// Creates <base>.tmp.<rand>
static FILE* create_tmp_file(StrBuf *path, const char *base)
{
size_t i;
const size_t attempt_limit = 100;
FILE *fh;
for(i = 0; i < attempt_limit; i++) {
size_t r = rand() % 9999;
strbuf_reset(path);
strbuf_sprintf(path, "%s.tmp.%04zu", base, r);
if(!futil_file_exists(path->b)) break;
}
if(i == attempt_limit)
die("Temporary files already exist (%zu tries): %s", attempt_limit, path->b);
if((fh = futil_fopen_create(path->b, "r+")) == NULL) {
die("Cannot write temporary file: %s [%s]", path->b, strerror(errno));
}
unlink(path->b); // Immediately unlink to hide temp file
return fh;
}
int ctx_clean(int argc, char **argv)
{
size_t nthreads = 0, use_ncols = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_ctx_path = NULL;
bool tip_cleaning = false, supernode_cleaning = false;
size_t min_keep_tip = 0;
Covg threshold = 0, fallback_thresh = 0;
const char *len_before_path = NULL, *len_after_path = NULL;
const char *covg_before_path = NULL, *covg_after_path = NULL;
// 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(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':
if(out_ctx_path != NULL) cmd_print_usage(NULL);
out_ctx_path = optarg;
break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'N': use_ncols = cmd_uint32_nonzero(cmd, optarg); break;
case 't': cmd_check(!nthreads, cmd); nthreads = cmd_uint32_nonzero(cmd, optarg); break;
case 'T':
cmd_check(!tip_cleaning, cmd);
min_keep_tip = cmd_uint32_nonzero(cmd, optarg);
tip_cleaning = true;
break;
case 'S':
cmd_check(!supernode_cleaning, cmd);
if(optarg != NULL) threshold = cmd_uint32_nonzero(cmd, optarg);
supernode_cleaning = true;
break;
case 'B': cmd_check(!fallback_thresh, cmd); fallback_thresh = cmd_uint32_nonzero(cmd, optarg); break;
case 'l': cmd_check(!len_before_path, cmd); len_before_path = optarg; break;
case 'L': cmd_check(!len_after_path, cmd); len_after_path = optarg; break;
case 'c': cmd_check(!covg_before_path, cmd); covg_before_path = optarg; break;
case 'C': cmd_check(!covg_after_path, cmd); covg_after_path = optarg; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" clean -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage("Please give input graph files");
// Default behaviour
if(!tip_cleaning && !supernode_cleaning) {
if(out_ctx_path != NULL)
supernode_cleaning = tip_cleaning = true; // do both
else
warn("No cleaning being done: you did not specify --out <out.ctx>");
}
bool doing_cleaning = (supernode_cleaning || tip_cleaning);
if(doing_cleaning && out_ctx_path == NULL) {
cmd_print_usage("Please specify --out <out.ctx> for cleaned graph");
}
if(!doing_cleaning && (covg_after_path || len_after_path)) {
cmd_print_usage("You gave --len-after <out> / --covg-after <out> without "
"any cleaning (set -s, --supernodes or -t, --tips)");
}
if(doing_cleaning && strcmp(out_ctx_path,"-") != 0 &&
!futil_get_force() && futil_file_exists(out_ctx_path))
{
cmd_print_usage("Output file already exists: %s", out_ctx_path);
}
if(fallback_thresh && !supernode_cleaning)
cmd_print_usage("-B, --fallback <T> without --supernodes");
// Use remaining args as graph files
char **gfile_paths = argv + optind;
size_t i, j, num_gfiles = (size_t)(argc - optind);
// Open graph files
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(gfile_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
size_t kmer_size = gfiles[0].hdr.kmer_size;
// default to one colour for now
if(use_ncols == 0) use_ncols = 1;
// Flatten if we don't have to remember colours / output a graph
if(!doing_cleaning)
{
ncols = use_ncols = 1;
for(i = 0; i < num_gfiles; i++)
file_filter_flatten(&gfiles[i].fltr, 0);
}
if(ncols < use_ncols) {
warn("I only need %zu colour%s ('--ncols %zu' ignored)",
