maxBufferSize, *maxBufferSize,NULL, defaultImplementationID(), defaultImplementationName(), defaultImplementationVersion(),NULL, wais_response); /* write it */ *sendBuf = writeInitResponseAPDU(reply,*sendBuf,sendBufLen); /* free everything */ freeInitAPDU(anInit); freeInitResponseAPDU(reply); freeWAISInitResponse(wais_response); } } /*----------------------------------------------------------------------*/ static boolean isRemoteDB(db) char * db; { return(strchr(db,'@') != NULL); } /*----------------------------------------------------------------------*/ #include "wmessage.h" struct { char host[256]; long port; FILE *connection; } last_connection; static void forwardSearch(aSearch,sendBuf,sendBufLen,waisProtocolVersion) SearchAPDU* aSearch; char** sendBuf; long* sendBufLen; long waisProtocolVersion; { FILE *connection; char hostname[1000], db[1000], *p, *p2; long port, len, rlen; char message[30000], response[30000]; p = strchr(aSearch->DatabaseNames[0], '@'); strncpy(db, aSearch->DatabaseNames[0], p-aSearch->DatabaseNames[0]); db[p-aSearch->DatabaseNames[0]] = 0; p2 = strchr(p+1, ':'); if(p2 == NULL) { strcpy(hostname, p+1); port = 210; } else { strncpy(hostname, p+1, p2-(p+1)); hostname[p2-(p+1)] = 0; port = atoi(p2+1); } strcpy(aSearch->DatabaseNames[0], db); rlen = len = 30000; writeSearchAPDU(aSearch, message+HEADER_LENGTH, &len); len = 30000-len; if(hostname[0] != 0) { if(strcmp(hostname, last_connection.host) == 0 && port == last_connection.port) connection = last_connection.connection; else { if (last_connection.connection != NULL) close_connection(last_connection.connection); strcpy(last_connection.host, hostname); last_connection.port = port; last_connection.connection = (FILE*)connect_to_server(hostname, port); connection = last_connection.connection; if(connection != NULL) { char userInfo[500], hostname[80], init_message[1000]; gethostname(hostname, 80); #ifdef TELL_USER sprintf(userInfo, "server forwarding %s, from host: %s, user: %s", VERSION, hostname, getenv("USER")); #else sprintf(userInfo, "server forwarding %s, from host: %s", VERSION, hostname); #endif init_connection(init_message, response, 1000L, connection, userInfo); } } if(connection != NULL) { len = interpret_message(message, len, response, rlen, connection, false); } else { static diagnosticRecord* diags[2] = {NULL, NULL}; SearchResponseAPDU* response = NULL; WAISSearchResponse* wais_response = NULL; char message[255]; sprintf(message, "Database not available: %s@%s:%d.", db, last_connection.host, last_connection.port); diags[0] = makeDiag(true,D_RecordNotAuthorizedToBeSent, message); wais_response = makeWAISSearchResponse(NULL,NULL,NULL, NULL,NULL,NULL,NULL,diags); response = makeSearchResponseAPDU(0L,0L, 1L, 0L,UNUSED,FAILURE, aSearch->ReferenceID, wais_response); *sendBuf = writeSearchResponseAPDU(response,*sendBuf,sendBufLen); freeSearchResponseAPDU(response); freeWAISSearchResponse(wais_response); waislog(WLOG_HIGH, WLOG_ERROR, message); return; } } else len = interpret_message(message, len, response, rlen, NULL, false); bcopy(response+HEADER_LENGTH, *sendBuf, len); *sendBuf+=len; } /*----------------------------------------------------------------------*/ static void handleSearch _AP((char** recBuf,char** sendBuf, long* sendBufLen,long waisProtocolVersion, char *index_directory)); static void handleSearch(recBuf,sendBuf,sendBufLen,waisProtocolVersion,index_directory) char** recBuf; char** sendBuf; long* sendBufLen; long waisProtocolVersion; char *index_directory; /* figure out what kind of search this is, (query or retrieval) and dispatch to the appropriate function */ { SearchAPDU* aSearch = NULL; /* read the search data */ *recBuf = readSearchAPDU(&aSearch,*recBuf); if (*recBuf == NULL) { *sendBuf = NULL; /* error in the read */ return; } else { /* dispatch on the query type */ if((aSearch->DatabaseNames != NULL) && (aSearch->DatabaseNames[0] != NULL) && isRemoteDB(aSearch->DatabaseNames[0])) forwardSearch(aSearch,sendBuf,sendBufLen,waisProtocolVersion); else { if (strcmp(aSearch->QueryType,QT_TextRetrievalQuery) == 0) { handleElementRetrieval(aSearch,sendBuf,sendBufLen, waisProtocolVersion, index_directory); } else if (strcmp(aSearch->QueryType,QT_RelevanceFeedbackQuery) == 0) { char *seeds, *s; s = seeds = s_strdup(((WAISSearch *)aSearch->Query)->SeedWords); while(*s != 0) { if(*s == '\n' || *s == '\r') *s = ' '; s++; } if(aSearch->DatabaseNames != NULL && aSearch->DatabaseNames[0] != NULL) { waislog(WLOG_LOW, WLOG_SEARCH, "Search! Database: %s, Seed Words: %s", aSearch->DatabaseNames[0], seeds); } else { waislog(WLOG_LOW, WLOG_SEARCH, "Search! Database: None, Seed Words: %s", seeds); } handleRelevanceFeedbackSearch(aSearch,sendBuf,sendBufLen, waisProtocolVersion, index_directory); } else { waislog(WLOG_HIGH, WLOG_ERROR, "Unknown search type"); *sendBuf = NULL; /* error - unknown search type */ } fflush(stderr); } } } /*----------------------------------------------------------------------*/ static boolean needs_help _AP ((char *question)); static boolean needs_help(question) char *question; /* returns true if the user wants help */ { if(question[0] == '\0') /* null question, must need help */ return(true); if(question[0] == '?') return(true); if(strlen(question) < 20){ if((NULL != strstr(question, "help")) || (NULL != strstr(question, "HELP")) || (NULL != strstr(question, "Help"))){ return(true); } } return(false); } /* returns a help header to be returned or NULL if not possible */ static WAISDocumentHeader *help_header _AP((char *database_name, char *index_directory)); static WAISDocumentHeader *help_header(database_name, index_directory) char *database_name; char *index_directory; { /* make a help document */ hit help; char local_id[MAX_FILENAME_LEN + 60]; strncpy(help.filename, merge_pathnames(database_name,index_directory), MAX_FILENAME_LEN); strncat(help.filename, source_ext, MAX_FILENAME_LEN); /* printf("help filename %s", help.filename); */ strncpy(help.headline, "Information on database: ", MAX_FILENAME_LEN); strncat(help.headline, pathname_name(database_name), MAX_FILENAME_LEN); sprintf(local_id, "%ld %ld %s", 0L, 0L, help.filename); if(probe_file(help.filename)) { DocID* theDocID = NULL; long length; long lines; char **type = NULL; help.start_character = 0; help.end_character = 0; { FILE *stream = s_fopen(help.filename, "r"); lines = count_lines(stream); length = file_length(stream); s_fclose(stream); } type = (char**)s_malloc((size_t)(sizeof(char*) * 2)); type[0] = s_strdup("WSRC"); type[1] = NULL; /* then there is a source structure to return */ theDocID = makeDocID(); theDocID->originalDatabase = stringToAny(database_name); /* XXX */ theDocID->originalLocalID = stringToAny(local_id); return(makeWAISDocumentHeader(anyFromDocID(theDocID), UNUSED, MAX_NORMAL_SCORE, UNUSED, length,lines, type, s_strdup(database_name), /* XXX */ NULL, /* date */ s_strdup(help.headline), NULL)); } else return(NULL); } /* picks a set of random documents from the database static void pick_random_documents(aSearch, headers, &headerNum) { } */ /*----------------------------------------------------------------------*/ static void handleRelevanceFeedbackSearch _AP((SearchAPDU* aSearch,char** sendBuf,long* sendBufLen, long waisProtocolVersion, char *index_directory)); static void handleRelevanceFeedbackSearch(aSearch,sendBuf,sendBufLen, waisProtocolVersion, index_directory) SearchAPDU* aSearch; char** sendBuf; long* sendBufLen; long waisProtocolVersion; char *index_directory; { SearchResponseAPDU* response = NULL; WAISSearchResponse* wais_response = NULL; WAISDocumentHeader** headers = NULL; long headerNum = 0; char* seedwords_used = NULL; diagnosticRecord** diags = NULL; char *seed_words_used = s_strdup(((WAISSearch *)aSearch->Query)->SeedWords); boolean search_status; /* construct a response list */ headers = (WAISDocumentHeader**) s_malloc((size_t) (sizeof(WAISDocumentHeader*) * (1 + (unsigned long) ((WAISSearch *)aSearch->Query)->MaxDocumentsRetrieved))); headers[0] = NULL; /* handle help queries */ if(needs_help(seed_words_used)){ char *database_name = (aSearch->DatabaseNames == NULL) ? INFO_DATABASE_NAME : aSearch->DatabaseNames[0]; WAISDocumentHeader *header = help_header(database_name, index_directory); if(NULL != header){ headers[headerNum++] = header; headers[headerNum] = NULL; } } if(seed_words_used[0] == '\0' && ((WAISSearch *)aSearch->Query)->Docs == NULL){ /* pick_random_documents(aSearch, headers, &headerNum); */ } else /* run the search on the database. If a new search engine were to be used, this is where it would be hooked in */ search_status = run_search(aSearch, headers,&diags, index_directory, &seed_words_used, waisProtocolVersion, &headerNum); { /* generate report on results. */ char *message; long size, i; /* calculate total length needed for log report */ for(size = 0L, i = 0; i < headerNum; i++) size+=(headers[i]->DocumentID->size+2); if (size > 0) { message = s_malloc(size); message[0] = 0; for (i = 0; i < headerNum; i++) { char docname[MAX_FILE_NAME_LEN+50]; sprintf(docname, "%s", anyToString(GetLocalID(docIDFromAny(headers[i]->DocumentID)))); s_strncat(message, docname, headers[i]->DocumentID->size, size); if ( i < headerNum-1) strcat(message, ", "); } waislog(WLOG_LOW, WLOG_RESULTS, "Returned %d results: %s", headerNum, message); s_free(message); } else waislog(WLOG_LOW, WLOG_RESULTS, "Returned 0 results. Aww."); } wais_response = makeWAISSearchResponse(seedwords_used,headers,NULL, NULL,NULL,NULL,NULL,diags); response = makeSearchResponseAPDU(search_status,0L, headerNum + ((diags == NULL) ? 0 : 1), 0L,UNUSED,SUCCESS, aSearch->ReferenceID, wais_response); /* write it */ *sendBuf = writeSearchResponseAPDU(response,*sendBuf,sendBufLen); freeWAISSearch((WAISSearch*)aSearch->Query); freeSearchAPDU(aSearch); freeSearchResponseAPDU(response); freeWAISSearchResponse(wais_response); /* free headers & seed_words_used */ } /*----------------------------------------------------------------------*/ /* this is defined above -brewster static void handleElementRetrieval _AP((SearchAPDU* aSearch,char** sendBuf,long* sendBufLen, long waisProtocolVersion)); */ static void handleElementRetrieval(aSearch,sendBuf,sendBufLen,waisProtocolVersion, index_directory) SearchAPDU* aSearch; char** sendBuf; long* sendBufLen; long waisProtocolVersion; char *index_directory; /* this is a type 1 query of the restricted form specified in the WAIS-protocol. Interpret it and write out an appropriate search response. (note the valid element sets are Document-Text,Document-Headlines, and Document-Codes but we only support text for now). */ { SearchResponseAPDU* response = NULL; WAISSearchResponse* wais_response = NULL; DocObj** docs = NULL; DocObj* doc = NULL; char *databaseName; void **elementList = NULL; void *element = NULL; diagnosticRecord** diags = NULL; diagnosticRecord* diag = NULL; long numDiags = 0L; long