// if one of the files is reading from stdin, sum_kmers_ptr is set to 0 // `max_cols_ptr` is used to return the most colours being loaded from a single file // returns the number of colours being loaded in total size_t graph_files_open(char **graph_paths, GraphFileReader *gfiles, size_t num_gfiles, size_t *max_kmers_ptr, size_t *sum_kmers_ptr) { size_t i, ctx_max_kmers = 0, ctx_sum_kmers = 0; bool ctx_uses_stdin = false; size_t ncols = 0; for(i = 0; i < num_gfiles; i++) { memset(&gfiles[i], 0, sizeof(GraphFileReader)); graph_file_open2(&gfiles[i], graph_paths[i], "r", ncols); if(gfiles[0].hdr.kmer_size != gfiles[i].hdr.kmer_size) { cmd_print_usage("Kmer sizes don't match [%u vs %u]", gfiles[0].hdr.kmer_size, gfiles[i].hdr.kmer_size); } ncols = MAX2(ncols, file_filter_into_ncols(&gfiles[i].fltr)); ctx_max_kmers = MAX2(ctx_max_kmers, graph_file_nkmers(&gfiles[i])); ctx_sum_kmers += graph_file_nkmers(&gfiles[i]); ctx_uses_stdin |= file_filter_isstdin(&gfiles[i].fltr); } if(ctx_uses_stdin) ctx_sum_kmers = SIZE_MAX; *max_kmers_ptr = ctx_max_kmers; *sum_kmers_ptr = ctx_sum_kmers; return ncols; }
int ctx_sort(int argc, char **argv) { const char *out_path = NULL; struct MemArgs memargs = MEM_ARGS_INIT; // Arg parsing char cmd[100]; char shortopts[300]; cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts)); int c; // silence error messages from getopt_long // opterr = 0; while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) { cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd)); switch(c) { case 0: /* flag set */ break; case 'h': cmd_print_usage(NULL); break; case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break; case 'm': cmd_mem_args_set_memory(&memargs, optarg); break; case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break; case 'o': cmd_check(!out_path, cmd); out_path = optarg; break; case ':': /* BADARG */ case '?': /* BADCH getopt_long has already printed error */ // cmd_print_usage(NULL); die("`"CMD" sort -h` for help. Bad option: %s", argv[optind-1]); default: die("Bad option: [%c]: %s", c, cmd); } } if(optind+1 != argc) cmd_print_usage("Require exactly one input graph file (.ctx)"); const char *ctx_path = argv[optind]; // // Open Graph file // GraphFileReader gfile; memset(&gfile, 0, sizeof(GraphFileReader)); graph_file_open2(&gfile, ctx_path, out_path ? "r" : "r+", true, 0); if(!file_filter_is_direct(&gfile.fltr)) die("Cannot open graph file with a filter ('in.ctx:blah' syntax)"); size_t num_kmers, memory; // Reading from a stream if(gfile.num_of_kmers < 0) { if(!memargs.num_kmers_set) die("If reading from a stream, must give -n <num_kmers>"); num_kmers = memargs.num_kmers; } else num_kmers = gfile.num_of_kmers; // Open output path (if given) FILE *fout = out_path ? futil_fopen_create(out_path, "w") : NULL; size_t i; size_t ncols = gfile.hdr.num_of_cols; size_t kmer_mem = sizeof(BinaryKmer) + (sizeof(Edges)+sizeof(Covg))*ncols; memory = (sizeof(char*) + kmer_mem) * num_kmers; char mem_str[50]; bytes_to_str(memory, 1, mem_str); if(memory > memargs.mem_to_use) die("Require at least %s memory", mem_str); status("[memory] Total: %s", mem_str); char *mem = ctx_malloc(kmer_mem * num_kmers); char **kmers = ctx_malloc(num_kmers*sizeof(char*)); // Read in whole file // if(graph_file_fseek(gfile, gfile.hdr_size, SEEK_SET) != 0) die("fseek failed"); size_t nkread = gfr_fread_bytes(&gfile, mem, num_kmers*kmer_mem); if(nkread != num_kmers*kmer_mem) die("Could