int ctx_infer_edges(int argc, char **argv)
{
size_t num_of_threads = DEFAULT_NTHREADS;
struct MemArgs memargs = MEM_ARGS_INIT;
char *out_ctx_path = NULL;
bool add_pop_edges = false, add_all_edges = false;
// Arg parsing
char cmd[100];
char shortopts[100];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'o': cmd_check(!out_ctx_path,cmd); out_ctx_path = optarg; break;
case 't': num_of_threads = cmd_uint32_nonzero(cmd, optarg); break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'A': add_all_edges = true; break;
case 'P': add_pop_edges = true; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" inferedges -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
// Default to adding all edges
if(!add_pop_edges && !add_all_edges) add_all_edges = true;
// Can only specify one of --pop --all
if(add_pop_edges && add_all_edges)
cmd_print_usage("Please specify only one of --all --pop");
// Check that optind+1 == argc
if(optind+1 > argc)
cmd_print_usage("Expected exactly one graph file");
else if(optind+1 < argc)
cmd_print_usage("Expected only one graph file. What is this: '%s'", argv[optind]);
//
// Open graph file
//
char *graph_path = argv[optind];
status("Reading graph: %s", graph_path);
if(strchr(graph_path,':') != NULL)
cmd_print_usage("Cannot use ':' in input graph for `"CMD" inferedges`");
GraphFileReader file;
memset(&file, 0, sizeof(file));
file_filter_open(&file.fltr, graph_path);
// Use stat to detect if we are reading from a stream
struct stat st;
bool reading_stream = (stat(file.fltr.path.b, &st) != 0);
// Mode r+ means open (not create) for update (read & write)
graph_file_open2(&file, graph_path, reading_stream ? "r" : "r+", 0);
if(!file_filter_is_direct(&file.fltr))
cmd_print_usage("Inferedges with filter not implemented - sorry");
bool editing_file = !(out_ctx_path || reading_stream);
FILE *fout = NULL;
// Editing input file or writing a new file
if(!editing_file)
fout = futil_fopen_create(out_ctx_path ? out_ctx_path : "-", "w");
// Print output status
if(fout == stdout) status("Writing to STDOUT");
else if(fout != NULL) status("Writing to: %s", out_ctx_path);
else status("Editing file in place: %s", graph_path);
status("Inferring all missing %sedges", add_pop_edges ? "population " : "");
//
// Decide on memory
//
const size_t ncols = file.hdr.num_of_cols;
size_t kmers_in_hash, graph_mem, bits_per_kmer;
// reading stream: all covgs + edges
// reading file: one bit per kmer per colour for 'in colour'
bits_per_kmer = sizeof(BinaryKmer)*8;
if(reading_stream) {
bits_per_kmer += ncols * 8 * (sizeof(Edges) + sizeof(Covg));
} else {
bits_per_kmer += ncols; // in colour
}
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
file.num_of_kmers, file.num_of_kmers,
memargs.mem_to_use_set, &graph_mem);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Allocate memory
//
int alloc_flags = reading_stream ? DBG_ALLOC_EDGES | DBG_ALLOC_COVGS
: DBG_ALLOC_NODE_IN_COL;
dBGraph db_graph;
db_graph_alloc(&db_graph, file.hdr.kmer_size,
ncols, reading_stream ? ncols : 1,
kmers_in_hash, alloc_flags);
LoadingStats stats = LOAD_STATS_INIT_MACRO;
