void _construct_graph_with_paths(dBGraph *graph,
size_t kmer_size, size_t ncols,
char **seqs, size_t nseqs,
CorrectAlnParam path_params)
{
size_t i;
db_graph_alloc(graph, kmer_size, ncols, ncols, 1024);
// Graph data
graph->bktlocks = ctx_calloc(roundup_bits2bytes(graph->ht.num_of_buckets), 1);
graph->col_edges = ctx_calloc(graph->ht.capacity * ncols, sizeof(Edges));
graph->col_covgs = ctx_calloc(graph->ht.capacity * ncols, sizeof(Covg));
graph->node_in_cols = ctx_calloc(roundup_bits2bytes(graph->ht.capacity) * ncols, 1);
// Path data
path_store_alloc(&graph->pstore, 1024, true, graph->ht.capacity, ncols);
graph->pstore.kmer_locks = ctx_calloc(roundup_bits2bytes(graph->ht.capacity), 1);
// Build graph
for(i = 0; i < nseqs; i++)
build_graph_from_str_mt(graph, 0, seqs[i], strlen(seqs[i]));
graph->num_of_cols_used = MAX2(graph->num_of_cols_used, 1);
GenPathWorker *gen_path_wrkr = gen_paths_workers_alloc(1, graph, NULL);
for(i = 0; i < nseqs; i++)
gen_paths_from_str_mt(gen_path_wrkr, seqs[i], path_params);
gen_paths_workers_dealloc(gen_path_wrkr, 1);
}
void graph_crawler_alloc(GraphCrawler *crawler, const dBGraph *db_graph)
{
ctx_assert(db_graph->node_in_cols != NULL);
size_t ncols = db_graph->num_of_cols;
int *col_paths = ctx_calloc(ncols, sizeof(int));
GCMultiColPath *multicol_paths = ctx_calloc(ncols, sizeof(GCMultiColPath));
GCUniColPath *unicol_paths = ctx_calloc(ncols, sizeof(GCUniColPath));
uint32_t *col_list = ctx_calloc(ncols, sizeof(uint32_t));
GraphCrawler tmp = {.num_paths = 0,
.col_paths = col_paths,
.multicol_paths = multicol_paths,
.unicol_paths = unicol_paths,
.col_list = col_list};
memcpy(crawler, &tmp, sizeof(GraphCrawler));
graph_cache_alloc(&crawler->cache, db_graph);
graph_walker_alloc(&crawler->wlk, db_graph);
rpt_walker_alloc(&crawler->rptwlk, db_graph->ht.capacity, 22); // 4MB
}
void graph_crawler_dealloc(GraphCrawler *crawler)
{
ctx_free(crawler->col_paths);
ctx_free(crawler->multicol_paths);
ctx_free(crawler->unicol_paths);
ctx_free(crawler->col_list);
graph_cache_dealloc(&crawler->cache);
graph_walker_dealloc(&crawler->wlk);
rpt_walker_dealloc(&crawler->rptwlk);
memset(crawler, 0, sizeof(GraphCrawler)); // reset
}
CallDecomp* call_decomp_init(htsFile *vcffh, bcf_hdr_t *vcfhdr)
{
CallDecomp *dc = ctx_calloc(1, sizeof(CallDecomp));
dc->nw_aligner = needleman_wunsch_new();
dc->aln = alignment_create(1024);
dc->scoring = ctx_calloc(1, sizeof(dc->scoring[0]));
scoring_system_default(dc->scoring);
dc->vcffh = vcffh;
dc->vcfhdr = vcfhdr;
dc->v = bcf_init();
strbuf_alloc(&dc->sbuf, 256);
return dc;
}
size_t infer_edges(size_t nthreads, bool add_all_edges, const dBGraph *db_graph)
{
ctx_assert(db_graph->node_in_cols != NULL);
ctx_assert(db_graph->col_edges != NULL);
size_t i, num_nodes_modified = 0;
status("[inferedges] Processing stream");
InferEdgesWorker *wrkrs = ctx_calloc(nthreads, sizeof(InferEdgesWorker));
for(i = 0; i < nthreads; i++) {
InferEdgesWorker tmp = {.threadid = i, .nthreads = nthreads,
.add_all_edges = add_all_edges,
.db_graph = db_graph,
.num_nodes_modified = 0};
memcpy(&wrkrs[i], &tmp, sizeof(InferEdgesWorker));
}
util_run_threads(wrkrs, nthreads, sizeof(InferEdgesWorker),
nthreads, infer_edges_worker);
// Sum up nodes modified
for(i = 0; i < nthreads; i++)
num_nodes_modified += wrkrs[i].num_nodes_modified;
ctx_free(wrkrs);
return num_nodes_modified;
}
// If num_paths != 0, we ensure at least num_paths capacity
// @split_linked_lists whether you intend to have traverse linked list and
// all linked list separate
void gpath_store_alloc(GPathStore *gpstore, size_t ncols, size_t graph_capacity,
size_t num_paths, size_t mem,
bool count_nseen, bool split_linked_lists)
{
memset(gpstore, 0, sizeof(*gpstore));
size_t store_mem = graph_capacity*sizeof(GPath*) * (split_linked_lists ? 2 : 1);
if(store_mem > mem)
die("Need at least %zu bytes (only got %zu bytes)", store_mem, mem);
size_t gpset_mem = mem - store_mem;
// Don't allow resizing - this allows use to the thread safe
// and control memory usage (die when we fill up allowed memory)
if(num_paths)
gpath_set_alloc2(&gpstore->gpset, ncols, num_paths, gpset_mem, false, count_nseen);
else
gpath_set_alloc(&gpstore->gpset, ncols, gpset_mem, false, count_nseen);
gpstore->graph_capacity = graph_capacity;
// paths_traverse is always a subset of paths_all
gpstore->paths_all = ctx_calloc(graph_capacity, sizeof(GPath*));
gpstore->paths_traverse = gpstore->paths_all;
}
static void run_exp_abc(const dBGraph *db_graph, bool prime_AB,
size_t nthreads, size_t num_repeats,
size_t max_AB_dist, bool print_failed_contigs)
{
ExpABCWorker *wrkrs = ctx_calloc(nthreads, sizeof(ExpABCWorker));
size_t i, j;
if(max_AB_dist == 0) max_AB_dist = SIZE_MAX;
for(i = 0; i < nthreads; i++) {
wrkrs[i].colour = 0;
wrkrs[i].nthreads = nthreads;
wrkrs[i].db_graph = db_graph;
wrkrs[i].prime_AB = prime_AB;
wrkrs[i].num_limit = num_repeats / nthreads;
wrkrs[i].max_AB_dist = max_AB_dist;
wrkrs[i].print_failed_contigs = print_failed_contigs;
db_node_buf_alloc(&wrkrs[i].nbuf, 1024);
graph_walker_alloc(&wrkrs[i].gwlk, db_graph);
rpt_walker_alloc(&wrkrs[i].rptwlk, db_graph->ht.capacity, 22); // 4MB
}
util_run_threads(wrkrs, nthreads, sizeof(ExpABCWorker),
nthreads, run_exp_abc_thread);
// Merge results
size_t num_tests = 0, results[NUM_RESULT_VALUES] = {0};
size_t ab_fail_state[GRPHWLK_NUM_STATES] = {0};
size_t bc_fail_state[GRPHWLK_NUM_STATES] = {0};
for(i = 0; i < nthreads; i++) {
num_tests += wrkrs[i].num_tests;
for(j = 0; j < NUM_RESULT_VALUES; j++) results[j] += wrkrs[i].results[j];
for(j = 0; j < GRPHWLK_NUM_STATES; j++) ab_fail_state[j] += wrkrs[i].ab_fail_state[j];
for(j = 0; j < GRPHWLK_NUM_STATES; j++) bc_fail_state[j] += wrkrs[i].bc_fail_state[j];
db_node_buf_dealloc(&wrkrs[i].nbuf);
graph_walker_dealloc(&wrkrs[i].gwlk);
rpt_walker_dealloc(&wrkrs[i].rptwlk);
}
// Print results
char nrunstr[50];
ulong_to_str(num_tests, nrunstr);
status("Ran %s tests with %zu threads", nrunstr, nthreads);
const char *titles[] = {"RES_ABC_SUCCESS", "RES_AB_WRONG",
"RES_AB_FAILED", "RES_BC_WRONG",
"RES_BC_FAILED", "RES_BC_OVERSHOT",
"RES_LOST_IN_RPT", "RES_NO_TRAVERSAL"};
util_print_nums(titles, results, NUM_RESULT_VALUES, 30);
status("AB_FAILED:");
graph_step_print_state_hist(ab_fail_state);
status("BC_FAILED:");
graph_step_print_state_hist(bc_fail_state);
ctx_free(wrkrs);
}
static void supernode_cleaner_alloc(SupernodeCleaner *cl, size_t nthreads,
size_t covg_threshold, size_t min_keep_tip,
uint8_t *keep_flags,
const dBGraph *db_graph)
{
size_t i;
CovgBuffer *cbufs = ctx_calloc(nthreads, sizeof(CovgBuffer));
for(i = 0; i < nthreads; i++)
covg_buf_alloc(&cbufs[i], 1024);
uint64_t *covg_hist_init, *covg_hist_cleaned;
uint64_t *mean_covg_hist_init, *mean_covg_hist_cleaned;
uint64_t *len_hist_init, *len_hist_cleaned;
covg_hist_init = ctx_calloc(DUMP_COVG_ARRSIZE, sizeof(uint64_t));
covg_hist_cleaned = ctx_calloc(DUMP_COVG_ARRSIZE, sizeof(uint64_t));
mean_covg_hist_init = ctx_calloc(DUMP_MEAN_COVG_ARRSIZE, sizeof(uint64_t));
mean_covg_hist_cleaned = ctx_calloc(DUMP_MEAN_COVG_ARRSIZE, sizeof(uint64_t));
len_hist_init = ctx_calloc(DUMP_LEN_ARRSIZE, sizeof(uint64_t));
len_hist_cleaned = ctx_calloc(DUMP_LEN_ARRSIZE, sizeof(uint64_t));
SupernodeCleaner tmp = {.nthreads = nthreads,
.covg_threshold = covg_threshold,
.min_keep_tip = min_keep_tip,
.cbufs = cbufs,
.covg_hist_init = covg_hist_init,
.covg_hist_cleaned = covg_hist_cleaned,
.covg_arrsize = DUMP_COVG_ARRSIZE,
.mean_covg_hist_init = mean_covg_hist_init,
.mean_covg_hist_cleaned = mean_covg_hist_cleaned,
.mean_covg_arrsize = DUMP_MEAN_COVG_ARRSIZE,
.len_hist_init = len_hist_init,
.len_hist_cleaned = len_hist_cleaned,
.len_arrsize = DUMP_LEN_ARRSIZE,
.keep_flags = keep_flags,
.num_tips = 0,
.num_low_covg_snodes = 0,
.num_tip_and_low_snodes = 0,
.num_tip_kmers = 0,
.num_low_covg_snode_kmers = 0,
.num_tip_and_low_snode_kmers = 0,
.db_graph = db_graph};
memcpy(cl, &tmp, sizeof(SupernodeCleaner));
}
static void supernode_cleaner_dealloc(SupernodeCleaner *cl)
{
size_t i;
for(i = 0; i < cl->nthreads; i++)
covg_buf_dealloc(&cl->cbufs[i]);
ctx_free(cl->cbufs);
ctx_free(cl->covg_hist_init);
ctx_free(cl->covg_hist_cleaned);
ctx_free(cl->mean_covg_hist_init);
ctx_free(cl->mean_covg_hist_cleaned);
ctx_free(cl->len_hist_init);
ctx_free(cl->len_hist_cleaned);
memset(cl, 0, sizeof(SupernodeCleaner));
}
static void unitig_cleaner_alloc(UnitigCleaner *cl, size_t nthreads,
size_t covg_threshold, size_t min_keep_tip,
uint8_t *keep_flags,
const dBGraph *db_graph)
{
size_t i;
CovgBuffer *cbufs = ctx_calloc(nthreads, sizeof(CovgBuffer));
