static int construct_bioseq_files(GtBioseq *bs, GtStr *bioseq_indexname,
GtError *err)
{
GtStr *sequence_filename;
GtEncseqEncoder *ee;
GtStrArray *indexfn;
int had_err = 0;
gt_error_check(err);
/* register the signal handler to remove incomplete files upon termination */
if (!bs->use_stdin) {
gt_bioseq_index_filename = gt_str_get(bs->sequence_file);
gt_sig_register_all(remove_bioseq_files);
}
/* if stdin is used as input, we need to create a tempfile containing the
sequence as GtEncseq cannot be built from stdin directly */
if (bs->use_stdin) {
GtStr *tmpfilename;
FILE *tmpfile = NULL;
int i;
char buf[BUFSIZ];
tmpfilename = gt_str_new();
tmpfile = gt_xtmpfp(tmpfilename);
gt_assert(tmpfile);
i = 1;
while (i > 0) {
i = fread(buf, 1, BUFSIZ, stdin);
if (i > 0) fwrite(buf, 1, i, tmpfile);
}
gt_fa_xfclose(tmpfile);
sequence_filename = tmpfilename;
} else {
sequence_filename = gt_str_ref(bs->sequence_file);
}
gt_assert(gt_str_length(sequence_filename) > 0);
ee = gt_encseq_encoder_new();
gt_encseq_encoder_enable_description_support(ee);
gt_encseq_encoder_enable_md5_support(ee);
gt_encseq_encoder_enable_multiseq_support(ee);
gt_encseq_encoder_enable_lossless_support(ee);
indexfn = gt_str_array_new();
gt_str_array_add(indexfn, sequence_filename);
gt_str_delete(sequence_filename);
had_err = gt_encseq_encoder_encode(ee, indexfn,
gt_str_get(bioseq_indexname), err);
/* unregister the signal handler */
if (!bs->use_stdin)
gt_sig_unregister_all();
gt_str_array_delete(indexfn);
gt_encseq_encoder_delete(ee);
return had_err;
}
GtHcrEncoder *gt_hcr_encoder_new(GtStrArray *files, GtAlphabet *alpha,
bool descs, GtQualRange qrange, GtTimer *timer,
GtError *err)
{
GtBaseQualDistr *bqd;
GtHcrEncoder *hcr_enc;
GtSeqIterator *seqit;
GtStrArray *file;
int had_err = 0,
status;
GtUword len1,
len2,
i,
num_of_reads = 0;
const GtUchar *seq,
*qual;
char *desc;
gt_error_check(err);
gt_assert(alpha && files);
if (timer != NULL)
gt_timer_show_progress(timer, "get <base,qual> distr", stdout);
if (qrange.start != GT_UNDEF_UINT)
if (qrange.start == qrange.end) {
gt_error_set(err, "qrange.start must unequal qrange.end");
return NULL;
}
hcr_enc = gt_malloc(sizeof (GtHcrEncoder));
hcr_enc->files = files;
hcr_enc->num_of_files = gt_str_array_size(files);
hcr_enc->num_of_reads = 0;
hcr_enc->page_sampling = false;
hcr_enc->regular_sampling = false;
hcr_enc->sampling_rate = 0;
hcr_enc->pagesize = gt_pagesize();
if (descs) {
hcr_enc->encdesc_encoder = gt_encdesc_encoder_new();
if (timer != NULL)
gt_encdesc_encoder_set_timer(hcr_enc->encdesc_encoder, timer);
}
else
hcr_enc->encdesc_encoder = NULL;
hcr_enc->seq_encoder = gt_malloc(sizeof (GtHcrSeqEncoder));
hcr_enc->seq_encoder->alpha = alpha;
hcr_enc->seq_encoder->sampling = NULL;
hcr_enc->seq_encoder->fileinfos = gt_calloc((size_t) hcr_enc->num_of_files,
sizeof (*(hcr_enc->seq_encoder->fileinfos)));
hcr_enc->seq_encoder->qrange = qrange;
bqd = hcr_base_qual_distr_new(alpha, qrange);
/* check if reads in the same file are of same length and get
<base, quality> pair distribution */
for (i = 0; i < hcr_enc->num_of_files; i++) {
file = gt_str_array_new();
gt_str_array_add(file, gt_str_array_get_str(files, i));
seqit = gt_seq_iterator_fastq_new(file, err);
