void construct_lcp_PHI(cache_config& config) { static_assert(t_width == 0 or t_width == 8 , "construct_lcp_PHI: width must be `0` for integer alphabet and `8` for byte alphabet"); typedef int_vector<>::size_type size_type; typedef int_vector<t_width> text_type; const char* KEY_TEXT = key_text_trait<t_width>::KEY_TEXT; int_vector_buffer<> sa_buf(config.file_map[conf::KEY_SA]); size_type n = sa_buf.size(); assert(n > 0); if (1 == n) { // Handle special case: Input only the sentinel character. int_vector<> lcp(1, 0); store_to_cache(lcp, conf::KEY_LCP, config); return; } // (1) Calculate PHI (stored in array plcp) int_vector<> plcp(n, 0, sa_buf.width()); for (size_type i=0, sai_1 = 0; i < n; ++i) { size_type sai = sa_buf[i]; plcp[ sai ] = sai_1; sai_1 = sai; } // (2) Load text from disk text_type text; load_from_cache(text, KEY_TEXT, config); // (3) Calculate permuted LCP array (text order), called PLCP size_type max_l = 0; for (size_type i=0, l=0; i < n-1; ++i) { size_type phii = plcp[i]; while (text[i+l] == text[phii+l]) { ++l; } plcp[i] = l; if (l) { max_l = std::max(max_l, l); --l; } } util::clear(text); uint8_t lcp_width = bits::hi(max_l)+1; // (4) Transform PLCP into LCP std::string lcp_file = cache_file_name(conf::KEY_LCP, config); size_type buffer_size = 1000000; // buffer_size is a multiple of 8! int_vector_buffer<> lcp_buf(lcp_file, std::ios::out, buffer_size, lcp_width); // open buffer for lcp lcp_buf[0] = 0; sa_buf.buffersize(buffer_size); for (size_type i=1; i < n; ++i) { size_type sai = sa_buf[i]; lcp_buf[i] = plcp[sai]; } lcp_buf.close(); register_cache_file(conf::KEY_LCP, config); }
void construct_sa(cache_config& config) { static_assert(t_width == 0 or t_width == 8 , "construct_sa: width must be `0` for integer alphabet and `8` for byte alphabet"); const char* KEY_TEXT = key_text_trait<t_width>::KEY_TEXT; if (t_width == 8) { typedef int_vector<t_width> text_type; text_type text; load_from_cache(text, KEY_TEXT, config); // call divsufsort int_vector<> sa(text.size(), 0, bits::hi(text.size())+1); algorithm::calculate_sa((const unsigned char*)text.data(), text.size(), sa); store_to_cache(sa, conf::KEY_SA, config); } else if (t_width == 0) { // call qsufsort int_vector<> sa; sdsl::qsufsort::construct_sa(sa, config.file_map[KEY_TEXT].c_str(), 0); store_to_cache(sa, conf::KEY_SA, config); } else { std::cerr << "Unknown alphabet type" << std::endl; } }
void construct(t_index& idx, const std::string& file, cache_config& config, uint8_t num_bytes, lcp_tag) { auto event = memory_monitor::event("construct compressed LCP"); const char* KEY_TEXT = key_text_trait<t_width>::KEY_TEXT; typedef int_vector<t_width> text_type; { // (2) check, if the longest common prefix array is cached auto event = memory_monitor::event("LCP"); if (!cache_file_exists(conf::KEY_LCP, config)) { { auto event = memory_monitor::event("parse input text"); // (1) check, if the text is cached if (!cache_file_exists(KEY_TEXT, config)) { text_type text; load_vector_from_file(text, file, num_bytes); if (contains_no_zero_symbol(text, file)) { append_zero_symbol(text); store_to_cache(text,KEY_TEXT, config); } } register_cache_file(KEY_TEXT, config); } { // (2) check, if the suffix array is cached auto event = memory_monitor::event("SA"); if (!cache_file_exists(conf::KEY_SA, config)) { construct_sa<t_width>(config); } register_cache_file(conf::KEY_SA, config); } if (t_width==8) { construct_lcp_semi_extern_PHI(config); } else { construct_lcp_PHI<t_width>(config); } } register_cache_file(conf::KEY_LCP, config); } { auto event = memory_monitor::event("compressed LCP"); t_index tmp(config); tmp.swap(idx); } if (config.delete_files) { auto event = memory_monitor::event("delete temporary files"); util::delete_all_files(config.file_map); } }