for(i = 0; i < nthreads; i++)
covg_buf_alloc(&cbufs[i], 1024);
uint64_t *kmer_covgs_init, *kmer_covgs_clean;
uint64_t *unitig_covgs_init, *unitig_covg_clean;
uint64_t *len_hist_init, *len_hist_clean;
kmer_covgs_init = ctx_calloc(DUMP_COVG_ARRSIZE, sizeof(uint64_t));
kmer_covgs_clean = ctx_calloc(DUMP_COVG_ARRSIZE, sizeof(uint64_t));
unitig_covgs_init = ctx_calloc(DUMP_COVG_ARRSIZE, sizeof(uint64_t));
unitig_covg_clean = ctx_calloc(DUMP_COVG_ARRSIZE, sizeof(uint64_t));
len_hist_init = ctx_calloc(DUMP_LEN_ARRSIZE, sizeof(uint64_t));
len_hist_clean = ctx_calloc(DUMP_LEN_ARRSIZE, sizeof(uint64_t));
UnitigCleaner tmp = {.nthreads = nthreads,
.covg_threshold = covg_threshold,
.min_keep_tip = min_keep_tip,
.cbufs = cbufs,
.kmer_covgs_init = kmer_covgs_init,
.kmer_covgs_clean = kmer_covgs_clean,
.unitig_covgs_init = unitig_covgs_init,
.unitig_covg_clean = unitig_covg_clean,
.covg_arrsize = DUMP_COVG_ARRSIZE,
.len_hist_init = len_hist_init,
.len_hist_clean = len_hist_clean,
.len_arrsize = DUMP_LEN_ARRSIZE,
.keep_flags = keep_flags,
.num_tips = 0,
.num_low_covg_snodes = 0,
.num_tip_and_low_snodes = 0,
.num_tip_kmers = 0,
.num_low_covg_snode_kmers = 0,
.num_tip_and_low_snode_kmers = 0,
.db_graph = db_graph};
memcpy(cl, &tmp, sizeof(UnitigCleaner));
}
static void unitig_cleaner_dealloc(UnitigCleaner *cl)
{
size_t i;
for(i = 0; i < cl->nthreads; i++)
covg_buf_dealloc(&cl->cbufs[i]);
ctx_free(cl->cbufs);
ctx_free(cl->kmer_covgs_init);
ctx_free(cl->kmer_covgs_clean);
ctx_free(cl->unitig_covgs_init);
ctx_free(cl->unitig_covg_clean);
ctx_free(cl->len_hist_init);
ctx_free(cl->len_hist_clean);
memset(cl, 0, sizeof(UnitigCleaner));
}
void hash_table_alloc(HashTable *ht, uint64_t req_capacity)
{
uint64_t num_of_buckets, capacity;
uint8_t bucket_size;
capacity = hash_table_cap(req_capacity, &num_of_buckets, &bucket_size);
uint_fast32_t hash_mask = (uint_fast32_t)(num_of_buckets - 1);
size_t mem = capacity * sizeof(BinaryKmer) +
num_of_buckets * sizeof(uint8_t[2]);
char num_bkts_str[100], bkt_size_str[100], cap_str[100], mem_str[100];
ulong_to_str(num_of_buckets, num_bkts_str);
ulong_to_str(bucket_size, bkt_size_str);
ulong_to_str(capacity, cap_str);
bytes_to_str(mem, 1, mem_str);
status("[hasht] Allocating table with %s entries, using %s", cap_str, mem_str);
status("[hasht] number of buckets: %s, bucket size: %s", num_bkts_str, bkt_size_str);
// calloc is required for bucket_data to set the first element of each bucket
// to the 0th pos
BinaryKmer *table = ctx_calloc(capacity, sizeof(BinaryKmer));
uint8_t (*const buckets)[2] = ctx_calloc(num_of_buckets, sizeof(uint8_t[2]));
HashTable data = {
.table = table,
.num_of_buckets = num_of_buckets,
.hash_mask = hash_mask,
.bucket_size = bucket_size,
.capacity = capacity,
.buckets = buckets,
.num_kmers = 0,
.collisions = {0},
.seed = rand()};
memcpy(ht, &data, sizeof(data));
}
void hash_table_dealloc(HashTable *hash_table)
{
ctx_free(hash_table->table);
ctx_free(hash_table->buckets);
}
// Does not allocate ugraph
void unitig_printer_init(UnitigPrinter *printer, const dBGraph *db_graph,
size_t nthreads, UnitigSyntax syntax, FILE *fout)
{
memset(printer, 0, sizeof(UnitigPrinter));
printer->db_graph = db_graph;
printer->syntax = syntax;
printer->fout = fout;
printer->nthreads = nthreads;
printer->num_unitigs = 0;
printer->visited = ctx_calloc(roundup_bits2bytes(db_graph->ht.capacity), 1);
if(pthread_mutex_init(&printer->outlock, NULL) != 0) die("Mutex init failed");
}
/**
* Save paths to a file.
* @param gzout gzFile to write to
* @param path path of output file
* @param save_path_seq if true, save seq= and juncpos= for links, requires
* exactly one colour in the graph
* @param hdrs is array of JSON headers of input files
*/
void gpath_save(gzFile gzout, const char *path,
size_t nthreads, bool save_path_seq,
const char *cmdstr, cJSON *cmdhdr,
cJSON **hdrs, size_t nhdrs,
const ZeroSizeBuffer *contig_hists, size_t ncols,
dBGraph *db_graph)
{
ctx_assert(nthreads > 0);
ctx_assert(gpath_set_has_nseen(&db_graph->gpstore.gpset));
ctx_assert(ncols == db_graph->gpstore.gpset.ncols);
ctx_assert(!save_path_seq || db_graph->num_of_cols == 1); // save_path => 1 colour
char npaths_str[50];
ulong_to_str(db_graph->gpstore.num_paths, npaths_str);
status("Saving %s paths to: %s", npaths_str, path);
status(" using %zu threads", nthreads);
// Write header
cJSON *json = gpath_save_mkhdr(path, cmdstr, cmdhdr, hdrs, nhdrs,
contig_hists, ncols, db_graph);
json_hdr_gzprint(json, gzout);
cJSON_Delete(json);
// Print comments about the format
gzputs(gzout, ctp_explanation_comment);
// Multithreaded
GPathSaver *wrkrs = ctx_calloc(nthreads, sizeof(GPathSaver));
pthread_mutex_t outlock;
size_t i;
if(pthread_mutex_init(&outlock, NULL) != 0) die("Mutex init failed");
for(i = 0; i < nthreads; i++) {
wrkrs[i] = (GPathSaver){.threadid = i,
.nthreads = nthreads,
.save_seq = save_path_seq,
.gzout = gzout,
.outlock = &outlock,
.db_graph = db_graph};
}
// Iterate over kmers writing paths
util_run_threads(wrkrs, nthreads, sizeof(*wrkrs), nthreads, gpath_save_thread);
pthread_mutex_destroy(&outlock);
ctx_free(wrkrs);
status("[GPathSave] Graph paths saved to %s", path);
}
void gpath_store_split_read_write(GPathStore *gpstore)
{
if(gpstore->paths_traverse == gpstore->paths_all)
{
status("[GPathStore] Creating separate read/write GraphPath linked lists");
if(gpstore->num_paths == 0)
{
gpstore->paths_traverse = NULL;
}
else
{
// Copy current paths over to path set to be updated
status("[GPathStore] (allocating new linked lists)");
size_t mem = gpstore->graph_capacity * sizeof(GPath*);
gpstore->paths_traverse = ctx_calloc(gpstore->graph_capacity, sizeof(GPath*));
memcpy(gpstore->paths_traverse, gpstore->paths_all, mem);
}
}
}
static void pull_out_supernodes(const char **seq, const char **ans, size_t n,
const dBGraph *graph)
{
dBNodeBuffer nbuf;
db_node_buf_alloc(&nbuf, 1024);
// 1. Check pulling out supernodes works for iterating over the graph
uint64_t *visited;
visited = ctx_calloc(roundup_bits2words64(graph->ht.capacity), 8);
HASH_ITERATE(&graph->ht, supernode_from_kmer,
&nbuf, visited, graph, ans, n);
ctx_free(visited);
// 2. Check pulling out supernodes works when we iterate over inputs
size_t i, j, len;
dBNode node;
char tmpstr[SNODEBUF];
for(i = 0; i < n; i++) {
len = strlen(seq[i]);
for(j = 0; j+graph->kmer_size <= len; j++)
{
// Find node
node = db_graph_find_str(graph, seq[i]+j);
TASSERT(node.key != HASH_NOT_FOUND);
// Fetch supernode
db_node_buf_reset(&nbuf);
supernode_find(node.key, &nbuf, graph);
supernode_normalise(nbuf.b, nbuf.len, graph);
// Compare
TASSERT(nbuf.len < SNODEBUF);
db_nodes_to_str(nbuf.b, nbuf.len, graph, tmpstr);
if(strcmp(tmpstr, ans[i]) != 0) {
test_status("Got: %s from ans[i]:%s\n", tmpstr, ans[i]);
}
TASSERT(strcmp(tmpstr, ans[i]) == 0);
}
}
db_node_buf_dealloc(&nbuf);
}
int ctx_correct(int argc, char **argv)
{
size_t i, j;
struct ReadThreadCmdArgs args = READ_THREAD_CMD_ARGS_INIT;
read_thread_args_alloc(&args);
read_thread_args_parse(&args, argc, argv, longopts, true);
GraphFileReader *gfile = &args.gfile;
PathFileBuffer *pfiles = &args.pfiles;
CorrectAlnInputBuffer *inputs = &args.inputs;
size_t ctx_total_cols = gfile->hdr.num_of_cols;
size_t ctx_num_kmers = gfile->num_of_kmers;
if(args.colour > ctx_total_cols)
cmd_print_usage("-c %zu is too big [> %zu]", args.colour, ctx_total_cols);
size_t ctp_usedcols = 0;
for(i = 0; i < pfiles->len; i++) {
if(!file_filter_iscolloaded(&pfiles->data[i].fltr, args.colour)) {
cmd_print_usage("Path file doesn't load into colour %zu: %s",
args.colour, pfiles->data[i].fltr.orig_path.buff);
}
ctp_usedcols = MAX2(ctp_usedcols, path_file_usedcols(&pfiles->data[i]));
}
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem, path_mem, total_mem;
// 1 bit needed per kmer if we need to keep track of noreseed
bits_per_kmer = sizeof(Edges)*8 + ctx_num_kmers + sizeof(uint64_t)*8;
kmers_in_hash = cmd_get_kmers_in_hash2(args.memargs.mem_to_use,
args.memargs.mem_to_use_set,
args.memargs.num_kmers,
args.memargs.num_kmers_set,
bits_per_kmer,
ctx_num_kmers, ctx_num_kmers,
false, &graph_mem);
// Paths memory
path_mem = path_files_mem_required(pfiles->data, pfiles->len, false, false,
ctp_usedcols, 0);
cmd_print_mem(path_mem, "paths");
// Total memory
total_mem = graph_mem + path_mem;
cmd_check_mem_limit(args.memargs.mem_to_use, total_mem);
//
// Check we can read all output files
//
// Open output files
SeqOutput *outputs = ctx_calloc(inputs->len, sizeof(SeqOutput));
bool output_files_exist = false;
for(i = 0; i < inputs->len; i++)
{
CorrectAlnInput *input = &inputs->data[i];