ncols, util_plural_str(ncols), use_ncols);
use_ncols = ncols;
}
char max_kmers_str[100];
ulong_to_str(ctx_max_kmers, max_kmers_str);
status("%zu input graph%s, max kmers: %s, using %zu colours",
num_gfiles, util_plural_str(num_gfiles), max_kmers_str, use_ncols);
// If no arguments given we default to removing tips < 2*kmer_size
if(tip_cleaning && min_keep_tip == 0)
min_keep_tip = 2 * kmer_size;
// Warn if any graph files already cleaned
size_t fromcol, intocol;
ErrorCleaning *cleaning;
for(i = 0; i < num_gfiles; i++) {
for(j = 0; j < file_filter_num(&gfiles[i].fltr); j++) {
fromcol = file_filter_fromcol(&gfiles[i].fltr, j);
cleaning = &gfiles[i].hdr.ginfo[fromcol].cleaning;
if(cleaning->cleaned_snodes && supernode_cleaning) {
warn("%s:%zu already has supernode cleaning with threshold: <%zu",
file_filter_path(&gfiles[i].fltr), fromcol,
(size_t)cleaning->clean_snodes_thresh);
}
if(cleaning->cleaned_tips && tip_cleaning) {
warn("%s:%zu already has had tip cleaned",
file_filter_path(&gfiles[i].fltr), fromcol);
}
}
}
// Print steps
size_t step = 0;
status("Actions:\n");
if(covg_before_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_before_path);
if(len_before_path != NULL)
status("%zu. Saving supernode length distribution to: %s", step++, len_before_path);
if(tip_cleaning)
status("%zu. Cleaning tips shorter than %zu nodes", step++, min_keep_tip);
if(supernode_cleaning && threshold > 0)
status("%zu. Cleaning supernodes with coverage < %u", step++, threshold);
if(supernode_cleaning && threshold <= 0)
status("%zu. Cleaning supernodes with auto-detected threshold", step++);
if(covg_after_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_after_path);
if(len_after_path != NULL)
status("%zu. Saving supernode length distribution to: %s", step++, len_after_path);
//
// Decide memory usage
//
bool all_colours_loaded = (ncols <= use_ncols);
bool use_mem_limit = (memargs.mem_to_use_set && num_gfiles > 1) || !ctx_max_kmers;
size_t kmers_in_hash, bits_per_kmer, graph_mem;
size_t per_kmer_per_col_bits = (sizeof(BinaryKmer)+sizeof(Covg)+sizeof(Edges)) * 8;
size_t pop_edges_per_kmer_bits = (all_colours_loaded ? 0 : sizeof(Edges) * 8);
bits_per_kmer = per_kmer_per_col_bits * use_ncols + pop_edges_per_kmer_bits;
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,
use_mem_limit, &graph_mem);
// Maximise the number of colours we load to fill the mem
size_t max_usencols = (memargs.mem_to_use*8 - pop_edges_per_kmer_bits * kmers_in_hash) /
(per_kmer_per_col_bits * kmers_in_hash);
use_ncols = MIN2(max_usencols, ncols);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Check output files are writable
//
futil_create_output(out_ctx_path);
// Does nothing if arg is NULL
futil_create_output(covg_before_path);
futil_create_output(covg_after_path);
futil_create_output(len_before_path);
futil_create_output(len_after_path);
// Create db_graph
// Load as many colours as possible
// Use an extra set of edge to take intersections
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, use_ncols, use_ncols,
kmers_in_hash, DBG_ALLOC_COVGS);
// Edges is a special case
size_t num_edges = db_graph.ht.capacity * (use_ncols + !all_colours_loaded);
db_graph.col_edges = ctx_calloc(num_edges, sizeof(Edges));
// Load graph into a single colour
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};
// Construct cleaned graph header
GraphFileHeader outhdr;
memset(&outhdr, 0, sizeof(GraphFileHeader));
outhdr.version = CTX_GRAPH_FILEFORMAT;
outhdr.kmer_size = db_graph.kmer_size;
outhdr.num_of_cols = ncols;
outhdr.num_of_bitfields = (db_graph.kmer_size*2+63)/64;
graph_header_alloc(&outhdr, ncols);
// Merge info into header
size_t gcol = 0;
for(i = 0; i < num_gfiles; i++) {
for(j = 0; j < file_filter_num(&gfiles[i].fltr); j++, gcol++) {