GraphLoadingPrefs gprefs = {.db_graph = &db_graph,
.boolean_covgs = false,
.must_exist_in_graph = false,
.must_exist_in_edges = NULL,
.empty_colours = false};
// We need to load the graph for both --pop and --all since we need to check
// if the next kmer is in each of the colours
graph_load(&file, gprefs, &stats);
if(add_pop_edges) status("Inferring edges from population...\n");
else status("Inferring all missing edges...\n");
size_t num_kmers_edited;
if(reading_stream)
{
ctx_assert(fout != NULL);
num_kmers_edited = infer_edges(num_of_threads, add_all_edges, &db_graph);
graph_write_header(fout, &file.hdr);
graph_write_all_kmers(fout, &db_graph);
}
else if(fout == NULL) {
num_kmers_edited = inferedges_on_mmap(&db_graph, add_all_edges, &file);
} else {
num_kmers_edited = inferedges_on_file(&db_graph, add_all_edges, &file, fout);
}
if(fout != NULL && fout != stdout) fclose(fout);
char modified_str[100], kmers_str[100];
ulong_to_str(num_kmers_edited, modified_str);
ulong_to_str(db_graph.ht.num_kmers, kmers_str);
double modified_rate = 0;
if(db_graph.ht.num_kmers)
modified_rate = (100.0 * num_kmers_edited) / db_graph.ht.num_kmers;
status("%s of %s (%.2f%%) nodes modified\n",
modified_str, kmers_str, modified_rate);
if(editing_file)
{
// Close and re-open
fclose(file.fh);
file.fh = NULL;
futil_update_timestamp(file.fltr.path.b);
}
graph_file_close(&file);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int main(int argc, char **argv)
{
cortex_init();
cmd_init(argc, argv);
ctx_msg_out = NULL;
ctx_tst_out = stdout;
test_status("Tests running k=%i..%i...", get_min_kmer_size(), get_max_kmer_size());
test_status("[version] "VERSION_STATUS_STR"\n");
// Binary Kmer tests should work for all values of MAXK
test_bkmer_functions();
test_hash_table();
#if MAX_KMER_SIZE == 31
// not kmer dependent
test_util();
test_dna_functions();
test_binary_seq_functions();
// only written in k=31
test_db_node();
test_build_graph();
test_db_unitig();
test_subgraph();
test_cleaning();
test_paths();
// test_path_sets(); // TODO: replace with test_path_subset()
test_graph_walker();
test_corrected_aln();
test_repeat_walker();
test_graph_crawler();
test_bubble_caller();
test_kmer_occur();
test_infer_edges_tests();
#endif
cmd_destroy();
// Check we free'd all our memory
size_t still_alloced = alloc_get_num_allocs() - alloc_get_num_frees();
TASSERT2(still_alloced == 0, "%zu not free'd", still_alloced);
// Finished
char num_test_str[100], num_passed_str[100];
size_t tests_num_passed = tests_num_run - tests_num_failed;
ulong_to_str(tests_num_run, num_test_str);
ulong_to_str(tests_num_passed, num_passed_str);
test_status("Tests passed: %s / %s (%.1f%%)", num_passed_str, num_test_str,
(100.0*tests_num_passed)/tests_num_run);
if(tests_num_failed) test_status("%zu tests failed", tests_num_failed);
else test_status("All tests passed.");
cortex_destroy();
// Return 1 if any tests failed, 0 on success
return tests_num_failed ? 1 : 0;
}
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;
size_t mem_to_use = memargs.mem_to_use;
bits_per_kmer = sizeof(BinaryKmer)*8 +