input->crt_params.ctxcol = input->crt_params.ctpcol = args.colour;
SeqOutput *output = &outputs[i];
seq_output_alloc(output);
seq_output_set_paths(output, input->out_base,
async_task_pe_output(&input->files));
input->output = output;
// output check prints warnings and returns true if errors
output_files_exist |= seq_output_files_exist_check(output);
}
// Abandon if some of the output files already exist
if(output_files_exist) die("Output files already exist");
// Attempt to open all files
for(i = 0; i < inputs->len && seq_output_open(&outputs[i]); i++) {}
// Check if something went wrong - if so remove all output files
if(i < inputs->len) {
for(j = 0; j < i; j++) seq_output_delete(&outputs[i]);
die("Couldn't open output files");
}
//
// Allocate memory
//
dBGraph db_graph;
db_graph_alloc(&db_graph, gfile->hdr.kmer_size, ctx_total_cols, 1, kmers_in_hash);
size_t bytes_per_col = roundup_bits2bytes(db_graph.ht.capacity);
db_graph.col_edges = ctx_calloc(db_graph.ht.capacity, sizeof(Edges));
db_graph.node_in_cols = ctx_calloc(bytes_per_col * ctx_total_cols, 1);
// Paths
path_store_alloc(&db_graph.pstore, path_mem, false,
db_graph.ht.capacity, ctp_usedcols);
//
// Load Graph and Path files
//
LoadingStats gstats = 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 = true};
// Load graph, print stats, close file
graph_load(gfile, gprefs, &gstats);
hash_table_print_stats_brief(&db_graph.ht);
graph_file_close(gfile);
// Load path files (does nothing if num_fpiles == 0)
paths_format_merge(pfiles->data, pfiles->len, false, false,
args.num_of_threads, &db_graph);
//
// Run alignment
//
correct_reads(args.num_of_threads, MAX_IO_THREADS,
inputs->data, inputs->len,
&db_graph);
// Close and free output files
for(i = 0; i < inputs->len; i++) seq_output_dealloc(&outputs[i]);
ctx_free(outputs);
read_thread_args_dealloc(&args);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
// Returns 0 on success, otherwise != 0
int ctx_unitigs(int argc, char **argv)
{
size_t nthreads = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
UnitigSyntax syntax = PRINT_FASTA;
bool dot_use_points = false;
// 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 'o': cmd_check(!out_path, cmd); out_path = optarg; 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(!syntax, cmd); syntax = PRINT_FASTA; break;
case 'g': cmd_check(!syntax, cmd); syntax = PRINT_GFA; break;
case 'd': cmd_check(!syntax, cmd); syntax = PRINT_DOT; break;
case 'P': cmd_check(!dot_use_points, cmd); dot_use_points = true; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
die("`"CMD" unitigs -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(dot_use_points && syntax == PRINT_FASTA)
cmd_print_usage("--point is only for use with --dot");
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage(NULL);
size_t i, num_gfiles = (size_t)(argc - optind);
char **gfile_paths = argv + optind;
if(dot_use_points && syntax != PRINT_DOT)
cmd_print_usage("--points only valid with --graphviz / --dot");
ctx_assert(num_gfiles > 0);
// Open graph files
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t ctx_max_kmers = 0, ctx_sum_kmers = 0;
graph_files_open(gfile_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem;
bits_per_kmer = sizeof(BinaryKmer)*8 + sizeof(Edges)*8 + 1;
if(syntax != PRINT_FASTA) bits_per_kmer += sizeof(UnitigEnd) * 8;
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);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
status("Output in %s format to %s\n", syntax_strs[syntax],
futil_outpath_str(out_path));
//
// Open output file
//
// Print to stdout unless --out <out> is specified
FILE *fout = futil_fopen_create(out_path, "w");
//
// Allocate memory
//
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, 1, 1, kmers_in_hash,
DBG_ALLOC_EDGES);
UnitigPrinter printer;
unitig_printer_init(&printer, &db_graph, nthreads, syntax, fout);
if(syntax == PRINT_DOT || syntax == PRINT_GFA)
unitig_graph_alloc(&printer.ugraph, &db_graph);
// Load graphs
GraphLoadingPrefs gprefs = {.db_graph = &db_graph,
.boolean_covgs = false,
.must_exist_in_graph = false,
.empty_colours = false};
for(i = 0; i < num_gfiles; i++) {
file_filter_flatten(&gfiles[i].fltr, 0);
graph_load(&gfiles[i], gprefs, NULL);
graph_file_close(&gfiles[i]);
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
switch(syntax)
{
case PRINT_FASTA:
status("Printing unitgs in FASTA using %zu threads", nthreads);
supernodes_iterate(nthreads, printer.visited, &db_graph,
print_unitig_fasta, &printer);
break;
case PRINT_GFA:
print_gfa_syntax(&printer);
break;
case PRINT_DOT:
print_dot_syntax(&printer, dot_use_points);
break;
default:
die("Invalid print syntax: %i", syntax);
}
char num_unitigs_str[50];
ulong_to_str(printer.num_unitigs, num_unitigs_str);
status("Dumped %s unitigs\n", num_unitigs_str);
fclose(fout);
unitig_printer_destroy(&printer);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
// If there is no 'into' filter (e.g. 0,1:in.ctx 0,1 is 'into filter'),
// load into `into_offset..into_offset+N-1`
void file_filter_set_cols(FileFilter *fltr, size_t filencols, size_t into_offset)
{
size_t i;
const char *path_start_c, *path_end_c;
file_filter_deconstruct_path(fltr->input.b, &path_start_c, &path_end_c);
// This is a hack to get non-const pointer
size_t offset_start = path_start_c - fltr->input.b;
size_t offset_end = path_end_c - fltr->input.b;
char *path_start = fltr->input.b + offset_start;
char *path_end = fltr->input.b + offset_end;
char *from_fltr = (*path_end == ':' ? path_end+1 : NULL);
char *into_fltr = (path_start > fltr->input.b ? fltr->input.b : NULL);
size_t ncols;
if(from_fltr) {
int s = range_get_num(from_fltr, filencols-1);
if(s < 0)
die("Invalid filter path: %s (file cols: %zu)", fltr->input.b, filencols);
ncols = s;
} else {
ncols = filencols;
}
if(into_fltr) {
*(path_start-1) = '\0';
int s = range_get_num(into_fltr, SIZE_MAX);
*(path_start-1) = ':';
if(s < 0 || (s != 1 && (size_t)s != ncols))
die("Invalid filter path: %s (s:%i ncols:%zu)", fltr->input.b, s, ncols);
}
fltr->filencols = filencols;
filter_buf_capacity(&fltr->filter, ncols);
file_filter_num(fltr) = ncols;
size_t *tmp = ctx_calloc(ncols, sizeof(size_t));
if(from_fltr)
{
if(range_parse_array(from_fltr, tmp, filencols-1) == -1)
die("Invalid filter path: %s", fltr->input.b);
for(i = 0; i < ncols; i++)
file_filter_fromcol(fltr, i) = tmp[i];
}
else {
for(i = 0; i < ncols; i++)
file_filter_fromcol(fltr, i) = i;
}
if(into_fltr)
{
*(path_start-1) = '\0';
int s = range_parse_array_fill(into_fltr, tmp, SIZE_MAX, ncols);
*(path_start-1) = ':';
if(s < 0 || (size_t)s != ncols)
die("Invalid filter path: %s (s:%i ncols:%zu)", fltr->input.b, s, ncols);
for(i = 0; i < ncols; i++)
file_filter_intocol(fltr, i) = tmp[i];
}
else {
for(i = 0; i < ncols; i++)
file_filter_intocol(fltr, i) = into_offset + i;
}
ctx_free(tmp);
file_filter_update(fltr);
// for(i = 0; i < file_filter_num(fltr); i++)
// status("%u -> %u", file_filter_fromcol(fltr,i), file_filter_intocol(fltr,i));
}
static void parse_entries(gzFile gzin, FILE *fout)
{
CallFileEntry centry;
call_file_entry_alloc(¢ry);
ChromPosBuffer chrposbuf;
chrompos_buf_alloc(&chrposbuf, 32);
StrBuf tmpbuf, flank3pbuf;
strbuf_alloc(&tmpbuf, 1024);
strbuf_alloc(&flank3pbuf, 1024);
const char *flank5p, *flank3p;
size_t flank5p_len, flank3p_len;
size_t cpy_flnk_5p, cpy_flnk_3p;
const read_t *chrom = NULL;
size_t ref_start = 0, ref_end = 0;
bool mapped = false, fw_strand = false;
const char **genotypes = NULL;
if(!input_bubble_format)
genotypes = ctx_calloc(num_samples, sizeof(char*));
for(; call_file_read(gzin, input_path, ¢ry); num_entries_read++)
{
size_t nlines = call_file_num_lines(¢ry);
ctx_assert2(!(nlines&1) && nlines >= 6, "Too few lines: %zu", nlines);
flank5p = call_file_get_line(¢ry,1);
flank5p_len = call_file_line_len(¢ry,1);
cpy_flnk_5p = cpy_flnk_3p = 0;
// Read a corresponding SAM entry
if(input_bubble_format)
{
// Trim down alleles, add to 3p flank
bubble_trim_alleles(¢ry, &flank3pbuf);
flank3p = flank3pbuf.b;
flank3p_len = flank3pbuf.end;
mapped = sam_fetch_coords(¢ry, flank5p, flank5p_len, flank3p, flank3p_len,
&cpy_flnk_5p, &cpy_flnk_3p,
&chrom, &ref_start, &ref_end, &fw_strand);
}
else {
flank3p = call_file_get_line(¢ry, 3);
flank3p_len = call_file_line_len(¢ry, 3);
mapped = brkpnt_fetch_coords(¢ry, &chrposbuf,
&chrom, &ref_start, &ref_end, &fw_strand,
&cpy_flnk_5p, &cpy_flnk_3p);
}
if(mapped)
{
// Get call id
const char *hdrline = call_file_get_line(¢ry, 0);
char callid[100];
int r = get_callid_str(hdrline, input_bubble_format, callid, sizeof(callid));