fromcol = file_filter_fromcol(&gfiles[i].fltr, j);
intocol = file_filter_intocol(&gfiles[i].fltr, j);
graph_info_merge(&outhdr.ginfo[intocol], &gfiles[i].hdr.ginfo[fromcol]);
}
}
if(ncols > use_ncols) {
graph_files_load_flat(gfiles, num_gfiles, gprefs, &stats);
} else {
for(i = 0; i < num_gfiles; i++)
graph_load(&gfiles[i], gprefs, &stats);
}
char num_kmers_str[100];
ulong_to_str(db_graph.ht.num_kmers, num_kmers_str);
status("Total kmers loaded: %s\n", num_kmers_str);
size_t initial_nkmers = db_graph.ht.num_kmers;
hash_table_print_stats(&db_graph.ht);
uint8_t *visited = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
uint8_t *keep = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
if((supernode_cleaning && threshold <= 0) || covg_before_path || len_before_path)
{
// Get coverage distribution and estimate cleaning threshold
int est_threshold = cleaning_get_threshold(nthreads,
covg_before_path,
len_before_path,
visited, &db_graph);
if(est_threshold < 0) status("Cannot find recommended cleaning threshold");
else status("Recommended cleaning threshold is: %i", est_threshold);
// Use estimated threshold if threshold not set
if(threshold <= 0) {
if(fallback_thresh > 0 && est_threshold < (int)fallback_thresh) {
status("Using fallback threshold: %i", fallback_thresh);
threshold = fallback_thresh;
}
else if(est_threshold >= 0) threshold = est_threshold;
}
}
// Die if we failed to find suitable cleaning threshold
if(supernode_cleaning && threshold <= 0)
die("Need cleaning threshold (--supernodes=<D> or --fallback <D>)");
if(doing_cleaning) {
// Clean graph of tips (if min_keep_tip > 0) and supernodes (if threshold > 0)
clean_graph(nthreads, threshold, min_keep_tip,
covg_after_path, len_after_path,
visited, keep, &db_graph);
}
ctx_free(visited);
ctx_free(keep);
if(doing_cleaning)
{
// Output graph file
Edges *intersect_edges = NULL;
bool kmers_loaded = true;
size_t col, thresh;
// Set output header ginfo cleaned
for(col = 0; col < ncols; col++)
{
cleaning = &outhdr.ginfo[col].cleaning;
cleaning->cleaned_snodes |= supernode_cleaning;
cleaning->cleaned_tips |= tip_cleaning;
// if(tip_cleaning) {
// strbuf_append_str(&outhdr.ginfo[col].sample_name, ".tipclean");
// }
if(supernode_cleaning) {
thresh = cleaning->clean_snodes_thresh;
thresh = cleaning->cleaned_snodes ? MAX2(thresh, (uint32_t)threshold)
: (uint32_t)threshold;
cleaning->clean_snodes_thresh = thresh;
// char name_append[200];
// sprintf(name_append, ".supclean%zu", thresh);
// strbuf_append_str(&outhdr.ginfo[col].sample_name, name_append);
}
}
if(!all_colours_loaded)
{
// We haven't loaded all the colours
// intersect_edges are edges to mask with
// resets graph edges
intersect_edges = db_graph.col_edges;
db_graph.col_edges += db_graph.ht.capacity;
}
// Print stats on removed kmers
size_t removed_nkmers = initial_nkmers - db_graph.ht.num_kmers;
double removed_pct = (100.0 * removed_nkmers) / initial_nkmers;
char removed_str[100], init_str[100];
ulong_to_str(removed_nkmers, removed_str);
ulong_to_str(initial_nkmers, init_str);
status("Removed %s of %s (%.2f%%) kmers", removed_str, init_str, removed_pct);
graph_files_merge(out_ctx_path, gfiles, num_gfiles,
kmers_loaded, all_colours_loaded,
intersect_edges, &outhdr, &db_graph);
// Swap back
if(!all_colours_loaded)
db_graph.col_edges = intersect_edges;
}
ctx_check(db_graph.ht.num_kmers == hash_table_count_kmers(&db_graph.ht));
graph_header_dealloc(&outhdr);
for(i = 0; i < num_gfiles; i++) graph_file_close(&gfiles[i]);
ctx_free(gfiles);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
static void dgen_parse_cmdline(int argc, char **argv)
{
// DEV: could make it optional to write the HLA types at the internal nodes.