if (!seqit) {
gt_error_set(err, "cannot initialize GtSeqIteratorFastQ object");
had_err = -1;
}
if (!had_err) {
gt_seq_iterator_set_symbolmap(seqit, gt_alphabet_symbolmap(alpha));
gt_seq_iterator_set_quality_buffer(seqit, &qual);
status = gt_seq_iterator_next(seqit, &seq, &len1, &desc, err);
if (status == 1) {
num_of_reads = 1UL;
while (!had_err) {
status = gt_seq_iterator_next(seqit, &seq, &len2, &desc, err);
if (status == -1)
had_err = -1;
if (status != 1)
break;
if (len2 != len1) {
gt_error_set(err, "reads have to be of equal length");
had_err = -1;
break;
}
if (hcr_base_qual_distr_add(bqd, qual, seq, len1) != 0)
had_err = -1;
len1 = len2;
num_of_reads++;
}
}
else if (status == -1)
had_err = -1;
if (!had_err) {
if (i == 0)
hcr_enc->seq_encoder->fileinfos[i].readnum = num_of_reads;
else
hcr_enc->seq_encoder->fileinfos[i].readnum =
hcr_enc->seq_encoder->fileinfos[i - 1].readnum + num_of_reads;
hcr_enc->seq_encoder->fileinfos[i].readlength = len1;
}
}
hcr_enc->num_of_reads += num_of_reads;
gt_str_array_delete(file);
gt_seq_iterator_delete(seqit);
}
if (!had_err)
hcr_base_qual_distr_trim(bqd);
if (!had_err) {
if (timer != NULL)
gt_timer_show_progress(timer, "build huffman tree for sequences and"
" qualities", stdout);
hcr_enc->seq_encoder->huffman =
gt_huffman_new(bqd,
hcr_base_qual_distr_func,
(GtUword) bqd->ncols * bqd->nrows);
}
if (!had_err) {
hcr_enc->seq_encoder->qual_offset = bqd->qual_offset;
hcr_base_qual_distr_delete(bqd);
return hcr_enc;
}
return NULL;
}
static int update_seq_col_if_necessary(GtRegionMapping *rm, GtStr *seqid,
GtError *err)
{
int had_err = 0;
gt_error_check(err);
gt_assert(rm && seqid);
/* for mappings, we need to load the changed sequence, if needed... */
if (rm->mapping) {
if (!rm->sequence_file || (gt_str_cmp(rm->sequence_name, seqid))) {
gt_str_delete(rm->sequence_file);
/* ignore MD5 hashes when using region mappings */
if (gt_md5_seqid_has_prefix(gt_str_get(seqid))) {
rm->sequence_file = region_mapping_map(rm,
gt_str_get(seqid)
+GT_MD5_SEQID_TOTAL_LEN,
err);
} else
rm->sequence_file = region_mapping_map(rm, gt_str_get(seqid), err);
if (!rm->sequence_file)
had_err = -1;
else {
/* load new seqcol */
if (!rm->sequence_filenames)
rm->sequence_filenames = gt_str_array_new();
else
gt_str_array_reset(rm->sequence_filenames);
gt_str_array_add(rm->sequence_filenames, rm->sequence_file);
if (!rm->sequence_name)
rm->sequence_name = gt_str_new();
else
gt_str_reset(rm->sequence_name);
gt_str_append_str(rm->sequence_name, seqid);
gt_seq_col_delete(rm->seq_col);
rm->seq_col = gt_bioseq_col_new(rm->sequence_filenames, err);
if (!rm->seq_col)
had_err = -1;
}
}
} else {
/* ...otherwise, just make sure the seqcol is loaded */
if (!rm->seq_col) {
if (rm->encseq) {
if (!(rm->seq_col = gt_encseq_col_new(rm->encseq, err)))
had_err = -1;
} else {
gt_assert(rm->sequence_filenames);
if (!(rm->seq_col = gt_bioseq_col_new(rm->sequence_filenames, err)))
had_err = -1;
}
}
if (!had_err && rm->usedesc) {
if (rm->seqid2seqnum_mapping)
gt_seqid2seqnum_mapping_delete(rm->seqid2seqnum_mapping);
rm->seqid2seqnum_mapping =
gt_seqid2seqnum_mapping_new_seqcol(rm->seq_col, err);
if (!rm->seqid2seqnum_mapping) {
had_err = -1;
}
}
}
return had_err;
}