if(r == -1) die("Poorly formatted: %s", hdrline);
if(r == -2) die("Call id string is too long: %s", hdrline);
align_entry(¢ry, callid, flank5p, flank5p_len, flank3p, flank3p_len,
cpy_flnk_5p, cpy_flnk_3p,
chrom, ref_start, ref_end, fw_strand,
&tmpbuf, genotypes,
fout);
}
}
ctx_free(genotypes);
call_file_entry_dealloc(¢ry);
chrompos_buf_dealloc(&chrposbuf);
strbuf_dealloc(&tmpbuf);
strbuf_dealloc(&flank3pbuf);
}
int ctx_pop_bubbles(int argc, char **argv)
{
size_t nthreads = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
int32_t max_covg = -1; // max mean coverage to remove <=0 => ignore
int32_t max_klen = -1; // max length (kmers) to remove <=0 => ignore
int32_t max_kdiff = -1; // max diff between bubble branch lengths <0 => ignore
// 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 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); 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 'C': cmd_check(max_covg<0, cmd); max_covg = cmd_uint32(cmd, optarg); break;
case 'L': cmd_check(max_klen<0, cmd); max_klen = cmd_uint32(cmd, optarg); break;
case 'D': cmd_check(max_kdiff<0, cmd); max_kdiff = cmd_uint32(cmd, optarg); break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" pop -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage("Require input graph files (.ctx)");
//
// Open graph files
//
const size_t num_gfiles = argc - optind;
char **graph_paths = argv + optind;
ctx_assert(num_gfiles > 0);
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t i, ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(graph_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
bool reread_graph_to_filter = (num_gfiles == 1 &&
strcmp(file_filter_path(&gfiles[0].fltr),"-") != 0);
if(reread_graph_to_filter) {
file_filter_flatten(&gfiles[0].fltr, 0);
ncols = 1;
}
// Check graphs are compatible
graphs_gpaths_compatible(gfiles, num_gfiles, NULL, 0, -1);
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem;
bits_per_kmer = sizeof(BinaryKmer)*8 +
sizeof(Covg)*8*ncols +
sizeof(Edges)*8*ncols +
2; // 1 bit for visited, 1 for removed
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,
false, &graph_mem);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
// Check out_path is writable
futil_create_output(out_path);
// Allocate memory
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, ncols,
kmers_in_hash, DBG_ALLOC_EDGES | DBG_ALLOC_COVGS);
size_t nkwords = roundup_bits2bytes(db_graph.ht.capacity);
uint8_t *visited = ctx_calloc(1, nkwords);
uint8_t *rmvbits = ctx_calloc(1, nkwords);
//
// Load graphs
//
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
gprefs.empty_colours = true;
for(i = 0; i < num_gfiles; i++) {
graph_load(&gfiles[i], gprefs, NULL);
graph_file_close(&gfiles[i]);
gprefs.empty_colours = false;
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
PopBubblesPrefs prefs = {.max_rmv_covg = max_covg,
.max_rmv_klen = max_klen,
.max_rmv_kdiff = max_kdiff};
size_t npopped = 0;
char npopped_str[50];
status("Popping bubbles...");
npopped = pop_bubbles(&db_graph, nthreads, prefs, visited, rmvbits);
ulong_to_str(npopped, npopped_str);
status("Popped %s bubbles", npopped_str);
size_t nkmers0 = db_graph.ht.num_kmers;
status("Removing nodes...");
for(i = 0; i < nkwords; i++) rmvbits[i] = ~rmvbits[i];
prune_nodes_lacking_flag(nthreads, rmvbits, &db_graph);
size_t nkmers1 = db_graph.ht.num_kmers;
ctx_assert(nkmers1 <= nkmers0);
char nkmers0str[50], nkmers1str[50], ndiffstr[50];
ulong_to_str(nkmers0, nkmers0str);
ulong_to_str(nkmers1, nkmers1str);
ulong_to_str(nkmers0-nkmers1, ndiffstr);
status("Number of kmers %s -> %s (-%s)", nkmers0str, nkmers1str, ndiffstr);
if(reread_graph_to_filter)
{
status("Streaming filtered file to: %s\n", out_path);
GraphFileReader gfile;
memset(&gfile, 0, sizeof(GraphFileReader));
graph_file_open(&gfile, graph_paths[0]);
graph_writer_stream_mkhdr(out_path, &gfile, &db_graph,
db_graph.col_edges, NULL);
graph_file_close(&gfile);
}
else
{
status("Saving to: %s\n", out_path);
graph_writer_save_mkhdr(out_path, &db_graph, CTX_GRAPH_FILEFORMAT, NULL,
0, ncols);
}
ctx_free(visited);
ctx_free(rmvbits);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
static BreakpointCaller* brkpt_callers_new(size_t num_callers,
gzFile gzout,
size_t min_ref_flank,
size_t max_ref_flank,
const KOGraph kograph,
const dBGraph *db_graph)
{
ctx_assert(num_callers > 0);
const size_t ncols = db_graph->num_of_cols;
BreakpointCaller *callers = ctx_malloc(num_callers * sizeof(BreakpointCaller));
pthread_mutex_t *out_lock = ctx_malloc(sizeof(pthread_mutex_t));
if(pthread_mutex_init(out_lock, NULL) != 0) die("mutex init failed");
size_t *callid = ctx_calloc(1, sizeof(size_t));
// Each colour in each caller can have a GraphCache path at once
PathRefRun *path_ref_runs = ctx_calloc(num_callers*MAX_REFRUNS_PER_CALLER(ncols),
sizeof(PathRefRun));
size_t i;
for(i = 0; i < num_callers; i++)
{
BreakpointCaller tmp = {.threadid = i,
.nthreads = num_callers,
.kograph = kograph,
.db_graph = db_graph,
.gzout = gzout,
.out_lock = out_lock,
.callid = callid,
.allele_refs = path_ref_runs,
.flank5p_refs = path_ref_runs+MAX_REFRUNS_PER_ORIENT(ncols),
.min_ref_nkmers = min_ref_flank,
.max_ref_nkmers = max_ref_flank};
memcpy(&callers[i], &tmp, sizeof(BreakpointCaller));
path_ref_runs += MAX_REFRUNS_PER_CALLER(ncols);
db_node_buf_alloc(&callers[i].allelebuf, 1024);
db_node_buf_alloc(&callers[i].flank5pbuf, 1024);
kmer_run_buf_alloc(&callers[i].koruns_5p, 128);
kmer_run_buf_alloc(&callers[i].koruns_5p_ended, 128);
kmer_run_buf_alloc(&callers[i].koruns_3p, 128);
kmer_run_buf_alloc(&callers[i].koruns_3p_ended, 128);
kmer_run_buf_alloc(&callers[i].allele_run_buf, 128);
kmer_run_buf_alloc(&callers[i].flank5p_run_buf, 128);
graph_crawler_alloc(&callers[i].crawlers[0], db_graph);
graph_crawler_alloc(&callers[i].crawlers[1], db_graph);
}
return callers;
}
static void brkpt_callers_destroy(BreakpointCaller *callers, size_t num_callers)
{
size_t i;
for(i = 0; i < num_callers; i++) {
db_node_buf_dealloc(&callers[i].allelebuf);
db_node_buf_dealloc(&callers[i].flank5pbuf);
kmer_run_buf_dealloc(&callers[i].koruns_5p);
kmer_run_buf_dealloc(&callers[i].koruns_5p_ended);
kmer_run_buf_dealloc(&callers[i].koruns_3p);
kmer_run_buf_dealloc(&callers[i].koruns_3p_ended);
kmer_run_buf_dealloc(&callers[i].allele_run_buf);
kmer_run_buf_dealloc(&callers[i].flank5p_run_buf);
graph_crawler_dealloc(&callers[i].crawlers[0]);
graph_crawler_dealloc(&callers[i].crawlers[1]);
}
pthread_mutex_destroy(callers[0].out_lock);
ctx_free(callers[0].out_lock);
ctx_free(callers[0].callid);
ctx_free(callers[0].allele_refs);
ctx_free(callers);
}
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;
}
return job_mem + corrector_mem + junc_mem + packed_mem + sizeof(GenPathWorker);
}
// GenPathWorker stores four buffers of size n
#define gworker_seq_buf(n) (binary_seq_mem(n)*4)
static void _gen_paths_worker_alloc(GenPathWorker *wrkr, dBGraph *db_graph)
{
GenPathWorker tmp = {.db_graph = db_graph};
correct_aln_worker_alloc(&tmp.corrector, true, db_graph);
// Junction data
// only fw arrays are malloc'd, rv point to fw
tmp.junc_arrsize = INIT_BUFLEN;
tmp.pck_fw = ctx_calloc(2, gworker_seq_buf(tmp.junc_arrsize));
tmp.pck_rv = tmp.pck_fw + gworker_seq_buf(tmp.junc_arrsize);
tmp.pos_fw = ctx_malloc(tmp.junc_arrsize * sizeof(*tmp.pos_fw) * 2);
tmp.pos_rv = tmp.pos_fw + tmp.junc_arrsize;
tmp.num_fw = tmp.num_rv = 0;
memcpy(wrkr, &tmp, sizeof(GenPathWorker));
}
static void _gen_paths_worker_dealloc(GenPathWorker *wrkr)
{
correct_aln_worker_dealloc(&wrkr->corrector);
ctx_free(wrkr->pck_fw);
ctx_free(wrkr->pos_fw);
}
BubbleCaller* bubble_callers_new(size_t num_callers,
BubbleCallingPrefs prefs,
gzFile gzout,
const dBGraph *db_graph)
{
ctx_assert(num_callers > 0);
// Max usage is 4 * max_allele_len * cols
size_t i;
size_t max_path_len = MAX2(prefs.max_flank_len, prefs.max_allele_len);
BubbleCaller *callers = ctx_malloc(num_callers * sizeof(BubbleCaller));
pthread_mutex_t *out_lock = ctx_malloc(sizeof(pthread_mutex_t));
if(pthread_mutex_init(out_lock, NULL) != 0) die("mutex init failed");
size_t *num_bubbles_ptr = ctx_calloc(1, sizeof(size_t));
for(i = 0; i < num_callers; i++)
{
BubbleCaller tmp = {.threadid = i, .nthreads = num_callers,
.haploid_seen = ctx_calloc(1+prefs.num_haploid, sizeof(bool)),
.num_bubbles_ptr = num_bubbles_ptr,
.prefs = prefs,
.db_graph = db_graph, .gzout = gzout,
.out_lock = out_lock};
memcpy(&callers[i], &tmp, sizeof(BubbleCaller));
// First two buffers don't actually need to grow
db_node_buf_alloc(&callers[i].flank5p, prefs.max_flank_len);