csa_sada<t_enc_vec, t_dens, t_inv_dens, t_sa_sample_strat, t_isa, t_alphabet_strat>::csa_sada(cache_config& config) { create_buffer(); if (!cache_file_exists(key_trait<alphabet_type::int_width>::KEY_BWT, config)) { return; } int_vector_buffer<alphabet_type::int_width> bwt_buf(cache_file_name(key_trait<alphabet_type::int_width>::KEY_BWT,config)); size_type n = bwt_buf.size(); { auto event = memory_monitor::event("construct csa-alpbabet"); alphabet_type tmp_alphabet(bwt_buf, n); m_alphabet.swap(tmp_alphabet); } int_vector<> cnt_chr(sigma, 0, bits::hi(n)+1); for (typename alphabet_type::sigma_type i=0; i < sigma; ++i) { cnt_chr[i] = C[i]; } // calculate psi { auto event = memory_monitor::event("construct PSI"); // TODO: move PSI construct into construct_PSI.hpp int_vector<> psi(n, 0, bits::hi(n)+1); for (size_type i=0; i < n; ++i) { psi[ cnt_chr[ char2comp[bwt_buf[i]] ]++ ] = i; } std::string psi_file = cache_file_name(conf::KEY_PSI, config); if (!store_to_cache(psi, conf::KEY_PSI, config)) { return; } } { auto event = memory_monitor::event("encode PSI"); int_vector_buffer<> psi_buf(cache_file_name(conf::KEY_PSI, config)); t_enc_vec tmp_psi(psi_buf); m_psi.swap(tmp_psi); } { auto event = memory_monitor::event("sample SA"); sa_sample_type tmp_sa_sample(config); m_sa_sample.swap(tmp_sa_sample); } { auto event = memory_monitor::event("sample ISA"); isa_sample_type isa_s(config, &m_sa_sample); util::swap_support(m_isa_sample, isa_s, &m_sa_sample, (const sa_sample_type*)nullptr); } }
void construct(t_index& idx, const std::string& file, cache_config& config, uint8_t num_bytes, csa_tag) { auto event = memory_monitor::event("construct CSA"); const char* KEY_TEXT = key_text_trait<t_index::alphabet_category::WIDTH>::KEY_TEXT; const char* KEY_BWT = key_bwt_trait<t_index::alphabet_category::WIDTH>::KEY_BWT; typedef int_vector<t_index::alphabet_category::WIDTH> text_type; { auto event = memory_monitor::event("parse input text"); // (1) check, if the text is cached if (!cache_file_exists(KEY_TEXT, config)) { text_type text; load_vector_from_file(text, file, num_bytes); if (contains_no_zero_symbol(text, file)) { append_zero_symbol(text); store_to_cache(text,KEY_TEXT, config); } } register_cache_file(KEY_TEXT, config); } { // (2) check, if the suffix array is cached auto event = memory_monitor::event("SA"); if (!cache_file_exists(conf::KEY_SA, config)) { construct_sa<t_index::alphabet_category::WIDTH>(config); } register_cache_file(conf::KEY_SA, config); } { // (3) construct BWT auto event = memory_monitor::event("BWT"); if (!cache_file_exists(KEY_BWT, config)) { construct_bwt<t_index::alphabet_category::WIDTH>(config); } register_cache_file(KEY_BWT, config); } { // (4) use BWT to construct the CSA auto event = memory_monitor::event("construct CSA"); t_index tmp(config); idx.swap(tmp); } if (config.delete_files) { auto event = memory_monitor::event("delete temporary files"); util::delete_all_files(config.file_map); } }
