// Using file so can call fseek and don't need to load whole graph
static size_t inferedges_on_mmap(const dBGraph *db_graph, bool add_all_edges,
GraphFileReader *file)
{
ctx_assert(db_graph->num_of_cols == file->hdr.num_of_cols);
ctx_assert(file_filter_is_direct(&file->fltr));
ctx_assert2(!isatty(fileno(file->fh)), "Use inferedges_on_stream() instead");
ctx_assert(file->num_of_kmers >= 0);
ctx_assert(file->file_size >= 0);
status("[inferedges] Processing mmap file: %s [hdr: %zu bytes file: %zu bytes]",
file_filter_path(&file->fltr),
(size_t)file->hdr_size, (size_t)file->file_size);
if(fseek(file->fh, 0, SEEK_SET) != 0)
die("fseek failed: %s", strerror(errno));
// Open memory mapped file
void *mmap_ptr = mmap(NULL, file->file_size, PROT_WRITE, MAP_SHARED,
fileno(file->fh), 0);
if(mmap_ptr == MAP_FAILED)
die("Cannot memory map file: %s [%s]", file->fltr.path.b, strerror(errno));
const size_t ncols = file->hdr.num_of_cols;
BinaryKmer bkmer;
Edges edges[ncols];
Covg covgs[ncols];
bool updated;
size_t i, num_kmers = file->num_of_kmers, num_kmers_edited = 0;
size_t filekmersize = sizeof(BinaryKmer) + (sizeof(Edges)+sizeof(Covg)) * ncols;
char *ptr = (char*)mmap_ptr + file->hdr_size;
for(i = 0; i < num_kmers; i++, ptr += filekmersize)
{
char *fh_covgs = ptr + sizeof(BinaryKmer);
char *fh_edges = fh_covgs + sizeof(Covg)*ncols;
memcpy(bkmer.b, ptr, sizeof(BinaryKmer));
memcpy(covgs, fh_covgs, ncols * sizeof(Covg));
memcpy(edges, fh_edges, ncols * sizeof(Edges));
updated = (add_all_edges ? infer_all_edges(bkmer, edges, covgs, db_graph)
: infer_pop_edges(bkmer, edges, covgs, db_graph));
if(updated) {
memcpy(fh_covgs, covgs, ncols * sizeof(Covg));
memcpy(fh_edges, edges, ncols * sizeof(Edges));
num_kmers_edited++;
}
}
if(munmap(mmap_ptr, file->file_size) == -1)
die("Cannot release mmap file: %s [%s]", file->fltr.path.b, strerror(errno));
return num_kmers_edited;
}
// Print file filter description
void file_filter_status(const FileFilter *fltr)
{
size_t i;
pthread_mutex_lock(&ctx_biglock);
timestamp();
message("[FileFilter] Loading file %s [%u colour%s]", file_filter_path(fltr),
fltr->filencols, util_plural_str(fltr->filencols));
if(!file_filter_is_direct(fltr))
{
message(" with filter: %u->%u", file_filter_fromcol(fltr, 0),
file_filter_intocol(fltr, 0));
for(i = 1; i < file_filter_num(fltr); i++)
message(",%u->%u", file_filter_fromcol(fltr,i), file_filter_intocol(fltr,i));
}
message("\n");
pthread_mutex_unlock(&ctx_biglock);
}
// Using file so can call fseek and don't need to load whole graph
static size_t inferedges_on_file(const dBGraph *db_graph, bool add_all_edges,
GraphFileReader *file, FILE *fout)
{
ctx_assert(db_graph->num_of_cols == file->hdr.num_of_cols);
ctx_assert(file_filter_is_direct(&file->fltr));
ctx_assert2(!isatty(fileno(file->fh)), "Use inferedges_on_stream() instead");
ctx_assert(fout != NULL);
ctx_assert(fileno(file->fh) != fileno(fout));
status("[inferedges] Processing file: %s", file_filter_path(&file->fltr));
// Print header
graph_write_header(fout, &file->hdr);
// Read the input file again
if(fseek(file->fh, file->hdr_size, SEEK_SET) != 0)
die("fseek failed: %s", strerror(errno));
const size_t ncols = file->hdr.num_of_cols;
BinaryKmer bkmer;
Edges edges[ncols];
Covg covgs[ncols];
size_t num_kmers_edited = 0;
bool updated;
while(graph_file_read_reset(file, ncols, &bkmer, covgs, edges))
{