db_node_buf_alloc(&callers[i].pathbuf, max_path_len);
graph_walker_alloc(&callers[i].wlk, db_graph);
rpt_walker_alloc(&callers[i].rptwlk, db_graph->ht.capacity, 22); // 4MB
graph_cache_alloc(&callers[i].cache, db_graph);
cache_stepptr_buf_alloc(&callers[i].spp_forward, 1024);
cache_stepptr_buf_alloc(&callers[i].spp_reverse, 1024);
strbuf_alloc(&callers[i].output_buf, 2048);
}
return callers;
}
void bubble_callers_destroy(BubbleCaller *callers, size_t num_callers)
{
ctx_assert(num_callers > 0);
size_t i;
for(i = 0; i < num_callers; i++)
{
ctx_free(callers[i].haploid_seen);
db_node_buf_dealloc(&callers[i].flank5p);
db_node_buf_dealloc(&callers[i].pathbuf);
rpt_walker_dealloc(&callers[i].rptwlk);
graph_walker_dealloc(&callers[i].wlk);
graph_cache_dealloc(&callers[i].cache);
cache_stepptr_buf_dealloc(&callers[i].spp_forward);
cache_stepptr_buf_dealloc(&callers[i].spp_reverse);
strbuf_dealloc(&callers[i].output_buf);
}
pthread_mutex_destroy(callers[0].out_lock);
ctx_free(callers[0].out_lock);
ctx_free(callers[0].num_bubbles_ptr);
ctx_free(callers);
}
int ctx_correct(int argc, char **argv)
{
size_t i;
struct ReadThreadCmdArgs args;
read_thread_args_alloc(&args);
read_thread_args_parse(&args, argc, argv, longopts, true);
GraphFileReader *gfile = &args.gfile;
GPathFileBuffer *gpfiles = &args.gpfiles;
CorrectAlnInputBuffer *inputs = &args.inputs;
// Update colours in graph file - sample in 0, all others in 1
size_t ncols = gpath_load_sample_pop(gfile, 1, gpfiles->b, gpfiles->len,
args.colour);
// Check for compatibility between graph files and link files
graphs_gpaths_compatible(gfile, 1, gpfiles->b, gpfiles->len, 1);
int64_t ctx_num_kmers = gfile->num_of_kmers;
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem, path_mem, total_mem;
// 1 bit needed per kmer if we need to keep track of noreseed
bits_per_kmer = sizeof(BinaryKmer)*8 + sizeof(Edges)*8 +
(gpfiles->len > 0 ? sizeof(GPath*)*8 : 0) +
ncols; // in colour
kmers_in_hash = cmd_get_kmers_in_hash(args.memargs.mem_to_use,
args.memargs.mem_to_use_set,
args.memargs.num_kmers,
args.memargs.num_kmers_set,
bits_per_kmer,
ctx_num_kmers, ctx_num_kmers,
false, &graph_mem);
// Paths memory
size_t rem_mem = args.memargs.mem_to_use - MIN2(args.memargs.mem_to_use, graph_mem);
path_mem = gpath_reader_mem_req(gpfiles->b, gpfiles->len, ncols, rem_mem, false,
kmers_in_hash, false);
cmd_print_mem(path_mem, "paths");
// Shift path store memory from graphs->paths
graph_mem -= sizeof(GPath*)*kmers_in_hash;
path_mem += sizeof(GPath*)*kmers_in_hash;
// Total memory
total_mem = graph_mem + path_mem;
cmd_check_mem_limit(args.memargs.mem_to_use, total_mem);
//
// Check we can write all output files
//
// Open output files
SeqOutput *outputs = ctx_calloc(inputs->len, sizeof(SeqOutput));
bool err_occurred = false;
for(i = 0; i < inputs->len && !err_occurred; i++)
{
CorrectAlnInput *input = &inputs->b[i];
// We loaded target colour into colour zero
input->crt_params.ctxcol = input->crt_params.ctpcol = 0;
bool is_pe = asyncio_task_is_pe(&input->files);
err_occurred = !seqout_open(&outputs[i], input->out_base, args.fmt, is_pe);
input->output = &outputs[i];
}
// Abandon if some of the output files already exist
if(err_occurred) {
for(i = 0; i < inputs->len; i++)
seqout_close(&outputs[i], true);
die("Error creating output files");
}
//
// Allocate memory
//
dBGraph db_graph;
db_graph_alloc(&db_graph, gfile->hdr.kmer_size, ncols, 1, kmers_in_hash,
DBG_ALLOC_EDGES | DBG_ALLOC_NODE_IN_COL);
// Create a path store that does not tracks path counts
gpath_reader_alloc_gpstore(gpfiles->b, gpfiles->len, path_mem, false, &db_graph);
//
// Load Graph and link files
//
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
gprefs.empty_colours = true;
// Load graph, print stats, close file
graph_load(gfile, gprefs, NULL);
hash_table_print_stats_brief(&db_graph.ht);
graph_file_close(gfile);
// Load link files
for(i = 0; i < gpfiles->len; i++) {
gpath_reader_load(&gpfiles->b[i], GPATH_DIE_MISSING_KMERS, &db_graph);
gpath_reader_close(&gpfiles->b[i]);
}
//
// Run alignment
//
correct_reads(inputs->b, inputs->len,
args.dump_seq_sizes, args.dump_frag_sizes,
args.fq_zero, args.append_orig_seq,
args.nthreads, &db_graph);
// Close and free output files
for(i = 0; i < inputs->len; i++)
seqout_close(&outputs[i], false);
ctx_free(outputs);
// Closes input files
read_thread_args_dealloc(&args);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int ctx_calls2vcf(int argc, char **argv)
{
const char *in_path = NULL, *out_path = NULL, *out_type = NULL;
// Filtering parameters
int32_t min_mapq = -1, max_align_len = -1, max_allele_len = -1;
// Alignment parameters
int nwmatch = 1, nwmismatch = -2, nwgapopen = -4, nwgapextend = -1;
// ref paths
char const*const* ref_paths = NULL;
size_t nref_paths = 0;
// flank file
const char *sam_path = NULL;
//
// Things we figure out by looking at the input
//
bool isbubble = false;
// samples in VCF, (0 for bubble, does not include ref in breakpoint calls)
size_t i, kmer_size, num_samples;
//
// Reference genome
//
// Hash map of chromosome name -> sequence
ChromHash *genome;
ReadBuffer chroms;
// 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 'O': cmd_check(!out_type, cmd); out_type = optarg; break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'F': cmd_check(!sam_path,cmd); sam_path = optarg; break;
case 'Q': cmd_check(min_mapq < 0,cmd); min_mapq = cmd_uint32(cmd, optarg); break;
case 'A': cmd_check(max_align_len < 0,cmd); max_align_len = cmd_uint32(cmd, optarg); break;
case 'L': cmd_check(max_allele_len < 0,cmd); max_allele_len = cmd_uint32(cmd, optarg); break;
case 'm': nwmatch = cmd_int32(cmd, optarg); break;
case 'M': nwmismatch = cmd_int32(cmd, optarg); break;
case 'g': nwgapopen = cmd_int32(cmd, optarg); break;
case 'G': nwgapextend = cmd_int32(cmd, optarg); break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
die("`"CMD" "SUBCMD" -h` for help. Bad option: %s", argv[optind-1]);
default: ctx_assert2(0, "shouldn't reach here: %c", c);
}
}
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(max_align_len < 0) max_align_len = DEFAULT_MAX_ALIGN;
if(max_allele_len < 0) max_allele_len = DEFAULT_MAX_ALLELE;
if(optind+2 > argc)
cmd_print_usage("Require <in.txt.gz> and at least one reference");
in_path = argv[optind++];
ref_paths = (char const*const*)argv + optind;
nref_paths = argc - optind;
// These functions call die() on error
gzFile gzin = futil_gzopen(in_path, "r");
// Read call file header
cJSON *json = json_hdr_load(gzin, in_path);
// Check we can handle the kmer size
kmer_size = json_hdr_get_kmer_size(json, in_path);
db_graph_check_kmer_size(kmer_size, in_path);
// Get format (bubble or breakpoint file)
cJSON *json_fmt = json_hdr_get(json, "file_format", cJSON_String, in_path);
if(strcmp(json_fmt->valuestring,"CtxBreakpoints") == 0) isbubble = false;
else if(strcmp(json_fmt->valuestring,"CtxBubbles") == 0) isbubble = true;
else die("Unknown format: '%s'", json_fmt->valuestring);
status("Reading %s in %s format", futil_inpath_str(in_path),
isbubble ? "bubble" : "breakpoint");
if(isbubble) {
// bubble specific
if(sam_path == NULL)
cmd_print_usage("Require -F <flanks.sam> with bubble file");
if(min_mapq < 0) min_mapq = DEFAULT_MIN_MAPQ;
}
else {
// breakpoint specific
if(min_mapq >= 0)
cmd_print_usage("-Q,--min-mapq <Q> only valid with bubble calls");
}
// Open flank file if it exists
htsFile *samfh = NULL;
bam_hdr_t *bam_hdr = NULL;
bam1_t *mflank = NULL;
if(sam_path)
{
if((samfh = hts_open(sam_path, "r")) == NULL)
die("Cannot open SAM/BAM %s", sam_path);
// Load BAM header
bam_hdr = sam_hdr_read(samfh);
if(bam_hdr == NULL) die("Cannot load BAM header: %s", sam_path);
mflank = bam_init1();
}
// Output VCF has 0 samples if bubbles file, otherwise has N where N is
// number of samples/colours in the breakpoint graph
size_t num_graph_samples = json_hdr_get_ncols(json, in_path);
size_t num_graph_nonref = json_hdr_get_nonref_ncols(json, in_path);
num_samples = 0;
if(!isbubble) {
// If last colour has "is_ref", drop number of samples by one
num_samples = num_graph_nonref < num_graph_samples ? num_graph_samples-1
: num_graph_samples;
}
//
// Open output file
//
if(!out_path) out_path = "-";
int mode = vcf_misc_get_outtype(out_type, out_path);
futil_create_output(out_path);
htsFile *vcffh = hts_open(out_path, modes_htslib[mode]);
status("[calls2vcf] Reading %s call file with %zu samples",
isbubble ? "Bubble" : "Breakpoint", num_graph_samples);
status("[calls2vcf] %zu sample output to: %s format: %s",
num_samples, futil_outpath_str(out_path), hsmodes_htslib[mode]);