sizeof(KONodeList) + sizeof(KOccur); // see kmer_occur.h
if(mem_to_use < (size_t)est_num_bases) {
warn("You probably need at least %zu bytes (> %zu)",
(size_t)est_num_bases, memargs.mem_to_use);
}
else {
mem_to_use -= est_num_bases;
}
kmers_in_hash = cmd_get_kmers_in_hash(mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
0, est_num_bases,
true, &graph_mem);
// 1 byte per kmer for each base to load sequence files
size_t total_mem = kmers_in_hash*bits_per_kmer/8 + est_num_bases;
char memstr[50];
bytes_to_str(total_mem, 1, memstr);
status("[memory] total mem with input: %s\n", memstr);
cmd_check_mem_limit(memargs.mem_to_use, total_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_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 ? 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;
}
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_view(int argc, char **argv)
{
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// TODO:
// print_action actions[argc];
// bool read_kmers = false;
// 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 ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
cmd_print_usage("`"CMD" "SUBCMD" -h` for help. Bad option: %s", argv[optind-1]);
default:
cmd_print_usage("Programmer fail. Tell Isaac.");
}
}
if(print_kmers) parse_kmers = 1;
bool no_flags = (!print_info && !parse_kmers && !print_kmers);
if(no_flags) { print_info = parse_kmers = 1; }
if(optind+1 != argc) cmd_print_usage("Require one input graph file (.ctx)");
char *path = argv[optind];
size_t num_errors = 0, num_warnings = 0;
GraphFileReader gfile;
memset(&gfile, 0, sizeof(gfile));
int ret = graph_file_open(&gfile, path);
if(ret == 0) die("Cannot open file: %s", path);
if(print_info)
{
char fsize_str[50];
bytes_to_str((size_t)gfile.file_size, 0, fsize_str);
printf("Loading file: %s\n", file_filter_path(&gfile.fltr));
printf("File size: %s\n", fsize_str);
printf("----\n");
}
size_t i, col, ncols = file_filter_into_ncols(&gfile.fltr);
size_t kmer_size = gfile.hdr.kmer_size;
ctx_assert(ncols > 0);
GraphFileHeader hdr;
memset(&hdr, 0, sizeof(hdr));
graph_file_merge_header(&hdr, &gfile);
uint64_t nkmers_read = 0, nkmers_loaded = 0;
uint64_t num_all_zero_kmers = 0, num_zero_covg_kmers = 0;
uint64_t *col_nkmers, *col_sum_covgs;
col_nkmers = ctx_calloc(ncols, sizeof(col_nkmers[0]));
col_sum_covgs = ctx_calloc(ncols, sizeof(col_sum_covgs[0]));
// Print header
if(print_info) print_header(&hdr, gfile.num_of_kmers);
BinaryKmer bkmer;
Covg covgs[ncols], keep_kmer;
Edges edges[ncols];
bool direct_read = file_filter_is_direct(&gfile.fltr);
if(parse_kmers || print_kmers)
{
if(print_info && print_kmers) printf("----\n");
for(; graph_file_read_reset(&gfile, &bkmer, covgs, edges); nkmers_read++)
{
// If kmer has no covg in any samples -> don't load
keep_kmer = 0;
for(col = 0; col < ncols; col++) {
col_nkmers[col] += (covgs[col] > 0);
col_sum_covgs[col] += covgs[col];
keep_kmer |= covgs[col];
}
if(!direct_read && !keep_kmer) continue;
nkmers_loaded++;
/* Kmer Checks */
// graph_file_read_reset() already checks for:
// 1. oversized kmers