void construct_lcp_kasai(cache_config& config) { int_vector<> lcp; typedef int_vector<>::size_type size_type; construct_isa(config); { int_vector<t_width> text; if (!load_from_cache(text, key_text_trait<t_width>::KEY_TEXT, config)) { return; } int_vector_file_buffer<> isa_buf(config.file_map[constants::KEY_ISA], 1000000); // init isa file_buffer int_vector<> sa; if (!load_from_cache(sa, constants::KEY_SA, config)) { return; } // use Kasai algorithm to compute the lcp values for (size_type i=0,j=0,sa_1=0,l=0, r_sum=0, r=isa_buf.load_next_block(), n=isa_buf.int_vector_size; r_sum < n;) { for (; i < r_sum+r; ++i) { sa_1 = isa_buf[i-r_sum]; // = isa[i] if (sa_1) { j = sa[sa_1-1]; if (l) --l; assert(i!=j); while (text[i+l]==text[j+l]) { // i+l < n and j+l < n are not necessary, since text[n]=0 and text[i]!=0 (i<n) and i!=j ++l; } sa[ sa_1-1 ] = l; //overwrite sa array with lcp values } else { l = 0; sa[ n-1 ] = 0; } } r_sum += r; r = isa_buf.load_next_block(); } for (size_type i=sa.size(); i>1; --i) { sa[i-1] = sa[i-2]; } sa[0] = 0; lcp.swap(sa); } store_to_cache(lcp, constants::KEY_LCP, config); }
void construct(t_index& idx, const std::string& file, cache_config& config, uint8_t num_bytes, cst_tag) { auto event = memory_monitor::event("construct CST"); const char* KEY_TEXT = key_text_trait<t_index::alphabet_category::WIDTH>::KEY_TEXT; const char* KEY_BWT = key_bwt_trait<t_index::alphabet_category::WIDTH>::KEY_BWT; csa_tag csa_t; { // (1) check, if the compressed suffix array is cached typename t_index::csa_type csa; if (!cache_file_exists(std::string(conf::KEY_CSA)+"_"+util::class_to_hash(csa), config)) { cache_config csa_config(false, config.dir, config.id, config.file_map); construct(csa, file, csa_config, num_bytes, csa_t); auto event = memory_monitor::event("store CSA"); config.file_map = csa_config.file_map; store_to_cache(csa,std::string(conf::KEY_CSA)+"_"+util::class_to_hash(csa), config); } register_cache_file(std::string(conf::KEY_CSA)+"_"+util::class_to_hash(csa), config); } { // (2) check, if the longest common prefix array is cached auto event = memory_monitor::event("LCP"); register_cache_file(KEY_TEXT, config); register_cache_file(KEY_BWT, config); register_cache_file(conf::KEY_SA, config); if (!cache_file_exists(conf::KEY_LCP, config)) { if (t_index::alphabet_category::WIDTH==8) { construct_lcp_semi_extern_PHI(config); } else { construct_lcp_PHI<t_index::alphabet_category::WIDTH>(config); } } register_cache_file(conf::KEY_LCP, config); } { auto event = memory_monitor::event("CST"); t_index tmp(config); tmp.swap(idx); } if (config.delete_files) { auto event = memory_monitor::event("delete temporary files"); util::delete_all_files(config.file_map); } }
void construct_lcp_kasai(cache_config& config) { static_assert(t_width == 0 or t_width == 8 , "construct_lcp_kasai: width must be `0` for integer alphabet and `8` for byte alphabet"); int_vector<> lcp; typedef int_vector<>::size_type size_type; construct_isa(config); { int_vector<t_width> text; if (!load_from_cache(text, key_text_trait<t_width>::KEY_TEXT, config)) { return; } int_vector_buffer<> isa_buf(config.file_map[conf::KEY_ISA], std::ios::in, 1000000); // init isa file_buffer int_vector<> sa; if (!load_from_cache(sa, conf::KEY_SA, config)) { return; } // use Kasai algorithm to compute the lcp values for (size_type i=0,j=0,sa_1=0,l=0, n=isa_buf.size(); i < n; ++i) { sa_1 = isa_buf[i]; // = isa[i] if (sa_1) { j = sa[sa_1-1]; if (l) --l; assert(i!=j); while (text[i+l]==text[j+l]) { // i+l < n and j+l < n are not necessary, since text[n]=0 and text[i]!=0 (i<n) and i!=j ++l; } sa[ sa_1-1 ] = l; //overwrite sa array with lcp values } else { l = 0; sa[ n-1 ] = 0; } } for (size_type i=sa.size(); i>1; --i) { sa[i-1] = sa[i-2]; } sa[0] = 0; lcp.swap(sa); } store_to_cache(lcp, conf::KEY_LCP, config); }