if(isbubble) status("[calls2vcf] min. MAPQ: %i", min_mapq);
status("[calls2vcf] max alignment length: %i", max_align_len);
status("[calls2vcf] max VCF allele length: %i", max_allele_len);
status("[calls2vcf] alignment match:%i mismatch:%i gap open:%i extend:%i",
nwmatch, nwmismatch, nwgapopen, nwgapextend);
// Load reference genome
read_buf_alloc(&chroms, 1024);
genome = chrom_hash_init();
chrom_hash_load(ref_paths, nref_paths, &chroms, genome);
// convert to upper case
char *s;
for(i = 0; i < chroms.len; i++)
for(s = chroms.b[i].seq.b; *s; s++) *s = toupper(*s);
if(!isbubble) brkpnt_check_refs_match(json, genome, in_path);
bcf_hdr_t *vcfhdr = make_vcf_hdr(json, in_path, !isbubble, kmer_size,
ref_paths, nref_paths,
chroms.b, chroms.len);
if(bcf_hdr_write(vcffh, vcfhdr) != 0) die("Cannot write VCF header");
AlignedCall *call = acall_init();
CallDecomp *aligner = call_decomp_init(vcffh, vcfhdr);
scoring_t *scoring = call_decomp_get_scoring(aligner);
scoring_init(scoring, nwmatch, nwmismatch, nwgapopen, nwgapextend,
false, false, 0, 0, 0, 0);
CallFileEntry centry;
call_file_entry_alloc(¢ry);
char kmer_str[50];
sprintf(kmer_str, ";K%zu", kmer_size);
if(isbubble)
{
// Bubble calls
DecompBubble *bubbles = decomp_bubble_init();
// Set scoring for aligning 3' flank
scoring = decomp_bubble_get_scoring(bubbles);
scoring_init(scoring, nwmatch, nwmismatch, nwgapopen, nwgapextend,
true, true, 0, 0, 0, 0);
while(call_file_read(gzin, in_path, ¢ry)) {
do {
if(sam_read1(samfh, bam_hdr, mflank) < 0)
die("We've run out of SAM entries!");
} while(mflank->core.flag & (BAM_FSECONDARY | BAM_FSUPPLEMENTARY));
// Align call
strbuf_reset(&call->info);
decomp_bubble_call(bubbles, genome, kmer_size, min_mapq,
¢ry, mflank, bam_hdr, call);
strbuf_append_str(&call->info, kmer_str);
acall_decompose(aligner, call, max_align_len, max_allele_len);
}
// print bubble stats
DecompBubbleStats *bub_stats = ctx_calloc(1, sizeof(*bub_stats));
decomp_bubble_cpy_stats(bub_stats, bubbles);
print_bubble_stats(bub_stats);
ctx_free(bub_stats);
decomp_bubble_destroy(bubbles);
}
else
{
// Breakpoint calls
DecompBreakpoint *breakpoints = decomp_brkpt_init();
while(call_file_read(gzin, in_path, ¢ry)) {
strbuf_reset(&call->info);
decomp_brkpt_call(breakpoints, genome, num_samples, ¢ry, call);
strbuf_append_str(&call->info, kmer_str);
acall_decompose(aligner, call, max_align_len, max_allele_len);
}
// print bubble stats
DecompBreakpointStats *brk_stats = ctx_calloc(1, sizeof(*brk_stats));
decomp_brkpt_cpy_stats(brk_stats, breakpoints);
print_breakpoint_stats(brk_stats);
ctx_free(brk_stats);
decomp_brkpt_destroy(breakpoints);
}
// Print stats
DecomposeStats *astats = ctx_calloc(1, sizeof(*astats));
call_decomp_cpy_stats(astats, aligner);
print_acall_stats(astats);
ctx_free(astats);
call_file_entry_dealloc(¢ry);
call_decomp_destroy(aligner);
acall_destroy(call);
// Finished - clean up
cJSON_Delete(json);
gzclose(gzin);
bcf_hdr_destroy(vcfhdr);
hts_close(vcffh);
for(i = 0; i < chroms.len; i++) seq_read_dealloc(&chroms.b[i]);
read_buf_dealloc(&chroms);
chrom_hash_destroy(genome);
if(sam_path) {
hts_close(samfh);
bam_hdr_destroy(bam_hdr);
bam_destroy1(mflank);
}
return EXIT_SUCCESS;
}
int ctx_links(int argc, char **argv)
{
size_t limit = 0;
const char *link_out_path = NULL, *csv_out_path = NULL, *plot_out_path = NULL;
const char *thresh_path = NULL, *hist_path = NULL;
size_t hist_distsize = 0, hist_covgsize = 0;
size_t cutoff = 0;
bool clean = false;
// Arg parsing
char cmd[100];
char shortopts[300];
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(!link_out_path, cmd); link_out_path = optarg; break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'l': cmd_check(!csv_out_path, cmd); csv_out_path = optarg; break;
case 'c': cmd_check(!cutoff, cmd); cutoff = cmd_size(cmd, optarg); clean = true; break;
case 'L': cmd_check(!limit, cmd); limit = cmd_size(cmd, optarg); break;
case 'P': cmd_check(!plot_out_path, cmd); plot_out_path = optarg; break;
case 'T': cmd_check(!thresh_path, cmd); thresh_path = optarg; break;
case 'H': cmd_check(!hist_path, cmd); hist_path = optarg; break;
case 'C': cmd_check(!hist_covgsize, cmd); hist_covgsize = cmd_size(cmd, optarg); break;
case 'D': cmd_check(!hist_distsize, cmd); hist_distsize = cmd_size(cmd, optarg); break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" links -h` for help. Bad option: %s", argv[optind-1]);
default: ctx_assert2(0, "shouldn't reach here: %c", c);
}
}
if(hist_distsize && !hist_path) cmd_print_usage("--max-dist without --covg-hist");
if(hist_covgsize && !hist_path) cmd_print_usage("--max-covg without --covg-hist");
// Defaults
if(!hist_distsize) hist_distsize = DEFAULT_MAX_DIST;
if(!hist_covgsize) hist_covgsize = DEFAULT_MAX_COVG;
if(optind + 1 != argc) cmd_print_usage("Wrong number of arguments");
const char *ctp_path = argv[optind];
bool list = (csv_out_path != NULL);
bool plot = (plot_out_path != NULL);
bool save = (link_out_path != NULL);
bool hist_covg = (thresh_path != NULL || hist_path != NULL);
size_t plot_kmer_idx = (limit == 0 ? 0 : limit - 1);
if(clean && !save)
cmd_print_usage("Need to give --out <out.ctp.gz> with --clean");
if(!save && !list && !plot && !hist_covg)
cmd_print_usage("Please specify one of --plot, --list or --clean");
if(link_out_path && hist_covg && strcmp(link_out_path,"-") == 0)
cmd_print_usage("Outputing both cleaning threshold (-T) and links (-o) to STDOUT!");
// Open input file
FILE *list_fh = NULL, *plot_fh = NULL, *link_tmp_fh = NULL;
FILE *thresh_fh = NULL, *hist_fh = NULL;
gzFile link_gz = NULL;
// Check file don't exist or that we can overwrite
// Will ignore if path is null
bool err = false;
err |= futil_check_outfile(csv_out_path);
err |= futil_check_outfile(plot_out_path);
err |= futil_check_outfile(link_out_path);
err |= futil_check_outfile(thresh_path);
err |= futil_check_outfile(hist_path);
if(err) die("Use -f,--force to overwrite files");
StrBuf link_tmp_path;
strbuf_alloc(&link_tmp_path, 1024);
GPathReader ctpin;
memset(&ctpin, 0, sizeof(ctpin));
gpath_reader_open(&ctpin, ctp_path);
size_t ncols = file_filter_into_ncols(&ctpin.fltr);
size_t kmer_size = gpath_reader_get_kmer_size(&ctpin);
cJSON *newhdr = cJSON_Duplicate(ctpin.json, 1);
if(ncols != 1) die("Can only clean a single colour at a time. Sorry.");
uint64_t (*hists)[hist_covgsize] = NULL;
if(hist_covg) {
hists = ctx_calloc(hist_distsize, sizeof(hists[0]));
}
if(hist_path && (hist_fh = futil_fopen_create(hist_path, "w")) == NULL)
die("Cannot open file: %s", hist_path);
if(thresh_path && (thresh_fh = futil_fopen_create(thresh_path, "w")) == NULL)
die("Cannot open file: %s", thresh_path);
if(limit)
status("Limiting to the first %zu kmers", limit);
if(clean)
{
timestamp();
message(" Cleaning coverage below %zu", cutoff);
message("\n");
}
if(save)
{
// Check we can find the fields we need
cJSON *links_json = json_hdr_get(newhdr, "paths", cJSON_Object, link_out_path);
cJSON *nkmers_json = json_hdr_get(links_json, "num_kmers_with_paths", cJSON_Number, link_out_path);
cJSON *nlinks_json = json_hdr_get(links_json, "num_paths", cJSON_Number, link_out_path);
cJSON *nbytes_json = json_hdr_get(links_json, "path_bytes", cJSON_Number, link_out_path);
if(!nkmers_json || !nlinks_json || !nbytes_json)
die("Cannot find required header entries");
// Create a random temporary file
link_tmp_fh = create_tmp_file(&link_tmp_path, link_out_path);
status("Saving output to: %s", link_out_path);
status("Temporary output: %s", link_tmp_path.b);
// Open output file
if((link_gz = futil_gzopen_create(link_out_path, "w")) == NULL)
die("Cannot open output link file: %s", link_out_path);
// Need to open output file first so we can get absolute path
// Update the header to include this command
json_hdr_add_curr_cmd(newhdr, link_out_path);
}
if(list)
{
status("Listing to %s", csv_out_path);
if((list_fh = futil_fopen_create(csv_out_path, "w")) == NULL)
die("Cannot open output CSV file %s", csv_out_path);
// Print csv header
fprintf(list_fh, "SeqLen,Covg\n");
}
if(plot)
{
status("Plotting kmer %zu to %s", plot_kmer_idx, plot_out_path);
if((plot_fh = futil_fopen_create(plot_out_path, "w")) == NULL)
die("Cannot open output .dot file %s", plot_out_path);
}
SizeBuffer countbuf, jposbuf;
size_buf_alloc(&countbuf, 16);
size_buf_alloc(&jposbuf, 1024);
StrBuf kmerbuf, juncsbuf, seqbuf, outbuf;
strbuf_alloc(&kmerbuf, 1024);
strbuf_alloc(&juncsbuf, 1024);
strbuf_alloc(&seqbuf, 1024);
strbuf_alloc(&outbuf, 1024);
bool link_fw;
size_t njuncs;
size_t knum, nlinks, num_links_exp = 0;
LinkTree ltree;