numElements = 0L; long i; database* db; char* new_db_name; /* read the query */ docs = readWAISTextQuery((any*)aSearch->Query); databaseName = (aSearch->DatabaseNames == NULL) ? INFO_DATABASE_NAME : aSearch->DatabaseNames[0]; new_db_name = merge_pathnames(databaseName, index_directory); /* assemble the elements and construct a response */ for (i = 0L, doc = docs[i]; doc != NULL; doc = docs[++i]) { long errorCode; any* bufAny; long size; if (doc->Type != NULL && strcmp(doc->Type, "WAIS_NEXT") == 0) { char docname[MAX_FILE_NAME_LEN+50], *buffer; db = openDatabase(new_db_name, false, true); if ((size = next_doc(docname, anyToString(GetLocalID(docIDFromAny(doc->DocumentID))), db)) > 0) { buffer = s_malloc(strlen(docname)+50); sprintf(buffer, "%s, %d", docname, size); bufAny = makeAny(strlen(buffer)+1,buffer); element = (void*)makeWAISDocumentText(duplicateAny(doc->DocumentID),0L,bufAny); } else element = NULL; closeDatabase(db); } else if (doc->Type != NULL && strcmp(doc->Type, "WAIS_PREV") == 0) { char docname[MAX_FILE_NAME_LEN+50], *buffer; any* bufAny; long size; db = openDatabase(new_db_name, false, true); if ((size = previous_doc(docname, anyToString(GetLocalID(docIDFromAny(doc->DocumentID))), db)) > 0) { buffer = s_malloc(strlen(docname)+50); sprintf(buffer, "%s, %d", docname, size); bufAny = makeAny(strlen(buffer),buffer); element = (void*)makeWAISDocumentText(duplicateAny(doc->DocumentID),0L,bufAny); } else element = NULL; closeDatabase(db); } else if (doc->ChunkCode == CT_line) element = (void*)getDocumentText(doc, databaseName, &errorCode); else if (doc->ChunkCode == CT_byte) element = (void*)getData(doc, databaseName, &errorCode); if (element == NULL) { /* make a diagnostic record to return */ switch (errorCode) { case GDT_UnsupportedChunkType: diag = makeDiag(true,D_PresentRequestOutOfRange, "Bad ChunkType in Request"); break; case GDT_BadDocID: diag = makeDiag(true,D_PresentRequestOutOfRange, "Bad DocID in request"); break; case GDT_MissingDocID: diag = makeDiag(true,D_PresentRequestOutOfRange, "Missing DocID in request"); break; case GDT_BadRange: diag = makeDiag(true,D_PresentRequestOutOfRange, "Request out of range"); break; case GDT_MissingDatabase: diag = makeDiag(true,D_PresentRequestOutOfRange, "Database missing from request"); break; case GDT_BadDatabase: diag = makeDiag(true,D_PresentRequestOutOfRange, "File not present in specified database"); break; default: /* should never get here */ diag = NULL; break; }; diags = (diagnosticRecord**)s_realloc(diags,(size_t)(sizeof(diagnosticRecord*) * (numDiags + 2))); diags[numDiags++] = diag; diags[numDiags] = NULL; } else { if (elementList == NULL) /* create a list */ { elementList = (void**)s_malloc((size_t)sizeof(void*) * 2); } else /* grow the list */ { elementList = (void**)s_realloc((char*)elementList, (size_t)(sizeof(void*) * (numElements + 2))); } elementList[numElements++] = element; /* put it in the list */ elementList[numElements] = NULL; } } wais_response = makeWAISSearchResponse(NULL,NULL,NULL,NULL, (WAISDocumentText**)elementList,NULL, NULL,diags); response = makeSearchResponseAPDU(SUCCESS,0L,numElements + numDiags,0L,UNUSED, SUCCESS,aSearch->ReferenceID, wais_response); /* write it */ *sendBuf = writeSearchResponseAPDU(response,*sendBuf,sendBufLen); /* clean up */ freeAny((any*)aSearch->Query); /* have to explicitly free the user info */ freeSearchAPDU(aSearch); freeSearchResponseAPDU(response); freeWAISSearchResponse(wais_response); /* frees the elements constructed */ doList((void**)docs,freeDocObj); s_free(docs); } /*----------------------------------------------------------------------*/