void construct_lcp_PHI(cache_config& config) { typedef int_vector<>::size_type size_type; typedef int_vector<t_width> text_type; const char* KEY_TEXT = key_text_trait<t_width>::KEY_TEXT; int_vector_file_buffer<> sa_buf(config.file_map[constants::KEY_SA]); size_type n = sa_buf.int_vector_size; assert(n > 0); if (1 == n) { // Handle special case: Input only the sentinel character. int_vector<> lcp(1, 0); store_to_cache(lcp, constants::KEY_LCP, config); return; } // (1) Calculate PHI (stored in array plcp) int_vector<> plcp(n, 0, sa_buf.width); for (size_type i=0, r_sum=0, r=sa_buf.load_next_block(), sai_1 = 0; r_sum < n;) { for (; i < r_sum+r; ++i) { size_type sai = sa_buf[i-r_sum]; plcp[ sai ] = sai_1; sai_1 = sai; } r_sum += r; r = sa_buf.load_next_block(); } // (2) Load text from disk text_type text; load_from_cache(text, KEY_TEXT, config); // (3) Calculate permuted LCP array (text order), called PLCP size_type max_l = 0; for (size_type i=0, l=0; i < n-1; ++i) { size_type phii = plcp[i]; while (text[i+l] == text[phii+l]) { ++l; } plcp[i] = l; if (l) { max_l = std::max(max_l, l); --l; } } util::clear(text); uint8_t lcp_width = bits::hi(max_l)+1; // (4) Transform PLCP into LCP std::string lcp_file = cache_file_name(constants::KEY_LCP, config); osfstream lcp_out_buf(lcp_file, std::ios::binary | std::ios::app | std::ios::out); // open buffer for lcp size_type bit_size = n*lcp_width; lcp_out_buf.write((char*) &(bit_size), sizeof(bit_size)); // write size of vector lcp_out_buf.write((char*) &(lcp_width),sizeof(lcp_width)); // write int_width of vector size_type wb = 0; // bytes written into lcp int_vector size_type buffer_size = 1000000; // buffer_size is a multiple of 8! int_vector<> lcp_buf(buffer_size, 0, lcp_width); lcp_buf[0] = 0; sa_buf.reset(buffer_size); size_type r = 0;// sa_buf.load_next_block(); for (size_type i=1, r_sum=0; r_sum < n;) { for (; i < r_sum+r; ++i) { size_type sai = sa_buf[i-r_sum]; lcp_buf[ i-r_sum ] = plcp[sai]; } if (r > 0) { size_type cur_wb = (r*lcp_buf.width()+7)/8; lcp_out_buf.write((const char*)lcp_buf.data(), cur_wb); wb += cur_wb; } r_sum += r; r = sa_buf.load_next_block(); } if (wb%8) { lcp_out_buf.write("\0\0\0\0\0\0\0\0", 8-wb%8); } lcp_out_buf.close(); register_cache_file(constants::KEY_LCP, config); }
errno_t name2ip( in_addr_t *out, const char *name, int flags ) { int ia, ib, ic, id; if( 4 == sscanf( name, "%d.%d.%d.%d", &ia, &ib, &ic, &id ) ) { // No resolver required, ip4 addr given ipv4_addr iaddr = IPV4_DOTADDR_TO_ADDR( ia, ib, ic, id); *out = htonl( iaddr ); SHOW_FLOW( 2, "parsed %s to %s", name, inet_ntoa(* (struct in_addr*)out) ); return 0; } if(!inited) return ENXIO; int tries = 20; if(flags & RESOLVER_FLAG_NORETRY) tries = 1; ipv4_addr result; //ipv4_addr next_servers[MAX_DNS_SERVERS]; SHOW_FLOW( 1, "request '%s'", name ); ipv4_addr * sptr = servers; int sleft = MAX_DNS_SERVERS; if( !(flags & RESOLVER_FLAG_NORCACHE) ) if( lookup_cache( out, name ) == 0 ) { SHOW_FLOW0( 1, "got from cache"); return 0; } // On OS stop don't produce network traffic if( (flags & RESOLVER_FLAG_NOWAIT) || phantom_stop_level ) return ESRCH; while(tries--) { ipv4_addr server = *sptr++; if(sleft-- <= 0 || server == 0) { SHOW_ERROR0( 1, "No more places to look in, give up\n"); return ENOENT; } SHOW_FLOW( 2, "look in %s", inet_ntoa(* (struct in_addr*)&server) ); errno_t res = dns_request( (const unsigned char *)name, server, &result ); if( res == 0 )//|| result != 0 ) { SHOW_FLOW( 1, "answer is %s", inet_ntoa(* (struct in_addr*)&result) ); *out = result; if( !(flags & RESOLVER_FLAG_NOWCACHE) ) store_to_cache( result, name ); return 0; } } return ENOENT; }