updated = (add_all_edges ? infer_all_edges(bkmer, edges, covgs, db_graph)
: infer_pop_edges(bkmer, edges, covgs, db_graph));
graph_write_kmer(fout, file->hdr.num_of_bitfields, file->hdr.num_of_cols,
bkmer, covgs, edges);
num_kmers_edited += updated;
}
return num_kmers_edited;
}
int ctx_sort(int argc, char **argv)
{
const char *out_path = NULL;
struct MemArgs memargs = MEM_ARGS_INIT;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'o': cmd_check(!out_path, cmd); out_path = optarg; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" sort -h` for help. Bad option: %s", argv[optind-1]);
default: die("Bad option: [%c]: %s", c, cmd);
}
}
if(optind+1 != argc)
cmd_print_usage("Require exactly one input graph file (.ctx)");
const char *ctx_path = argv[optind];
//
// Open Graph file
//
GraphFileReader gfile;
memset(&gfile, 0, sizeof(GraphFileReader));
graph_file_open2(&gfile, ctx_path, out_path ? "r" : "r+", true, 0);
if(!file_filter_is_direct(&gfile.fltr))
die("Cannot open graph file with a filter ('in.ctx:blah' syntax)");
size_t num_kmers, memory;
// Reading from a stream
if(gfile.num_of_kmers < 0) {
if(!memargs.num_kmers_set)
die("If reading from a stream, must give -n <num_kmers>");
num_kmers = memargs.num_kmers;
}
else num_kmers = gfile.num_of_kmers;
// Open output path (if given)
FILE *fout = out_path ? futil_fopen_create(out_path, "w") : NULL;
size_t i;
size_t ncols = gfile.hdr.num_of_cols;
size_t kmer_mem = sizeof(BinaryKmer) + (sizeof(Edges)+sizeof(Covg))*ncols;
memory = (sizeof(char*) + kmer_mem) * num_kmers;
char mem_str[50];
bytes_to_str(memory, 1, mem_str);
if(memory > memargs.mem_to_use)
die("Require at least %s memory", mem_str);
status("[memory] Total: %s", mem_str);
char *mem = ctx_malloc(kmer_mem * num_kmers);
char **kmers = ctx_malloc(num_kmers*sizeof(char*));
// Read in whole file
// if(graph_file_fseek(gfile, gfile.hdr_size, SEEK_SET) != 0) die("fseek failed");
size_t nkread = gfr_fread_bytes(&gfile, mem, num_kmers*kmer_mem);
if(nkread != num_kmers*kmer_mem)
die("Could only read %zu bytes [<%zu]", nkread, num_kmers*kmer_mem);
// check we are at the end of the file
char tmpc;
if(gfr_fread_bytes(&gfile, &tmpc, 1) != 0) {
die("More kmers in file than believed (kmers: %zu ncols: %zu).",
num_kmers, ncols);
}
status("Read %zu kmers with %zu colour%s", num_kmers,
ncols, util_plural_str(ncols));
for(i = 0; i < num_kmers; i++)
kmers[i] = mem + kmer_mem*i;
sort_block(kmers, num_kmers);
// Print
if(out_path != NULL) {
// saving to a different destination - write header
graph_write_header(fout, &gfile.hdr);
}
else {
// Directly manipulating gfile.fh here, using it to write later
// Not doing any more reading
if(fseek(gfile.fh, gfile.hdr_size, SEEK_SET) != 0) die("fseek failed");
fout = gfile.fh;
}
for(i = 0; i < num_kmers; i++)
if(fwrite(kmers[i], 1, kmer_mem, fout) != kmer_mem)
die("Cannot write to file");
if(out_path) fclose(fout);
graph_file_close(&gfile);
ctx_free(kmers);
ctx_free(mem);
return EXIT_SUCCESS;
}
int ctx_index(int argc, char **argv)
{
const char *out_path = NULL;
size_t block_size = 0, block_kmers = 0;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'o': cmd_check(!out_path, cmd); out_path = optarg; break;
case 'b':
cmd_check(!block_kmers, cmd);
block_kmers = cmd_size_nonzero(cmd, optarg);
break;
case 's':
cmd_check(!block_size, cmd);
block_size = cmd_size_nonzero(cmd, optarg);