// DEV: talk to Gil about generating his approximation to the coalescent
// with recombination - different trees as you track along the viral sequence.
static struct option longopts[] =
{
{"prop_past_sampling", required_argument, NULL, 's'},
{"consensus_fasta", required_argument, NULL, 'c'},
{"root_fasta", required_argument, NULL, 'r'},
{"write_newick_tree", no_argument, NULL, 'n'},
{"write_tree_matrix", no_argument, NULL, 'g'},
{"num_queries", required_argument, NULL, 'q'},
{"prop_query", required_argument, NULL, 'f'},
{NULL, 0, NULL, 0}
};
char shortopts[] = "s:j:c:r:ngq:f:";
int optional;
while((optional = getopt_long(argc, argv, shortopts, longopts, NULL)) != -1) {
switch (optional) {
case 's': past_sampling = atof(optarg); break;
case 'c': cons_path = optarg; break;
case 'r': root_path = optarg; break;
case 'n': write_newick_tree_to_file = true; break;
case 'g': write_tree_to_file = true; break;
case 'q': num_queries = atoi(optarg); break;
case 'f': query_fraction = atof(optarg); break;
default: die("Unknown option: %c", optional);
}
}
if(argc - optind != 6) print_usage();
json_parameters_path = argv[optind];
printf("%s\n", json_parameters_path);
N = atoi(argv[optind+1]);
M = atoi(argv[optind+2]);
if(M > N)
die("Sample size larger than the infected population size: [M=%i, N=%i]", M, N);
if(past_sampling < 0) past_sampling = (Decimal) M/N;
if(past_sampling > 1)
die("Past sampling proportion is greater than 1: [M=%i, N=%i]", M, N);
lambda = atof(argv[optind+3]);
mu_tree = atof(argv[optind+4]);
out_dir = argv[optind+5];
if(!mkpath(out_dir, 0755)) die("Cannot create output dir: %s", out_dir);
time_t rawtime;
struct tm * timeinfo;
time(&rawtime);
timeinfo = localtime(&rawtime);
char date[100];
sprintf(date, "%d_%d_%d_%d_%d_%d", timeinfo->tm_year+1900,
timeinfo->tm_mon+1, timeinfo->tm_mday, timeinfo->tm_hour,
timeinfo->tm_min, timeinfo->tm_sec);
char output_data_path[PATH_MAX+1];
char newick_data_path[PATH_MAX+1];
char summary_path[PATH_MAX+1];
char json_summary_path[PATH_MAX+1];
char simulated_refs_path[PATH_MAX+1];
char simulated_root_path[PATH_MAX+1];
char simulated_queries_path[PATH_MAX+1];
char hla_query_path[PATH_MAX+1];
sprintf(output_data_path, "%s/%s_tree.txt", out_dir, date);
sprintf(newick_data_path, "%s/%s_newick.tre", out_dir, date);
sprintf(summary_path, "%s/%s_run_summary.txt", out_dir, date);
sprintf(json_summary_path, "%s/%s_summary.json", out_dir, date);
sprintf(simulated_refs_path, "%s/%s_simulated_birth_death_refs.fasta", out_dir, date);
sprintf(simulated_root_path, "%s/%s_simulated_birth_death_root.fasta", out_dir, date);
sprintf(simulated_queries_path, "%s/%s_simulated_birth_death_queries.fasta", out_dir, date);
sprintf(hla_query_path, "%s/%s_simulated_hla_queries_birth_death.csv", out_dir, date);
while(futil_file_exists(output_data_path))