ltree_alloc(<ree, kmer_size);
LinkTreeStats tree_stats;
memset(&tree_stats, 0, sizeof(tree_stats));
size_t init_num_links = 0, num_links = 0;
for(knum = 0; !limit || knum < limit; knum++)
{
ltree_reset(<ree);
if(!gpath_reader_read_kmer(&ctpin, &kmerbuf, &num_links_exp)) break;
ctx_assert2(kmerbuf.end == kmer_size, "Kmer incorrect length %zu != %zu",
kmerbuf.end, kmer_size);
// status("kmer: %s", kmerbuf.b);
for(nlinks = 0;
gpath_reader_read_link(&ctpin, &link_fw, &njuncs,
&countbuf, &juncsbuf,
&seqbuf, &jposbuf);
nlinks++)
{
ltree_add(<ree, link_fw, countbuf.b[0], jposbuf.b,
juncsbuf.b, seqbuf.b);
}
if(nlinks != num_links_exp)
warn("Links count mismatch %zu != %zu", nlinks, num_links_exp);
if(hist_covg)
{
ltree_update_covg_hists(<ree, (uint64_t*)hists,
hist_distsize, hist_covgsize);
}
if(clean)
{
ltree_clean(<ree, cutoff);
}
// Accumulate statistics
ltree_get_stats(<ree, &tree_stats);
num_links = tree_stats.num_links - init_num_links;
init_num_links = tree_stats.num_links;
if(list)
{
ltree_write_list(<ree, &outbuf);
if(fwrite(outbuf.b, 1, outbuf.end, list_fh) != outbuf.end)
die("Cannot write CSV file to: %s", csv_out_path);
strbuf_reset(&outbuf);
}
if(save && num_links)
{
ltree_write_ctp(<ree, kmerbuf.b, num_links, &outbuf);
if(fwrite(outbuf.b, 1, outbuf.end, link_tmp_fh) != outbuf.end)
die("Cannot write ctp file to: %s", link_tmp_path.b);
strbuf_reset(&outbuf);
}
if(plot && knum == plot_kmer_idx)
{
status("Plotting tree...");
ltree_write_dot(<ree, &outbuf);
if(fwrite(outbuf.b, 1, outbuf.end, plot_fh) != outbuf.end)
die("Cannot write plot DOT file to: %s", plot_out_path);
strbuf_reset(&outbuf);
}
}
gpath_reader_close(&ctpin);
cJSON *links_json = json_hdr_get(newhdr, "paths", cJSON_Object, link_out_path);
cJSON *nkmers_json = json_hdr_get(links_json, "num_kmers_with_paths", cJSON_Number, link_out_path);
cJSON *nlinks_json = json_hdr_get(links_json, "num_paths", cJSON_Number, link_out_path);
cJSON *nbytes_json = json_hdr_get(links_json, "path_bytes", cJSON_Number, link_out_path);
status("Number of kmers with links %li -> %zu", nkmers_json->valueint, tree_stats.num_trees_with_links);
status("Number of links %li -> %zu", nlinks_json->valueint, tree_stats.num_links);
status("Number of bytes %li -> %zu", nbytes_json->valueint, tree_stats.num_link_bytes);
if(save)
{
// Update JSON
nkmers_json->valuedouble = nkmers_json->valueint = tree_stats.num_trees_with_links;
nlinks_json->valuedouble = nlinks_json->valueint = tree_stats.num_links;
nbytes_json->valuedouble = nbytes_json->valueint = tree_stats.num_link_bytes;
char *json_str = cJSON_Print(newhdr);
if(gzputs(link_gz, json_str) != (int)strlen(json_str))
die("Cannot write ctp file to: %s", link_out_path);
free(json_str);
gzputs(link_gz, "\n\n");
gzputs(link_gz, ctp_explanation_comment);
gzputs(link_gz, "\n");
fseek(link_tmp_fh, 0, SEEK_SET);
char *tmp = ctx_malloc(4*ONE_MEGABYTE);
size_t s;
while((s = fread(tmp, 1, 4*ONE_MEGABYTE, link_tmp_fh)) > 0) {
if(gzwrite(link_gz, tmp, s) != (int)s)
die("Cannot write to output: %s", link_out_path);
}
ctx_free(tmp);
gzclose(link_gz);
fclose(link_tmp_fh);
}
// Write histogram to file
if(hist_fh)
{
size_t i, j;
fprintf(hist_fh, " ");
for(j = 1; j < hist_covgsize; j++) fprintf(hist_fh, ",covg.%02zu", j);
fprintf(hist_fh, "\n");
for(i = 1; i < hist_distsize; i++) {
fprintf(hist_fh, "dist.%02zu", i);
for(j = 1; j < hist_covgsize; j++) {
fprintf(hist_fh, ",%"PRIu64, hists[i][j]);
}
fprintf(hist_fh, "\n");
}
}
if(thresh_fh)
{
// Use median of first five cutoffs
print_suggest_cutoff(6, hist_covgsize, hists, thresh_fh);
}
if(hist_fh && hist_fh != stdout) fclose(hist_fh);
if(list)
{
fclose(list_fh);
}
if(plot)
{
fclose(plot_fh);
}
ctx_free(hists);
cJSON_Delete(newhdr);
strbuf_dealloc(&link_tmp_path);
ltree_dealloc(<ree);
size_buf_dealloc(&countbuf);
size_buf_dealloc(&jposbuf);
strbuf_dealloc(&kmerbuf);
strbuf_dealloc(&juncsbuf);
strbuf_dealloc(&seqbuf);
strbuf_dealloc(&outbuf);
return EXIT_SUCCESS;
}
int ctx_reads(int argc, char **argv)
{
parse_args(argc, argv);
//
// Open input graphs
//
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t i, ctx_max_kmers = 0, ctx_sum_kmers = 0;
graph_files_open(gfile_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
// Will exit and remove output files on error
inputs_attempt_open();
//
// Calculate memory use
//
size_t kmers_in_hash, graph_mem, bits_per_kmer = sizeof(BinaryKmer)*8;
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);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Set up graph
//
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, 1, 0, kmers_in_hash, 0);
// Load graphs
LoadingStats gstats = LOAD_STATS_INIT_MACRO;
GraphLoadingPrefs gprefs = {.db_graph = &db_graph,
.must_exist_in_graph = false,
.empty_colours = true,
.boolean_covgs = false};
for(i = 0; i < num_gfiles; i++) {
file_filter_flatten(&gfiles[i].fltr, 0);
graph_load(&gfiles[i], gprefs, &gstats);
graph_file_close(&gfiles[i]);
gprefs.empty_colours = false;
}
ctx_free(gfiles);
status("Printing reads that do %stouch the graph\n",
inputs.b[0].invert ? "not " : "");
//
// Filter reads using async io
//
LoadingStats seq_stats = LOAD_STATS_INIT_MACRO;
for(i = 0; i < inputs.len; i++) {
inputs.b[i].stats = &seq_stats;
inputs.b[i].db_graph = &db_graph;
}
// Deal with a set of files at once
size_t start, end;
for(start = 0; start < inputs.len; start += MAX_IO_THREADS)
{
// Can have different numbers of inputs vs threads
end = MIN2(inputs.len, start+MAX_IO_THREADS);
asyncio_run_pool(files.b+start, end-start, filter_reads, NULL, nthreads, 0);
}
size_t total_reads_printed = 0;
size_t total_reads = seq_stats.num_se_reads + seq_stats.num_pe_reads;
for(i = 0; i < inputs.len; i++)
total_reads_printed += inputs.b[i].num_of_reads_printed;
for(i = 0; i < inputs.len; i++) {
seqout_close(&inputs.b[i].seqout, false);
asyncio_task_close(&files.b[i]);
}
aln_reads_buf_dealloc(&inputs);
asyncio_buf_dealloc(&files);
status("Total printed %zu / %zu (%.2f%%) reads\n",
total_reads_printed, total_reads,
total_reads ? (100.0 * total_reads_printed) / total_reads : 0.0);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int ctx_rmsubstr(int argc, char **argv)
{
struct MemArgs memargs = MEM_ARGS_INIT;
size_t kmer_size = 0, nthreads = 0;
const char *output_file = NULL;
seq_format fmt = SEQ_FMT_FASTA;
bool invert = false;
// 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 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'o': cmd_check(!output_file, cmd); output_file = optarg; 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 'k': cmd_check(!kmer_size,cmd); kmer_size = cmd_uint32(cmd, optarg); break;
case 'F': cmd_check(fmt==SEQ_FMT_FASTA, cmd); fmt = cmd_parse_format(cmd, optarg); break;
case 'v': cmd_check(!invert,cmd); invert = true; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
cmd_print_usage("`"CMD" rmsubstr -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
// Defaults
if(!nthreads) nthreads = DEFAULT_NTHREADS;
if(!kmer_size) kmer_size = DEFAULT_KMER;
if(!(kmer_size&1)) cmd_print_usage("Kmer size must be odd");
if(kmer_size < MIN_KMER_SIZE) cmd_print_usage("Kmer size too small (recompile)");
if(kmer_size > MAX_KMER_SIZE) cmd_print_usage("Kmer size too large (recompile?)");
if(optind >= argc)
cmd_print_usage("Please specify at least one input sequence file (.fq, .fq etc.)");
size_t i, num_seq_files = argc - optind;
char **seq_paths = argv + optind;
seq_file_t **seq_files = ctx_calloc(num_seq_files, sizeof(seq_file_t*));
for(i = 0; i < num_seq_files; i++)
if((seq_files[i] = seq_open(seq_paths[i])) == NULL)
die("Cannot read sequence file %s", seq_paths[i]);
// Estimate number of bases
// set to -1 if we cannot calc
int64_t est_num_bases = seq_est_seq_bases(seq_files, num_seq_files);
if(est_num_bases < 0) {
warn("Cannot get file sizes, using pipes");
est_num_bases = memargs.num_kmers * IDEAL_OCCUPANCY;
}
status("[memory] Estimated number of bases: %li", (long)est_num_bases);
// Use file sizes to decide on memory
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem;
bits_per_kmer = sizeof(BinaryKmer)*8 +
sizeof(KONodeList) + sizeof(KOccur) + // see kmer_occur.h
8; // 1 byte per kmer for each base to load sequence files
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,
est_num_bases, est_num_bases,
false, &graph_mem);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Open output file
//
if(output_file == NULL) output_file = "-";
FILE *fout = futil_fopen_create(output_file, "w");
//
// Set up memory
//
dBGraph db_graph;
db_graph_alloc(&db_graph, kmer_size, 1, 0, kmers_in_hash, DBG_ALLOC_BKTLOCKS);