break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" index -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(optind+1 != argc)
cmd_print_usage("Require exactly one input graph file (.ctx)");
if(block_size && block_kmers)
cmd_print_usage("Cannot use --block-kmers and --block-size together");
const char *ctx_path = argv[optind];
//
// Open Graph file
//
GraphFileReader gfile;
memset(&gfile, 0, sizeof(GraphFileReader));
graph_file_open2(&gfile, ctx_path, "r+", true, 0);
if(!file_filter_is_direct(&gfile.fltr))
die("Cannot open graph file with a filter ('in.ctx:blah' syntax)");
// Open output file
FILE *fout = out_path ? futil_fopen_create(out_path, "w") : stdout;
// Start
size_t filencols = gfile.hdr.num_of_cols;
size_t kmer_size = gfile.hdr.kmer_size;
const char *path = file_filter_path(&gfile.fltr);
size_t ncols = file_filter_into_ncols(&gfile.fltr);
size_t kmer_mem = sizeof(BinaryKmer) + (sizeof(Edges)+sizeof(Covg))*filencols;
if(block_size) {
block_kmers = block_size / kmer_mem;
} else if(!block_size && !block_kmers) {
block_size = 4 * ONE_MEGABYTE;
block_kmers = block_size / kmer_mem;
}
// Update block-size
block_size = block_kmers * kmer_mem;
status("[index] block bytes: %zu kmers: %zu; kmer bytes: %zu, hdr: %zu",
block_size, block_kmers, kmer_mem, (size_t)gfile.hdr_size);
if(block_kmers == 0) die("Cannot set block_kmers to zero");
// Print header
fputs("#block_start\tnext_block\tfirst_kmer\tkmer_idx\tnext_kmer_idx\n", fout);
BinaryKmer bkmer = BINARY_KMER_ZERO_MACRO;
BinaryKmer prev_bkmer = BINARY_KMER_ZERO_MACRO;
Covg *covgs = ctx_malloc(ncols * sizeof(Covg));
Edges *edges = ctx_malloc(ncols * sizeof(Edges));
char bkmerstr[MAX_KMER_SIZE+1];
size_t rem_block = block_size - kmer_mem; // block after first kmer
char *tmp_mem = ctx_malloc(rem_block);
// Read in file, print index
size_t nblocks = 0;
size_t bl_bytes = 0, bl_kmers = 0;
size_t bl_byte_offset = gfile.hdr_size, bl_kmer_offset = 0;
while(1)
{
if(!graph_file_read(&gfile, &bkmer, covgs, edges)) {
status("Read kmer failed"); break; }
binary_kmer_to_str(bkmer, kmer_size, bkmerstr);
if(nblocks > 0 && !binary_kmer_less_than(prev_bkmer,bkmer))
die("File is not sorted: %s [%s]", bkmerstr, path);
// We've already read one kmer entry, read rest of block
bl_bytes = kmer_mem + gfr_fread_bytes(&gfile, tmp_mem, rem_block);
bl_kmers = 1 + bl_bytes / kmer_mem;
fprintf(fout, "%zu\t%zu\t%s\t%zu\t%zu\n",
bl_byte_offset, bl_byte_offset+bl_bytes, bkmerstr,
bl_kmer_offset, bl_kmer_offset+bl_kmers);
bl_byte_offset += bl_bytes;
bl_kmer_offset += bl_kmers;
nblocks++;
if(bl_kmers < block_kmers) {
status("last block %zu < %zu; %zu vs %zu",
bl_kmers, block_kmers, bl_bytes, block_size);
break;
}
prev_bkmer = bkmer;
}
ctx_free(covgs);
ctx_free(edges);
ctx_free(tmp_mem);
// done
char num_kmers_str[50], num_blocks_str[50];
char block_mem_str[50], block_kmers_str[50];
ulong_to_str(bl_kmer_offset, num_kmers_str);
ulong_to_str(nblocks, num_blocks_str);
bytes_to_str(block_size, 1, block_mem_str);
ulong_to_str(block_kmers, block_kmers_str);
status("Read %s kmers in %s block%s (block size %s / %s kmers)",
num_kmers_str, num_blocks_str, util_plural_str(nblocks),
block_mem_str, block_kmers_str);
if(fout != stdout) status("Saved to %s", out_path);
graph_file_close(&gfile);
fclose(fout);
return EXIT_SUCCESS;
}
int ctx_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 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;
}