only read %zu bytes [<%zu]", nkread, num_kmers*kmer_mem); // check we are at the end of the file char tmpc; if(gfr_fread_bytes(&gfile, &tmpc, 1) != 0) { die("More kmers in file than believed (kmers: %zu ncols: %zu).", num_kmers, ncols); } status("Read %zu kmers with %zu colour%s", num_kmers, ncols, util_plural_str(ncols)); for(i = 0; i < num_kmers; i++) kmers[i] = mem + kmer_mem*i; sort_block(kmers, num_kmers); // Print if(out_path != NULL) { // saving to a different destination - write header graph_write_header(fout, &gfile.hdr); } else { // Directly manipulating gfile.fh here, using it to write later // Not doing any more reading if(fseek(gfile.fh, gfile.hdr_size, SEEK_SET) != 0) die("fseek failed"); fout = gfile.fh; } for(i = 0; i < num_kmers; i++) if(fwrite(kmers[i], 1, kmer_mem, fout) != kmer_mem) die("Cannot write to file"); if(out_path) fclose(fout); graph_file_close(&gfile); ctx_free(kmers); ctx_free(mem); return EXIT_SUCCESS; }
int ctx_index(int argc, char **argv) { const char *out_path = NULL; size_t block_size = 0, block_kmers = 0; // Arg parsing char cmd[100]; char shortopts[300]; cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts)); int c; // silence error messages from getopt_long // opterr = 0; while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) { cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd)); switch(c) { case 0: /* flag set */ break; case 'h': cmd_print_usage(NULL); break; case 'o': cmd_check(!out_path, cmd); out_path = optarg; break; case 'b': cmd_check(!block_kmers, cmd); block_kmers = cmd_size_nonzero(cmd, optarg); break; case 's': cmd_check(!block_size, cmd); block_size = cmd_size_nonzero(cmd, optarg); break; case ':': /* BADARG */ case '?': /* BADCH getopt_long has already printed error */ // cmd_print_usage(NULL); die("`"CMD" index -h` for help. Bad option: %s", argv[optind-1]); default: abort(); } } if(optind+1 != argc) cmd_print_usage("Require exactly one input graph file (.ctx)"); if(block_size && block_kmers) cmd_print_usage("Cannot use --block-kmers and --block-size together"); const char *ctx_path = argv[optind]; // // Open Graph file // GraphFileReader gfile; memset(&gfile, 0, sizeof(GraphFileReader)); graph_file_open2(&gfile, ctx_path, "r+", true, 0); if(!file_filter_is_direct(&gfile.fltr)) die("Cannot open graph file with a filter ('in.ctx:blah' syntax)"); // Open output file FILE *fout = out_path ? futil_fopen_create(out_path, "w") : stdout; // Start size_t filencols = gfile.hdr.num_of_cols; size_t kmer_size = gfile.hdr.kmer_size; const char *path = file_filter_path(&gfile.fltr); size_t ncols = file_filter_into_ncols(&gfile.fltr); size_t kmer_mem = sizeof(BinaryKmer) + (sizeof(Edges)+sizeof(Covg))*filencols; if(block_size) { block_kmers = block_size / kmer_mem; } else if(!block_size && !block_kmers) { block_size = 4 * ONE_MEGABYTE; block_kmers = block_size / kmer_mem; } // Update block-size block_size = block_kmers * kmer_mem; status("[index] block bytes: %zu kmers: %zu; kmer bytes: %zu, hdr: %zu", block_size, block_kmers, kmer_mem, (size_t)gfile.hdr_size); if(block_kmers == 0) die("Cannot set block_kmers to zero"); // Print header fputs("#block_start\tnext_block\tfirst_kmer\tkmer_idx\tnext_kmer_idx\n", fout); BinaryKmer bkmer = BINARY_KMER_ZERO_MACRO; BinaryKmer prev_bkmer = BINARY_KMER_ZERO_MACRO; Covg *covgs = ctx_malloc(ncols * sizeof(Covg)); Edges *edges = ctx_malloc(ncols * sizeof(Edges)); char