//
// Load reference sequence into a read buffer
//
ReadBuffer rbuf;
read_buf_alloc(&rbuf, 1024);
seq_load_all_reads(seq_files, num_seq_files, &rbuf);
// Check for reads too short
for(i = 0; i < rbuf.len && rbuf.b[i].seq.end >= kmer_size; i++) {}
if(i < rbuf.len)
warn("Reads shorter than kmer size (%zu) will not be filtered", kmer_size);
KOGraph kograph = kograph_create(rbuf.b, rbuf.len, true, 0,
nthreads, &db_graph);
size_t num_reads = rbuf.len, num_reads_printed = 0, num_bad_reads = 0;
// Loop over reads printing those that are not substrings
int ret;
for(i = 0; i < rbuf.len; i++) {
ret = _is_substr(&rbuf, i, &kograph, &db_graph);
if(ret == -1) num_bad_reads++;
else if((ret && invert) || (!ret && !invert)) {
seqout_print_read(&rbuf.b[i], fmt, fout);
num_reads_printed++;
}
}
char num_reads_str[100], num_reads_printed_str[100], num_bad_reads_str[100];
ulong_to_str(num_reads, num_reads_str);
ulong_to_str(num_reads_printed, num_reads_printed_str);
ulong_to_str(num_bad_reads, num_bad_reads_str);
status("Printed %s / %s (%.1f%%) to %s",
num_reads_printed_str, num_reads_str,
!num_reads ? 0.0 : (100.0 * num_reads_printed) / num_reads,
futil_outpath_str(output_file));
if(num_bad_reads > 0) {
status("Bad reads: %s / %s (%.1f%%) - no kmer {ACGT} of length %zu",
num_bad_reads_str, num_reads_str,
(100.0 * num_bad_reads) / num_reads,
kmer_size);
}
fclose(fout);
kograph_dealloc(&kograph);
// Free sequence memory
for(i = 0; i < rbuf.len; i++) seq_read_dealloc(&rbuf.b[i]);
read_buf_dealloc(&rbuf);
ctx_free(seq_files);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int ctx_contigs(int argc, char **argv)
{
size_t nthreads = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
size_t i, contig_limit = 0, colour = 0;
bool cmd_reseed = false, cmd_no_reseed = false; // -r, -R
const char *conf_table_path = NULL; // save confidence table to here
bool use_missing_info_check = true, seed_with_unused_paths = false;
double min_step_confid = -1.0, min_cumul_confid = -1.0; // < 0 => no min
// Read length and expected depth for calculating confidences
size_t genome_size = 0;
seq_file_t *tmp_seed_file = NULL;
SeqFilePtrBuffer seed_buf;
seq_file_ptr_buf_alloc(&seed_buf, 16);
GPathReader tmp_gpfile;
GPathFileBuffer gpfiles;
gpfile_buf_alloc(&gpfiles, 8);
// Arg parsing
char cmd[100], 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 'o': cmd_check(!out_path,cmd); out_path = optarg; 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 'p':
memset(&tmp_gpfile, 0, sizeof(GPathReader));
gpath_reader_open(&tmp_gpfile, optarg);
gpfile_buf_push(&gpfiles, &tmp_gpfile, 1);
break;
case '1':
case 's': // --seed <in.fa>
if((tmp_seed_file = seq_open(optarg)) == NULL)
die("Cannot read --seed file: %s", optarg);
seq_file_ptr_buf_add(&seed_buf, tmp_seed_file);
break;
case 'r': cmd_check(!cmd_reseed,cmd); cmd_reseed = true; break;
case 'R': cmd_check(!cmd_no_reseed,cmd); cmd_no_reseed = true; break;
case 'N':
cmd_check(!contig_limit,cmd);
contig_limit = cmd_uint32_nonzero(cmd, optarg);
break;
case 'c': cmd_check(!colour,cmd); colour = cmd_uint32(cmd, optarg); break;
case 'G': cmd_check(!genome_size,cmd); genome_size = cmd_bases(cmd, optarg); break;
case 'S': cmd_check(!conf_table_path,cmd); conf_table_path = optarg; break;
case 'M': cmd_check(use_missing_info_check,cmd); use_missing_info_check = false; break;
case 'P': cmd_check(!seed_with_unused_paths,cmd); seed_with_unused_paths = true; break;
case 'C':
cmd_check(min_cumul_confid < 0,cmd);
min_cumul_confid = cmd_udouble(cmd,optarg);
if(min_cumul_confid > 1) die("%s must be 0 <= x <= 1", cmd);
break;
case 'T':
cmd_check(min_step_confid < 0,cmd);
min_step_confid = cmd_udouble(cmd,optarg);
if(min_step_confid > 1) die("%s must be 0 <= x <= 1", cmd);
break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
die("`"CMD" contigs -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(cmd_no_reseed && cmd_reseed)
cmd_print_usage("Cannot specify both -r and -R");
if(contig_limit && seed_with_unused_paths)
cmd_print_usage("Cannot combine --ncontigs with --use-seed-paths");
bool sample_with_replacement = cmd_reseed;
// Defaults
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(!seed_buf.len && !contig_limit && sample_with_replacement) {
cmd_print_usage("Please specify one or more of: "
"--no-reseed | --ncontigs | --seed <in.fa>");
}
if(optind >= argc) cmd_print_usage("Require input graph files (.ctx)");
//
// Open graph files
//
const size_t num_gfiles = argc - optind;
char **graph_paths = argv + optind;
ctx_assert(num_gfiles > 0);
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
graph_files_open(graph_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
// char *ctx_path = argv[optind];
//
// Open Graph file
//
// GraphFileReader gfile;
// memset(&gfile, 0, sizeof(GraphFileReader));
// graph_file_open(&gfile, ctx_path);
// Update colours in graph file - sample in 0, all others in 1
// never need more than two colours
ncols = gpath_load_sample_pop(gfiles, num_gfiles,
gpfiles.b, gpfiles.len, colour);
// Check for compatibility between graph files and path files
// pop_colour is colour 1
graphs_gpaths_compatible(gfiles, num_gfiles, gpfiles.b, gpfiles.len, 1);
if(!genome_size)
{
char nk_str[50];
if(ctx_max_kmers <= 0) die("Please pass --genome <G> if streaming");
genome_size = ctx_max_kmers;
ulong_to_str(genome_size, nk_str);
status("Taking number of kmers as genome size: %s", nk_str);
}
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem, path_mem, total_mem;
// 1 bit needed per kmer if we need to keep track of kmer usage
bits_per_kmer = sizeof(BinaryKmer)*8 + sizeof(Edges)*8 + sizeof(GPath*)*8 +
ncols + !sample_with_replacement;
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,
false, &graph_mem);
// Paths memory
size_t rem_mem = memargs.mem_to_use - MIN2(memargs.mem_to_use, graph_mem);
path_mem = gpath_reader_mem_req(gpfiles.b, gpfiles.len, ncols, rem_mem, false);
// Shift path store memory from graphs->paths
graph_mem -= sizeof(GPath*)*kmers_in_hash;
path_mem += sizeof(GPath*)*kmers_in_hash;
cmd_print_mem(path_mem, "paths");
// Total memory
total_mem = graph_mem + path_mem;
cmd_check_mem_limit(memargs.mem_to_use, total_mem);
// Load contig hist distribution from ctp files
ZeroSizeBuffer contig_hist;
memset(&contig_hist, 0, sizeof(contig_hist));
for(i = 0; i < gpfiles.len; i++) {
gpath_reader_load_contig_hist(gpfiles.b[i].json,
gpfiles.b[i].fltr.path.b,
file_filter_fromcol(&gpfiles.b[i].fltr, 0),
&contig_hist);
}
// Calculate confidences, only for one colour
ContigConfidenceTable conf_table;
conf_table_alloc(&conf_table, 1);
conf_table_update_hist(&conf_table, 0, genome_size,
contig_hist.b, contig_hist.len);
if(conf_table_path != NULL) {
conf_table_save(&conf_table, conf_table_path);
}
zsize_buf_dealloc(&contig_hist);
//
// Output file if printing
//
FILE *fout = out_path ? futil_fopen_create(out_path, "w") : NULL;
// Allocate
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, 1, kmers_in_hash,
DBG_ALLOC_EDGES | DBG_ALLOC_NODE_IN_COL);
// Paths
gpath_reader_alloc_gpstore(gpfiles.b, gpfiles.len, path_mem,
false, &db_graph);
uint8_t *visited = NULL;
if(!sample_with_replacement)
visited = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
// Load graph
LoadingStats stats = LOAD_STATS_INIT_MACRO;
GraphLoadingPrefs gprefs = {.db_graph = &db_graph,
.boolean_covgs = false,
.must_exist_in_graph = false,
.empty_colours = true};
for(i = 0; i < num_gfiles; i++) {
graph_load(&gfiles[i], gprefs, &stats);
graph_file_close(&gfiles[i]);
gprefs.empty_colours = false;
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
// Load path files
for(i = 0; i < gpfiles.len; i++) {
gpath_reader_load(&gpfiles.b[i], GPATH_DIE_MISSING_KMERS, &db_graph);
gpath_reader_close(&gpfiles.b[i]);
}
gpfile_buf_dealloc(&gpfiles);
AssembleContigStats assem_stats;
assemble_contigs_stats_init(&assem_stats);
assemble_contigs(nthreads, seed_buf.b, seed_buf.len,
contig_limit, visited,
use_missing_info_check, seed_with_unused_paths,
min_step_confid, min_cumul_confid,
fout, out_path, &assem_stats, &conf_table,
&db_graph, 0); // Sample always loaded into colour zero
if(fout && fout != stdout) fclose(fout);
assemble_contigs_stats_print(&assem_stats);
assemble_contigs_stats_destroy(&assem_stats);
conf_table_dealloc(&conf_table);
for(i = 0; i < seed_buf.len; i++)
seq_close(seed_buf.b[i]);
seq_file_ptr_buf_dealloc(&seed_buf);
ctx_free(visited);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
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;
}
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;
}