{
printf("This filename already exists, wait for a second to change the folder name.\n");
sleep(1);
time(&rawtime);
timeinfo = localtime(&rawtime);
sprintf(date, "%d_%d_%d_%d_%d_%d\n", timeinfo->tm_year+1900,
timeinfo->tm_mon+1, timeinfo->tm_mday, timeinfo->tm_hour,
timeinfo->tm_min, timeinfo->tm_sec);
sprintf(output_data_path, "%s/%s_tree.txt", out_dir, date);
sprintf(newick_data_path, "%s/%s_newick.tre", out_dir, date);
sprintf(summary_path, "%s/%s_summary.txt", out_dir, date);
sprintf(json_summary_path, "%s/%s_summary_dgen.json", out_dir, date);
sprintf(simulated_refs_path, "%s/%s_simulated_birth_death_refs.fasta", out_dir, date);
sprintf(simulated_root_path, "%s/%s_simulated_birth_death_root.fasta", out_dir, date);
sprintf(simulated_queries_path, "%s/%s_simulated_birth_death_queries.fasta", out_dir, date);
sprintf(hla_query_path, "%s/%s_simulated_hla_queries_birth_death.csv", out_dir, date);
}
if(write_tree_to_file == true) printf("Writing tree information to: %s\n", output_data_path);
if(write_newick_tree_to_file == true) printf("Writing Newick tree information to: %s\n", newick_data_path);
printf("Writing summary information to: %s\n", summary_path);
printf("Writing passed .json information to: %s\n", summary_path);
printf("Writing simulated reference sequences to: %s\n", simulated_refs_path);
printf("Writing simulated root sequence to: %s\n", simulated_root_path);
printf("Writing simulated query sequences to: %s\n", simulated_queries_path);
printf("Writing simulated HLA query data to: %s\n", hla_query_path);
if(write_tree_to_file == true) tree_data_file = fopen(output_data_path, "w");
if(write_newick_tree_to_file == true) newick_tree_data_file = fopen(newick_data_path, "w");
simulated_refs_file = fopen(simulated_refs_path, "w");
simulated_root_file = fopen(simulated_root_path, "w");
simulated_queries_file = fopen(simulated_queries_path, "w");
summary_file = fopen(summary_path, "w");
json_summary_file = fopen(json_summary_path, "w");
hla_query_file = fopen(hla_query_path, "w");
if(tree_data_file == NULL && write_tree_to_file == true)
die("Cannot open output file: %s", output_data_path);
if(newick_tree_data_file == NULL && write_newick_tree_to_file == true)
die("Cannot open output file: %s", newick_data_path);
if(summary_file == NULL) die("Cannot open output file: %s", summary_path);
if(json_summary_file == NULL) die("Cannot open output file: %s", json_summary_path);
if(hla_query_file == NULL) die("Cannot open output file: %s", hla_query_path);
if(simulated_refs_file == NULL) die("Cannot open output file: %s", simulated_refs_path);
if(simulated_root_file == NULL) die("Cannot open output file: %s", simulated_root_path);
if(simulated_queries_file == NULL) die("Cannot open output file: %s", simulated_queries_path);
// Add this information to the summary file.