// 2. kmers with covg 0 in all colours
// 3. edges without coverage in a colour
// Check for all-zeros (i.e. all As kmer: AAAAAA)
uint64_t kmer_words_or = 0;
for(i = 0; i < hdr.num_of_bitfields; i++)
kmer_words_or |= bkmer.b[i];
if(kmer_words_or == 0)
{
if(num_all_zero_kmers == 1)
{
loading_error("more than one all 'A's kmers seen [index: %"PRIu64"]\n",
nkmers_read);
}
num_all_zero_kmers++;
}
// Check covg is 0 for all colours
for(i = 0; i < ncols && covgs[i] == 0; i++);
num_zero_covg_kmers += (i == ncols);
// Print
if(print_kmers)
db_graph_print_kmer2(bkmer, covgs, edges, ncols, kmer_size, stdout);
}
}
// check for various reading errors
// if(errno != 0)
// loading_error("errno set [%i]: %s\n", (int)errno, strerror(errno));
int err = ferror(gfile.fh);
if(err != 0)
loading_error("occurred after file reading [%i]\n", err);
char nstr[50];
if(print_kmers || parse_kmers)
{
// file_size is set to -1 if we are reading from a stream,
// therefore won't be able to check number of kmers read
if(gfile.file_size != -1 && nkmers_read != (uint64_t)gfile.num_of_kmers) {
loading_warning("Expected %zu kmers, read %zu\n",
(size_t)gfile.num_of_kmers, (size_t)nkmers_read);
}
if(num_all_zero_kmers > 1)
{
loading_error("%s all-zero-kmers seen\n",
ulong_to_str(num_all_zero_kmers, nstr));
}
if(num_zero_covg_kmers > 0)
{
loading_warning("%s kmers have no coverage in any colour\n",
ulong_to_str(num_zero_covg_kmers, nstr));
}
}
// Count warnings printed by graph_file_reader.c
num_warnings += gfile.error_zero_covg;
num_warnings += gfile.error_missing_covg;
// Can only print these stats if we're read in the kmers
if((print_kmers || parse_kmers) && print_info)
{
// print kmer coverage per sample
printf("\n---- Per colour stats\n");
printf("num. kmers:");
for(col = 0; col < ncols; col++)
printf("\t%s", ulong_to_str(col_nkmers[col], nstr));
printf("\n");
printf("sum coverage:");
for(col = 0; col < ncols; col++)
printf("\t%s", ulong_to_str(col_sum_covgs[col], nstr));
printf("\n");
printf("kmer coverage:");
for(col = 0; col < ncols; col++)
printf("\t%.2f", safe_frac(col_sum_covgs[col], col_nkmers[col]));
printf("\n");
// Overall stats
uint64_t sum_covgs = 0;
double mean_kmer_covg = 0.0;
for(col = 0; col < ncols; col++) sum_covgs += col_sum_covgs[col];
mean_kmer_covg = nkmers_loaded ? (double)sum_covgs / nkmers_loaded : 0.0;
printf("\n---- Overall stats\n");
printf("Total kmers: %s\n", ulong_to_str(nkmers_loaded, nstr));
printf("Total coverage: %s\n", ulong_to_str(sum_covgs, nstr));
printf("Mean coverage: %s\n", double_to_str(mean_kmer_covg, 2, nstr));
}
if(print_info)
{
// Print memory stats
uint64_t mem, capacity, num_buckets, req_capacity;
uint8_t bucket_size;
req_capacity = (size_t)(gfile.num_of_kmers / IDEAL_OCCUPANCY);
capacity = hash_table_cap(req_capacity, &num_buckets, &bucket_size);
mem = ht_mem(bucket_size, num_buckets,
sizeof(BinaryKmer)*8 + ncols*(sizeof(Covg)+sizeof(Edges))*8);
char memstr[100], capacitystr[100], bucket_size_str[100], num_buckets_str[100];
bytes_to_str(mem, 1, memstr);
ulong_to_str(capacity, capacitystr);
ulong_to_str(bucket_size, bucket_size_str);