bkmerstr[MAX_KMER_SIZE+1]; size_t rem_block = block_size - kmer_mem; // block after first kmer char *tmp_mem = ctx_malloc(rem_block); // Read in file, print index size_t nblocks = 0; size_t bl_bytes = 0, bl_kmers = 0; size_t bl_byte_offset = gfile.hdr_size, bl_kmer_offset = 0; while(1) { if(!graph_file_read(&gfile, &bkmer, covgs, edges)) { status("Read kmer failed"); break; } binary_kmer_to_str(bkmer, kmer_size, bkmerstr); if(nblocks > 0 && !binary_kmer_less_than(prev_bkmer,bkmer)) die("File is not sorted: %s [%s]", bkmerstr, path); // We've already read one kmer entry, read rest of block bl_bytes = kmer_mem + gfr_fread_bytes(&gfile, tmp_mem, rem_block); bl_kmers = 1 + bl_bytes / kmer_mem; fprintf(fout, "%zu\t%zu\t%s\t%zu\t%zu\n", bl_byte_offset, bl_byte_offset+bl_bytes, bkmerstr, bl_kmer_offset, bl_kmer_offset+bl_kmers); bl_byte_offset += bl_bytes; bl_kmer_offset += bl_kmers; nblocks++; if(bl_kmers < block_kmers) { status("last block %zu < %zu; %zu vs %zu", bl_kmers, block_kmers, bl_bytes, block_size); break; } prev_bkmer = bkmer; } ctx_free(covgs); ctx_free(edges); ctx_free(tmp_mem); // done char num_kmers_str[50], num_blocks_str[50]; char block_mem_str[50], block_kmers_str[50]; ulong_to_str(bl_kmer_offset, num_kmers_str); ulong_to_str(nblocks, num_blocks_str); bytes_to_str(block_size, 1, block_mem_str); ulong_to_str(block_kmers, block_kmers_str); status("Read %s kmers in %s block%s (block size %s / %s kmers)", num_kmers_str, num_blocks_str, util_plural_str(nblocks), block_mem_str, block_kmers_str); if(fout != stdout) status("Saved to %s", out_path); graph_file_close(&gfile); fclose(fout); return EXIT_SUCCESS; }
int ctx_join(int argc, char **argv) { struct MemArgs memargs = MEM_ARGS_INIT; const char *out_path = NULL; size_t use_ncols = 0; GraphFileReader tmp_gfile; GraphFileBuffer isec_gfiles_buf; gfile_buf_alloc(&isec_gfiles_buf, 8); // Arg parsing char cmd[100], shortopts[100]; cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts)); int c; while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) { cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd)); switch(c) { case 0: /* flag set */ break; case 'h': cmd_print_usage(NULL); break; case 'o': cmd_check(!out_path, cmd); out_path = optarg; break; case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break; case 'm': cmd_mem_args_set_memory(&memargs, optarg); break; case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break; case 'N': cmd_check(!use_ncols, cmd); use_ncols = cmd_uint32_nonzero(cmd, optarg); break; case 'i': graph_file_reset(&tmp_gfile); graph_file_open(&tmp_gfile, optarg); if(file_filter_into_ncols(&tmp_gfile.fltr) > 1) warn("Flattening intersection graph into colour 0: %s", optarg); file_filter_flatten(&tmp_gfile.fltr, 0); gfile_buf_push(&isec_gfiles_buf, &tmp_gfile, 1); break; case ':': /* BADARG */ case '?': /* BADCH getopt_long has already printed error */ // cmd_print_usage(NULL); die("`"CMD" join -h` for help. Bad option: %s", argv[optind-1]); default: abort(); } } GraphFileReader *igfiles = isec_gfiles_buf.b; size_t num_igfiles = isec_gfiles_buf.len; if(!out_path) cmd_print_usage("--out <out.ctx> required"); if(optind >= argc) cmd_print_usage("Please specify at least one input graph file"); // optind .. argend-1 are graphs to load size_t num_gfiles = (size_t)(argc - optind); char **gfile_paths = argv + optind; GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader)); status("Probing %zu graph files and %zu intersect files", num_gfiles, num_igfiles); // Check all binaries are valid binaries with matching kmer size size_t i; size_t ctx_max_cols = 0; uint64_t min_intersect_num_kmers = 0, ctx_max_kmers = 0, ctx_sum_kmers = 0; for(i = 0; i < num_gfiles; i++) { graph_file_open2(&gfiles[i], gfile_paths[i], "r", true, ctx_max_cols); if(gfiles[0].hdr.kmer_size != gfiles[i].hdr.kmer_size) { cmd_print_usage("Kmer sizes don't match [%u vs %u]", gfiles[0].hdr.kmer_size, gfiles[i].hdr.kmer_size); } ctx_max_cols = MAX2(ctx_max_cols, file_filter_into_ncols(&gfiles[i].fltr)); ctx_max_kmers = MAX2(ctx_max_kmers, graph_file_nkmers(&gfiles[i])); ctx_sum_kmers += graph_file_nkmers(&gfiles[i]); } // Probe intersection graph files for(i = 0; i < num_igfiles; i++) { if(gfiles[0].hdr.kmer_size != igfiles[i].hdr.kmer_size) { cmd_print_usage("Kmer sizes don't match [%u vs %u]", gfiles[0].hdr.kmer_size, igfiles[i].hdr.kmer_size); } uint64_t nkmers = graph_file_nkmers(&igfiles[i]); if(i == 0) min_intersect_num_kmers = nkmers; else if(nkmers < min_intersect_num_kmers) { // Put smallest intersection binary first SWAP(igfiles[i], igfiles[0]); min_intersect_num_kmers = nkmers; } } bool take_intersect = (num_igfiles > 0); // If we are taking an intersection, // all kmers intersection kmers will need to be loaded if(take_intersect) ctx_max_kmers = ctx_sum_kmers = min_intersect_num_kmers; bool use_ncols_set = (use_ncols > 0); bool output_to_stdout = (strcmp(out_path,"-") == 0); // if(use_ncols == 0) use_ncols = 1; if(use_ncols_set) { if(use_ncols < ctx_max_cols && output_to_stdout) die("I need %zu colours if outputting to STDOUT (--ncols)", ctx_max_cols); if(use_ncols > ctx_max_cols) { warn("I only need %zu colour%s ('--ncols %zu' ignored)", ctx_max_cols, util_plural_str(ctx_max_cols), use_ncols); use_ncols = ctx_max_cols; } } else { use_ncols = output_to_stdout ? ctx_max_cols : 1; } // Check out_path is writable futil_create_output(out_path); status("Output %zu cols; from %zu files; intersecting %zu graphs; ", ctx_max_cols, num_gfiles, num_igfiles); if(num_gfiles == 1 && num_igfiles == 0) { // Loading only one file with no intersection files // Don't need to store a graph in memory, can filter as stream // Don't actually store anything in the de Bruijn graph, but we need to // pass it, so mock one up dBGraph db_graph; db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, file_filter_into_ncols(&gfiles[0].fltr), 0, 1024, 0); graph_writer_stream_mkhdr(out_path, &gfiles[0], &db_graph, NULL, NULL); graph_file_close(&gfiles[0]); gfile_buf_dealloc(&isec_gfiles_buf); ctx_free(gfiles); db_graph_dealloc(&db_graph); return EXIT_SUCCESS; } // // Decide on memory // size_t bits_per_kmer, kmers_in_hash, graph_mem; bits_per_kmer = sizeof(BinaryKmer)*8 + (sizeof(Covg) + sizeof(Edges)) * 8 * use_ncols; kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use, memargs.mem_to_use_set, memargs.num_kmers, memargs.num_kmers_set, bits_per_kmer, ctx_max_kmers, ctx_sum_kmers, true, &graph_mem); if(!use_ncols_set) { // Maximise use_ncols size_t max_usencols = (memargs.mem_to_use*8) / bits_per_kmer; use_ncols = MIN2(max_usencols, ctx_max_cols); bits_per_kmer = sizeof(BinaryKmer)*8 + (sizeof(Covg) + sizeof(Edges)) * 8 * use_ncols; // Re-check memory used kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use, memargs.mem_to_use_set, memargs.num_kmers, memargs.num_kmers_set, bits_per_kmer, ctx_max_kmers, ctx_sum_kmers, true, &graph_mem); } status("Using %zu colour%s in memory", use_ncols, util_plural_str(use_ncols)); cmd_check_mem_limit(memargs.mem_to_use, graph_mem); // Create db_graph dBGraph db_graph; Edges *intersect_edges = NULL; size_t edge_cols = (use_ncols + take_intersect); db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, use_ncols, use_ncols, kmers_in_hash, DBG_ALLOC_COVGS); // We allocate edges ourself since it's a special case db_graph.col_edges = ctx_calloc(db_graph.ht.capacity*edge_cols, sizeof(Edges)); // Load intersection binaries char *intsct_gname_ptr = NULL; StrBuf intersect_gname; strbuf_alloc(&intersect_gname, 1024); if(take_intersect) { GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph); gprefs.boolean_covgs = true; // covg++ only for(i = 0; i < num_igfiles; i++) { graph_load(&igfiles[i], gprefs, NULL); // Update intersect header // note: intersection graphs all load exactly one colour into colour 0 graph_info_make_intersect(&igfiles[i].hdr.ginfo[0], &intersect_gname); gprefs.must_exist_in_graph = true; gprefs.must_exist_in_edges = db_graph.col_edges; } if(num_igfiles > 1) { // Remove nodes where covg != num_igfiles HASH_ITERATE_SAFE(&db_graph.ht, remove_non_intersect_nodes, db_graph.col_covgs, (Covg)num_igfiles, &db_graph.ht); } status("Loaded intersection set\n"); intsct_gname_ptr = intersect_gname.b; for(i = 0; i < num_igfiles; i++) graph_file_close(&igfiles[i]); // Reset graph info for(i = 0; i < db_graph.num_of_cols; i++) graph_info_init(&db_graph.ginfo[i]); // Zero covgs memset(db_graph.col_covgs, 0, db_graph.ht.capacity * sizeof(Covg)); // Use union edges we loaded to intersect new edges intersect_edges = db_graph.col_edges; db_graph.col_edges += db_graph.ht.capacity; } bool kmers_loaded = take_intersect, colours_loaded = false; graph_writer_merge_mkhdr(out_path, gfiles, num_gfiles, kmers_loaded, colours_loaded, intersect_edges, intsct_gname_ptr, &db_graph); if(take_intersect) db_graph.col_edges -= db_graph.ht.capacity; for(i = 0; i < num_gfiles; i++) graph_file_close(&gfiles[i]); strbuf_dealloc(&intersect_gname); gfile_buf_dealloc(&isec_gfiles_buf); ctx_free(gfiles); db_graph_dealloc(&db_graph); return EXIT_SUCCESS; }
int graph_file_open(GraphFileReader *file, const char *path) { return graph_file_open2(file, path, "r", 0); }
static void parse_args(int argc, char **argv) { BuildGraphTask task; memset(&task, 0, sizeof(task)); task.prefs = SEQ_LOADING_PREFS_INIT; task.stats = SEQ_LOADING_STATS_INIT; uint8_t fq_offset = 0; int intocolour = -1; GraphFileReader tmp_gfile; // Arg parsing char cmd[100], shortopts[100]; cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts)); int c; bool sample_named = false, pref_unused = false; while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) { cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd)); switch(c) { case 0: /* flag set */ break; case 'h': cmd_print_usage(NULL); break; case 't': cmd_check(!nthreads,cmd); nthreads = cmd_uint32_nonzero(cmd, optarg); break; case 'm': cmd_mem_args_set_memory(&memargs, optarg); break; case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break; case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break; case 'k': cmd_check(!kmer_size,cmd); kmer_size = cmd_kmer_size(cmd, optarg); break; case 's': intocolour++; check_sample_name(optarg); sample_name_buf_add(&snamebuf, (SampleName){.colour = intocolour, .name = optarg}); sample_named = true; break; case '1': case '2': case 'i': pref_unused = false; if(!sample_named) cmd_print_usage("Please give sample name first [-s,--sample <name>]"); asyncio_task_parse(&task.files, c, optarg, fq_offset, NULL); task.prefs.colour = intocolour; add_task(&task); break; case 'M': if(!strcmp(optarg,"FF")) task.prefs.matedir = READPAIR_FF; else if(!strcmp(optarg,"FR")) task.prefs.matedir = READPAIR_FR; else if(!strcmp(optarg,"RF")) task.prefs.matedir = READPAIR_RF; else if(!strcmp(optarg,"RR")) task.prefs.matedir = READPAIR_RR; else die("-M,--matepair <orient> must be one of: FF,FR,RF,RR"); pref_unused = true; break; case 'O': fq_offset = cmd_uint8(cmd, optarg); pref_unused = true; break; case 'Q': task.prefs.fq_cutoff = cmd_uint8(cmd, optarg); pref_unused = true; break; case 'H': task.prefs.hp_cutoff = cmd_uint8(cmd, optarg); pref_unused = true; break; case 'p': task.prefs.remove_pcr_dups = true; pref_unused = true; break; case 'P': task.prefs.remove_pcr_dups = false; pref_unused = true; break; case 'g': if(intocolour == -1) intocolour = 0; graph_file_reset(&tmp_gfile); graph_file_open2(&tmp_gfile, optarg, "r", true, intocolour); intocolour = MAX2((size_t)intocolour, file_filter_into_ncols(&tmp_gfile.fltr)-1); gfile_buf_push(&gfilebuf, &tmp_gfile, 1); sample_named = false; break; case 'I': graph_file_reset(&tmp_gfile); graph_file_open(&tmp_gfile, optarg); if(file_filter_into_ncols(&tmp_gfile.fltr) > 1) warn("Flattening intersection graph into colour 0: %s", optarg); file_filter_flatten(&tmp_gfile.fltr, 0); gfile_buf_push(&gisecbuf, &tmp_gfile, 1); break; case ':': /* BADARG */ case '?': /* BADCH getopt_long has already printed error */ // cmd_print_usage(NULL); die("`"CMD" build -h` for help. Bad option: %s", argv[optind-1]); default: die("Bad option: %s", cmd); } }
int ctx_infer_edges(int argc, char **argv) { size_t num_of_threads = DEFAULT_NTHREADS; struct MemArgs memargs = MEM_ARGS_INIT; char *out_ctx_path = NULL; bool add_pop_edges = false, add_all_edges = false; // Arg parsing char cmd[100]; char shortopts[100]; cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts)); int c; while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) { cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd)); switch(c) { case 0: /* flag set */ break; case 'h': cmd_print_usage(NULL); break; case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break; case 'o': cmd_check(!out_ctx_path,cmd); out_ctx_path = optarg; break; case 't': num_of_threads = cmd_uint32_nonzero(cmd, optarg); break; case 'm': cmd_mem_args_set_memory(&memargs, optarg); break; case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break; case 'A': add_all_edges = true; break; case 'P': add_pop_edges = true; break; case ':': /* BADARG */ case '?': /* BADCH getopt_long has already printed error */ // cmd_print_usage(NULL); die("`"CMD" inferedges -h` for help. Bad option: %s", argv[optind-1]); default: abort(); } } // Default to adding all edges if(!add_pop_edges && !add_all_edges) add_all_edges = true; // Can only specify one of --pop --all if(add_pop_edges && add_all_edges) cmd_print_usage("Please specify only one of --all --pop"); // Check that optind+1 == argc if(optind+1 > argc) cmd_print_usage("Expected exactly one graph file"); else if(optind+1 < argc) cmd_print_usage("Expected only one graph file. What is this: '%s'", argv[optind]); // // Open graph file // char *graph_path = argv[optind]; status("Reading