fprintf(summary_file, "Total lineages at present: %i\n"
"Sampled lineages at present: %i\n"
"Historical sampling proportion: "DECPRINT"\n"
"Birth rate in the genealogy (lambda): "DECPRINT"\n"
"Death rate in the genealogy (mu_tree): "DECPRINT"\n",
N, M, past_sampling, lambda, mu_tree);
}
int ctx_clean(int argc, char **argv)
{
size_t nthreads = 0, use_ncols = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_ctx_path = NULL;
int min_keep_tip = -1, unitig_min = -1; // <0 => default, 0 => noclean
uint32_t fallback_thresh = 0;
const char *len_before_path = NULL, *len_after_path = NULL;
const char *covg_before_path = NULL, *covg_after_path = NULL;
// 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(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':
if(out_ctx_path != NULL) cmd_print_usage(NULL);
out_ctx_path = optarg;
break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'N': use_ncols = cmd_uint32_nonzero(cmd, optarg); break;
case 't': cmd_check(!nthreads, cmd); nthreads = cmd_uint32_nonzero(cmd, optarg); break;
case 'T':
cmd_check(min_keep_tip<0, cmd);
min_keep_tip = (optarg != NULL ? (int)cmd_uint32(cmd, optarg) : -1);
break;
case 'S':
case 'U':
cmd_check(unitig_min<0, cmd);
unitig_min = (optarg != NULL ? cmd_uint32(cmd, optarg) : -1);
break;
case 'B': cmd_check(!fallback_thresh, cmd); fallback_thresh = cmd_uint32_nonzero(cmd, optarg); break;
case 'l': cmd_check(!len_before_path, cmd); len_before_path = optarg; break;
case 'L': cmd_check(!len_after_path, cmd); len_after_path = optarg; break;
case 'c': cmd_check(!covg_before_path, cmd); covg_before_path = optarg; break;
case 'C': cmd_check(!covg_after_path, cmd); covg_after_path = optarg; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" clean -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage("Please give input graph files");
bool unitig_cleaning = (unitig_min != 0);
bool tip_cleaning = (min_keep_tip != 0);
bool doing_cleaning = (unitig_cleaning || tip_cleaning);
// If you ever want to estimate cleaning threshold without outputting
// a graph, change this to a warning
if(doing_cleaning && out_ctx_path == NULL) {
cmd_print_usage("Please specify --out <out.ctx> for cleaned graph");
// warn("No cleaning being done: you did not specify --out <out.ctx>");
}
if(!doing_cleaning && (covg_after_path || len_after_path)) {
warn("You gave --len-after <out> / --covg-after <out> without "
"any cleaning (set -U, --unitigs or -t, --tips)");
}
if(doing_cleaning && strcmp(out_ctx_path,"-") != 0 &&
!futil_get_force() && futil_file_exists(out_ctx_path))
{
cmd_print_usage("Output file already exists: %s", out_ctx_path);
}
if(fallback_thresh && !unitig_cleaning)
warn("-B, --fallback <T> without --unitigs");
// Use remaining args as graph files
char **gfile_paths = argv + optind;
size_t i, j, num_gfiles = (size_t)(argc - optind);
// Open graph files
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t col, ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(gfile_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
size_t kmer_size = gfiles[0].hdr.kmer_size;
// default to one colour for now
if(use_ncols == 0) use_ncols = 1;
// Flatten if we don't have to remember colours / output a graph
if(out_ctx_path == NULL)
{
ncols = use_ncols = 1;
for(i = 0; i < num_gfiles; i++)
file_filter_flatten(&gfiles[i].fltr, 0);
}
if(ncols < use_ncols) {
warn("I only need %zu colour%s ('--ncols %zu' ignored)",
ncols, util_plural_str(ncols), use_ncols);
use_ncols = ncols;
}
char max_kmers_str[100];