ulong_to_str(num_buckets, num_buckets_str);
size_t mem_height = (size_t)__builtin_ctzl(num_buckets);
printf("\n---- Memory\n");
printf("memory required: %s [capacity: %s]\n", memstr, capacitystr);
printf(" bucket size: %s; number of buckets: %s\n",
bucket_size_str, num_buckets_str);
printf(" --kmer_size %zu --mem_height %zu --mem_width %i\n",
kmer_size, mem_height, bucket_size);
}
if((print_kmers || parse_kmers) && print_info)
{
printf("\n----\n");
if(num_warnings > 0 || num_errors > 0) {
printf("Warnings: %zu; Errors: %zu\n",
(size_t)num_warnings, (size_t)num_errors);
}
if(num_errors == 0)
printf(num_warnings ? "Graph may be ok\n" : "Graph is valid\n");
}
ctx_free(col_nkmers);
ctx_free(col_sum_covgs);
// Close file (which zeros it)
graph_file_close(&gfile);
graph_header_dealloc(&hdr);
return num_errors ? EXIT_FAILURE : EXIT_SUCCESS;
}
void seq_parse_pe_sf(seq_file_t *sf1, seq_file_t *sf2, uint8_t ascii_fq_offset,
read_t *r1, read_t *r2,
void (*read_func)(read_t *_r1, read_t *_r2,
uint8_t _qoffset1, uint8_t _qoffset2,
void *_ptr),
void *reader_ptr)
{
if(sf2 == NULL) {
seq_parse_se_sf(sf1, ascii_fq_offset, r1, read_func, reader_ptr);
return;
}
status("[seq] Parsing sequence files %s %s\n",
futil_inpath_str(sf1->path), futil_inpath_str(sf2->path));
// Guess offset if needed
uint8_t qoffset1 = ascii_fq_offset, qoffset2 = ascii_fq_offset;
uint8_t qmin1 = ascii_fq_offset, qmin2 = ascii_fq_offset;
uint8_t qmax1 = 126, qmax2 = 126;
if(ascii_fq_offset == 0)
{
int fmt1, fmt2;
if((fmt1 = guess_fastq_format(sf1)) != -1) {
qmin1 = (uint8_t)FASTQ_MIN[fmt1];
qmax1 = (uint8_t)FASTQ_MAX[fmt1];
qoffset1 = (uint8_t)FASTQ_OFFSET[fmt1];
}
if((fmt2 = guess_fastq_format(sf2)) != -1) {
qmin2 = (uint8_t)FASTQ_MIN[fmt2];
qmax2 = (uint8_t)FASTQ_MAX[fmt2];
qoffset2 = (uint8_t)FASTQ_OFFSET[fmt2];
}
}
// warn_flags keeps track of which of the error msgs have been printed
// (only print each error msg once per file)
uint8_t warn_flags = 0;
int success1, success2;
size_t num_pe_pairs = 0;
while(1)
{
success1 = seq_read_primary(sf1, r1);
success2 = seq_read_primary(sf2, r2);
if(success1 < 0) warn("input error: %s", sf1->path);
if(success2 < 0) warn("input error: %s", sf2->path);
if(!success1 != !success2) {
warn("Different number of reads in pe files [%s; %s]\n",
sf1->path, sf2->path);
}
if(success1 <= 0 || success2 <= 0) break;
// PE
// We don't care about read orientation at this point
warn_flags = check_new_read(r1, qmin1, qmax1, sf1->path, warn_flags);
warn_flags = check_new_read(r2, qmin2, qmax2, sf2->path, warn_flags);
read_func(r1, r2, qoffset1, qoffset2, reader_ptr);
num_pe_pairs++;
}
char num_pe_pairs_str[100];
ulong_to_str(num_pe_pairs, num_pe_pairs_str);
status("[seq] Loaded %s read pairs (files: %s, %s)", num_pe_pairs_str,
futil_inpath_str(sf1->path), futil_inpath_str(sf2->path));
}
void seq_parse_interleaved_sf(seq_file_t *sf, uint8_t ascii_fq_offset,
read_t *r1, read_t *r2,
void (*read_func)(read_t *_r1, read_t *_r2,
uint8_t _qoffset1,
uint8_t _qoffset2,
void *_ptr),
void *reader_ptr)
{
status("[seq] Reading a (possibly) interleaved file (expect both S.E. & P.E. reads)");