graph: %s", graph_path); if(strchr(graph_path,':') != NULL) cmd_print_usage("Cannot use ':' in input graph for `"CMD" inferedges`"); GraphFileReader file; memset(&file, 0, sizeof(file)); file_filter_open(&file.fltr, graph_path); // Use stat to detect if we are reading from a stream struct stat st; bool reading_stream = (stat(file.fltr.path.b, &st) != 0); // Mode r+ means open (not create) for update (read & write) graph_file_open2(&file, graph_path, reading_stream ? "r" : "r+", 0); if(!file_filter_is_direct(&file.fltr)) cmd_print_usage("Inferedges with filter not implemented - sorry"); bool editing_file = !(out_ctx_path || reading_stream); FILE *fout = NULL; // Editing input file or writing a new file if(!editing_file) fout = futil_fopen_create(out_ctx_path ? out_ctx_path : "-", "w"); // Print output status if(fout == stdout) status("Writing to STDOUT"); else if(fout != NULL) status("Writing to: %s", out_ctx_path); else status("Editing file in place: %s", graph_path); status("Inferring all missing %sedges", add_pop_edges ? "population " : ""); // // Decide on memory // const size_t ncols = file.hdr.num_of_cols; size_t kmers_in_hash, graph_mem, bits_per_kmer; // reading stream: all covgs + edges // reading file: one bit per kmer per colour for 'in colour' bits_per_kmer = sizeof(BinaryKmer)*8; if(reading_stream) { bits_per_kmer += ncols * 8 * (sizeof(Edges) + sizeof(Covg)); } else { bits_per_kmer += ncols; // in colour } kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use, memargs.mem_to_use_set, memargs.num_kmers, memargs.num_kmers_set, bits_per_kmer, file.num_of_kmers, file.num_of_kmers, memargs.mem_to_use_set, &graph_mem); cmd_check_mem_limit(memargs.mem_to_use, graph_mem); // // Allocate memory // int alloc_flags = reading_stream ? DBG_ALLOC_EDGES | DBG_ALLOC_COVGS : DBG_ALLOC_NODE_IN_COL; dBGraph db_graph; db_graph_alloc(&db_graph, file.hdr.kmer_size, ncols, reading_stream ? ncols : 1, kmers_in_hash, alloc_flags); LoadingStats stats = LOAD_STATS_INIT_MACRO; GraphLoadingPrefs gprefs = {.db_graph = &db_graph, .boolean_covgs = false, .must_exist_in_graph = false, .must_exist_in_edges = NULL, .empty_colours = false}; // We need to load the graph for both --pop and --all since we need to check // if the next kmer is in each of the colours graph_load(&file, gprefs, &stats); if(add_pop_edges) status("Inferring edges from population...\n"); else status("Inferring all missing edges...\n"); size_t num_kmers_edited; if(reading_stream) { ctx_assert(fout != NULL); num_kmers_edited = infer_edges(num_of_threads, add_all_edges, &db_graph); graph_write_header(fout, &file.hdr); graph_write_all_kmers(fout, &db_graph); } else if(fout == NULL) { num_kmers_edited = inferedges_on_mmap(&db_graph, add_all_edges, &file); } else { num_kmers_edited = inferedges_on_file(&db_graph, add_all_edges, &file, fout); } if(fout != NULL && fout != stdout) fclose(fout); char modified_str[100], kmers_str[100]; ulong_to_str(num_kmers_edited, modified_str); ulong_to_str(db_graph.ht.num_kmers, kmers_str); double modified_rate = 0; if(db_graph.ht.num_kmers) modified_rate = (100.0 * num_kmers_edited) / db_graph.ht.num_kmers; status("%s of %s (%.2f%%) nodes modified\n", modified_str, kmers_str, modified_rate); if(editing_file) { // Close and re-open fclose(file.fh); file.fh = NULL; futil_update_timestamp(file.fltr.path.b); } graph_file_close(&file); db_graph_dealloc(&db_graph); return EXIT_SUCCESS; }