ulong_to_str(ctx_max_kmers, max_kmers_str);
status("%zu input graph%s, max kmers: %s, using %zu colours",
num_gfiles, util_plural_str(num_gfiles), max_kmers_str, use_ncols);
// If no arguments given we default to removing tips < 2*kmer_size
if(min_keep_tip < 0)
min_keep_tip = 2 * kmer_size;
// Warn if any graph files already cleaned
size_t fromcol;
ErrorCleaning *cleaning;
for(i = 0; i < num_gfiles; i++) {
for(j = 0; j < file_filter_num(&gfiles[i].fltr); j++) {
fromcol = file_filter_fromcol(&gfiles[i].fltr, j);
cleaning = &gfiles[i].hdr.ginfo[fromcol].cleaning;
if(cleaning->cleaned_snodes && unitig_cleaning) {
warn("%s:%zu already has unitig cleaning with threshold: <%zu",
file_filter_path(&gfiles[i].fltr), fromcol,
(size_t)cleaning->clean_snodes_thresh);
}
if(cleaning->cleaned_tips && tip_cleaning) {
warn("%s:%zu already has had tip cleaned",
file_filter_path(&gfiles[i].fltr), fromcol);
}
}
}
// Print steps
size_t step = 0;
status("Actions:\n");
if(covg_before_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_before_path);
if(len_before_path != NULL)
status("%zu. Saving unitig length distribution to: %s", step++, len_before_path);
if(min_keep_tip > 0)
status("%zu. Cleaning tips shorter than %i nodes", step++, min_keep_tip);
if(unitig_min > 0)
status("%zu. Cleaning unitigs with coverage < %i", step++, unitig_min);
if(unitig_min < 0)
status("%zu. Cleaning unitigs with auto-detected threshold", step++);
if(covg_after_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_after_path);
if(len_after_path != NULL)
status("%zu. Saving unitig length distribution to: %s", step++, len_after_path);
//
// Decide memory usage
//
bool all_colours_loaded = (ncols <= use_ncols);
bool use_mem_limit = (memargs.mem_to_use_set && num_gfiles > 1) || !ctx_max_kmers;
size_t kmers_in_hash, bits_per_kmer, graph_mem;
size_t per_col_bits = (sizeof(Covg)+sizeof(Edges)) * 8;
size_t extra_edge_bits = (all_colours_loaded ? 0 : sizeof(Edges) * 8);
bits_per_kmer = sizeof(BinaryKmer)*8 +
per_col_bits * use_ncols +
extra_edge_bits;
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,
use_mem_limit, &graph_mem);
// Maximise the number of colours we load to fill the mem
size_t max_usencols = (memargs.mem_to_use*8 -
sizeof(BinaryKmer)*8*kmers_in_hash +
extra_edge_bits*kmers_in_hash) /
(per_col_bits*kmers_in_hash);
use_ncols = MIN2(max_usencols, ncols);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Check output files are writable
//
futil_create_output(out_ctx_path);
// Does nothing if arg is NULL
futil_create_output(covg_before_path);
futil_create_output(covg_after_path);
futil_create_output(len_before_path);
futil_create_output(len_after_path);
// Create db_graph
// Load as many colours as possible
// Use an extra set of edge to take intersections
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, use_ncols, use_ncols,
kmers_in_hash, DBG_ALLOC_EDGES | DBG_ALLOC_COVGS);
// Extra edges required to hold union of kept edges
Edges *edges_union = NULL;
if(use_ncols < ncols)
edges_union = ctx_calloc(db_graph.ht.capacity, sizeof(Edges));
// Load graph into a single colour
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
// Construct cleaned graph header
GraphFileHeader outhdr;
memset(&outhdr, 0, sizeof(GraphFileHeader));
for(i = 0; i < num_gfiles; i++)
graph_file_merge_header(&outhdr, &gfiles[i]);
if(ncols > use_ncols)
{
db_graph.num_of_cols = db_graph.num_edge_cols = 1;
SWAP(edges_union, db_graph.col_edges);
graphs_load_files_flat(gfiles, num_gfiles, gprefs, NULL);