// Guess offset if needed
uint8_t qoffset = ascii_fq_offset;
uint8_t qmin = ascii_fq_offset, qmax = 126;
int format;
if(ascii_fq_offset == 0 && (format = guess_fastq_format(sf)) != -1)
{
qmin = (uint8_t)FASTQ_MIN[format];
qmax = (uint8_t)FASTQ_MAX[format];
qoffset = (uint8_t)FASTQ_OFFSET[format];
}
read_t *r[2] = {r1,r2};
int ridx = 0, s;
uint8_t warn_flags = 0;
size_t num_se_reads = 0, num_pe_pairs = 0;
while((s = seq_read_primary(sf, r[ridx])) > 0)
{
warn_flags = check_new_read(r[ridx], qmin, qmax, sf->path, warn_flags);
if(ridx)
{
// ridx == 1
if(seq_read_names_cmp(r[0]->name.b, r[1]->name.b) == 0) {
// Either read may be the first in the pair if from SAM/BAM
int r0 = (r[1]->from_sam && seq_read_bam(r[1])->core.flag & BAM_FREAD1);
read_func(r[r0], r[!r0], qoffset, qoffset, reader_ptr);
num_pe_pairs++;
ridx = 0;
} else {
read_func(r[0], NULL, qoffset, 0, reader_ptr);
num_se_reads++;
SWAP(r[0], r[1]);
ridx = 1;
}
}
else ridx = 1;
}
// Process last read
if(ridx == 1) {
read_func(r[0], NULL, qoffset, 0, reader_ptr);
num_se_reads++;
}
if(s < 0) warn("Input error: %s\n", sf->path);
char num_se_reads_str[100], num_pe_pairs_str[100];
ulong_to_str(num_pe_pairs, num_pe_pairs_str);
ulong_to_str(num_se_reads, num_se_reads_str);
status("[seq] Loaded %s reads and %s reads pairs (file: %s)",
num_se_reads_str, num_pe_pairs_str, futil_inpath_str(sf->path));
}
int ctx_vcfcov(int argc, char **argv)
{
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL, *out_type = NULL;
uint32_t max_allele_len = 0, max_gt_vars = 0;
char *ref_path = NULL;
bool low_mem = false;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
size_t i;
// 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 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'r': cmd_check(!ref_path, cmd); ref_path = optarg; break;
case 'L': cmd_check(!max_allele_len,cmd); max_allele_len = cmd_uint32(cmd,optarg); break;
case 'N': cmd_check(!max_gt_vars,cmd); max_gt_vars = cmd_uint32(cmd,optarg); break;
case 'M': cmd_check(!low_mem, cmd); low_mem = true; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" "SUBCMD" -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(ref_path == NULL) cmd_print_usage("Require a reference (-r,--ref <ref.fa>)");
if(optind+2 > argc) cmd_print_usage("Require VCF and graph files");
if(!max_allele_len) max_allele_len = DEFAULT_MAX_ALLELE_LEN;
if(!max_gt_vars) max_gt_vars = DEFAULT_MAX_GT_VARS;
status("[vcfcov] max allele length: %u; max number of variants: %u",
max_allele_len, max_gt_vars);
// open ref
// index fasta with: samtools faidx ref.fa
faidx_t *fai = fai_load(ref_path);
if(fai == NULL) die("Cannot load ref index: %s / %s.fai", ref_path, ref_path);
// Open input VCF file
const char *vcf_path = argv[optind++];
htsFile *vcffh = hts_open(vcf_path, "r");
if(vcffh == NULL) die("Cannot open VCF file: %s", vcf_path);
bcf_hdr_t *vcfhdr = bcf_hdr_read(vcffh);
if(vcfhdr == NULL) die("Cannot read VCF header: %s", vcf_path);
// Test we can close and reopen files
if(low_mem) {
if((vcffh = hts_open(vcf_path, "r")) == NULL)
die("Cannot re-open VCF file: %s", vcf_path);
if((vcfhdr = bcf_hdr_read(vcffh)) == NULL)
die("Cannot re-read VCF header: %s", vcf_path);
}
//
// 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;
ncols = graph_files_open(graph_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
// Check graph + paths 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;
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,
low_mem ? -1 : (int64_t)ctx_max_kmers,
ctx_sum_kmers,
true, &graph_mem);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Open output file
//
// v=>vcf, z=>compressed vcf, b=>bcf, bu=>uncompressed bcf
int mode = vcf_misc_get_outtype(out_type, out_path);
futil_create_output(out_path);
htsFile *outfh = hts_open(out_path, modes_htslib[mode]);
status("[vcfcov] Output format: %s", hsmodes_htslib[mode]);
// Allocate memory
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, 1, kmers_in_hash,
DBG_ALLOC_COVGS);
//
// Set up tag names
//
// *R => ref, *A => alt
sprintf(kcov_ref_tag, "K%zuR", db_graph.kmer_size); // mean coverage
sprintf(kcov_alt_tag, "K%zuA", db_graph.kmer_size);
// #SAMPLE=<ID=...,K29KCOV=...,K29NK=...,K29RLK>
// - K29_kcov is empirical kmer coverage
// - K29_nkmers is the number of kmers in the sample
// - mean_read_length is the mean read length in bases
char sample_kcov_tag[20], sample_nk_tag[20], sample_rlk_tag[20];
sprintf(sample_kcov_tag, "K%zu_kcov", db_graph.kmer_size); // mean coverage
sprintf(sample_nk_tag, "K%zu_nkmers", db_graph.kmer_size);
sprintf(sample_rlk_tag, "mean_read_length");
//
// Load kmers if we are using --low-mem
//
VcfCovStats st;
memset(&st, 0, sizeof(st));
VcfCovPrefs prefs = {.kcov_ref_tag = kcov_ref_tag,
.kcov_alt_tag = kcov_alt_tag,
.max_allele_len = max_allele_len,
.max_gt_vars = max_gt_vars,
.load_kmers_only = false};
if(low_mem)
{
status("[vcfcov] Loading kmers from VCF+ref");
prefs.load_kmers_only = true;
vcfcov_file(vcffh, vcfhdr, NULL, NULL, vcf_path, fai,
NULL, &prefs, &st, &db_graph);
// Close files
hts_close(vcffh);
bcf_hdr_destroy(vcfhdr);
// Re-open files
if((vcffh = hts_open(vcf_path, "r")) == NULL)
die("Cannot re-open VCF file: %s", vcf_path);
if((vcfhdr = bcf_hdr_read(vcffh)) == NULL)
die("Cannot re-read VCF header: %s", vcf_path);
prefs.load_kmers_only = false;
}
//
// Load graphs
//
GraphLoadingStats gstats;
memset(&gstats, 0, sizeof(gstats));
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
gprefs.must_exist_in_graph = low_mem;
for(i = 0; i < num_gfiles; i++) {
graph_load(&gfiles[i], gprefs, &gstats);
graph_file_close(&gfiles[i]);
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
//
// Set up VCF header / graph matchup
//
size_t *samplehdrids = ctx_malloc(db_graph.num_of_cols * sizeof(size_t));
// Add samples to vcf header
bcf_hdr_t *outhdr = bcf_hdr_dup(vcfhdr);
bcf_hrec_t *hrec;
int sid;
char hdrstr[200];
for(i = 0; i < db_graph.num_of_cols; i++) {
char *sname = db_graph.ginfo[i].sample_name.b;
if((sid = bcf_hdr_id2int(outhdr, BCF_DT_SAMPLE, sname)) < 0) {
bcf_hdr_add_sample(outhdr, sname);
sid = bcf_hdr_id2int(outhdr, BCF_DT_SAMPLE, sname);
}
samplehdrids[i] = sid;