SWAP(edges_union, db_graph.col_edges);
db_graph.num_of_cols = db_graph.num_edge_cols = use_ncols;
}
else {
for(i = 0; i < num_gfiles; i++)
graph_load(&gfiles[i], gprefs, NULL);
}
char num_kmers_str[100];
ulong_to_str(db_graph.ht.num_kmers, num_kmers_str);
status("Total kmers loaded: %s\n", num_kmers_str);
size_t initial_nkmers = db_graph.ht.num_kmers;
hash_table_print_stats(&db_graph.ht);
uint8_t *visited = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
uint8_t *keep = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
// Always estimate cleaning threshold
// if(unitig_min <= 0 || covg_before_path || len_before_path)
// {
// Get coverage distribution and estimate cleaning threshold
int est_min_covg = cleaning_get_threshold(nthreads,
covg_before_path,
len_before_path,
visited, &db_graph);
if(est_min_covg < 0) status("Cannot find recommended cleaning threshold");
else status("Recommended cleaning threshold is: %i", est_min_covg);
// Use estimated threshold if threshold not set
if(unitig_min < 0) {
if(fallback_thresh > 0 && est_min_covg < (int)fallback_thresh) {
status("Using fallback threshold: %i", fallback_thresh);
unitig_min = fallback_thresh;
}
else if(est_min_covg >= 0) unitig_min = est_min_covg;
}
// }
// Die if we failed to find suitable cleaning threshold
if(unitig_min < 0)
die("Need cleaning threshold (--unitigs=<D> or --fallback <D>)");
// Cleaning parameters should now be set (>0) or turned off (==0)
ctx_assert(unitig_min >= 0);
ctx_assert(min_keep_tip >= 0);
if(unitig_min || min_keep_tip)
{
// Clean graph of tips (if min_keep_tip > 0) and unitigs (if threshold > 0)
clean_graph(nthreads, unitig_min, min_keep_tip,
covg_after_path, len_after_path,
visited, keep, &db_graph);
}
ctx_free(visited);
ctx_free(keep);
if(out_ctx_path != NULL)
{
// Set output header ginfo cleaned
for(col = 0; col < ncols; col++)
{
cleaning = &outhdr.ginfo[col].cleaning;
cleaning->cleaned_snodes |= unitig_cleaning;
cleaning->cleaned_tips |= tip_cleaning;
// if(tip_cleaning) {
// strbuf_append_str(&outhdr.ginfo[col].sample_name, ".tipclean");
// }
if(unitig_cleaning) {
size_t thresh = cleaning->clean_snodes_thresh;
thresh = cleaning->cleaned_snodes ? MAX2(thresh, (uint32_t)unitig_min)
: (uint32_t)unitig_min;
cleaning->clean_snodes_thresh = thresh;
// char name_append[200];
// sprintf(name_append, ".supclean%zu", thresh);
// strbuf_append_str(&outhdr.ginfo[col].sample_name, name_append);
}
}
// Print stats on removed kmers
size_t removed_nkmers = initial_nkmers - db_graph.ht.num_kmers;
double removed_pct = (100.0 * removed_nkmers) / initial_nkmers;
char removed_str[100], init_str[100];
ulong_to_str(removed_nkmers, removed_str);
ulong_to_str(initial_nkmers, init_str);
status("Removed %s of %s (%.2f%%) kmers", removed_str, init_str, removed_pct);
// kmers_loaded=true
graph_writer_merge(out_ctx_path, gfiles, num_gfiles,
true, all_colours_loaded,
edges_union, &outhdr, &db_graph);
}
ctx_check(db_graph.ht.num_kmers == hash_table_count_kmers(&db_graph.ht));
// TODO: report kmer coverage for each sample
graph_header_dealloc(&outhdr);
for(i = 0; i < num_gfiles; i++) graph_file_close(&gfiles[i]);
ctx_free(gfiles);
ctx_free(edges_union);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
// Check file don't exist or that we can overwrite it. Ignores if path is NULL
bool futil_check_outfile(const char *file)
{
bool err = (file && !futil_get_force() && futil_file_exists(file));
if(err) warn("File already exists: %s", file);
return err;
}