// Add SAMPLE field
hrec = bcf_hdr_get_hrec(outhdr, BCF_HL_STR, "ID", sname, "SAMPLE");
if(hrec == NULL) {
sprintf(hdrstr, "##SAMPLE=<ID=%s,%s=%"PRIu64",%s=%"PRIu64",%s=%zu>", sname,
sample_kcov_tag,
gstats.nkmers[i] ? gstats.sumcov[i] / gstats.nkmers[i] : 0,
sample_nk_tag, gstats.nkmers[i],
sample_rlk_tag, (size_t)db_graph.ginfo[i].mean_read_length);
bcf_hdr_append(outhdr, hdrstr);
}
else {
// mean kcovg
sprintf(hdrstr, "%"PRIu64, gstats.sumcov[i] / gstats.nkmers[i]);
vcf_misc_add_update_hrec(hrec, sample_kcov_tag, hdrstr);
// num kmers
sprintf(hdrstr, "%"PRIu64, gstats.nkmers[i]);
vcf_misc_add_update_hrec(hrec, sample_nk_tag, hdrstr);
// mean read length in kmers
sprintf(hdrstr, "%zu", (size_t)db_graph.ginfo[i].mean_read_length);
vcf_misc_add_update_hrec(hrec, sample_rlk_tag, hdrstr);
}
status("[vcfcov] Colour %zu: %s [VCF column %zu]", i, sname, samplehdrids[i]);
}
// Add genotype format fields
// One field per alternative allele
sprintf(hdrstr, "##FORMAT=<ID=%s,Number=A,Type=Integer,"
"Description=\"Coverage on ref (k=%zu): sum(kmer_covs) / exp_num_kmers\">\n",
kcov_ref_tag, db_graph.kmer_size);
bcf_hdr_append(outhdr, hdrstr);
sprintf(hdrstr, "##FORMAT=<ID=%s,Number=A,Type=Integer,"
"Description=\"Coverage on alt (k=%zu): sum(kmer_covs) / exp_num_kmers\">\n",
kcov_alt_tag, db_graph.kmer_size);
bcf_hdr_append(outhdr, hdrstr);
bcf_hdr_set_version(outhdr, "VCFv4.2");
// Add command string to header
vcf_misc_hdr_add_cmd(outhdr, cmd_get_cmdline(), cmd_get_cwd());
if(bcf_hdr_write(outfh, outhdr) != 0)
die("Cannot write header to: %s", futil_outpath_str(out_path));
status("[vcfcov] Reading %s and adding coverage", vcf_path);
// Reset stats and get coverage
memset(&st, 0, sizeof(st));
vcfcov_file(vcffh, vcfhdr, outfh, outhdr, vcf_path, fai,
samplehdrids, &prefs, &st, &db_graph);
// Print statistics
char ns0[50], ns1[50];
status("[vcfcov] Read %s VCF lines", ulong_to_str(st.nvcf_lines, ns0));
status("[vcfcov] Read %s ALTs", ulong_to_str(st.nalts_read, ns0));
status("[vcfcov] Used %s kmers", ulong_to_str(st.ngt_kmers, ns0));
status("[vcfcov] ALTs used: %s / %s (%.2f%%)",
ulong_to_str(st.nalts_loaded, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_loaded) / st.nalts_read : 0.0);
status("[vcfcov] ALTs too long (>%ubp): %s / %s (%.2f%%)", max_allele_len,
ulong_to_str(st.nalts_too_long, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_too_long) / st.nalts_read : 0.0);
status("[vcfcov] ALTs too dense (>%u within %zubp): %s / %s (%.2f%%)",
max_gt_vars, db_graph.kmer_size,
ulong_to_str(st.nalts_no_covg, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_no_covg) / st.nalts_read : 0.0);
status("[vcfcov] ALTs printed with coverage: %s / %s (%.2f%%)",
ulong_to_str(st.nalts_with_covg, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_with_covg) / st.nalts_read : 0.0);
status("[vcfcov] Saved to: %s\n", out_path);
ctx_free(samplehdrids);
graph_loading_stats_destroy(&gstats);
bcf_hdr_destroy(vcfhdr);
bcf_hdr_destroy(outhdr);
hts_close(vcffh);
hts_close(outfh);
fai_destroy(fai);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
