* samtools-0.1.12-9 (r891)

[samtools.git] / bcftools / bcfutils.c

diff --git a/bcftools/bcfutils.c b/bcftools/bcfutils.c
index dd1e89f16cf1c48ac8f90ceae35bae312d76ba54..fb70d81ae8c1c50d765176860c55719388f5b260 100644 (file)
--- a/bcftools/bcfutils.c
+++ b/bcftools/bcfutils.c
@@ -1,3 +1,5 @@
+#include <string.h>
+#include <math.h>
 #include "bcf.h"
 #include "kstring.h"
 #include "khash.h"
@@ -48,11 +50,10 @@ int bcf_str2id_add(void *_hash, const char *str)
        return kh_val(hash, k);
 }
 
-int bcf_shrink_alt(int n_smpl, bcf1_t *b, int n)
+int bcf_shrink_alt(bcf1_t *b, int n)
 {
        char *p;
-       int i, j, k;
-       int *z;
+       int i, j, k, *z, n_smpl = b->n_smpl;
        if (b->n_alleles <= n) return -1;
        if (n > 1) {
                for (p = b->alt, k = 1; *p; ++p)
@@ -79,14 +80,14 @@ int bcf_shrink_alt(int n_smpl, bcf1_t *b, int n)
                } // FIXME: to add GL
        }
        b->n_alleles = n;
-       bcf_sync(n_smpl, b);
+       bcf_sync(b);
        return 0;
 }
 
-int bcf_gl2pl(int n_smpl, bcf1_t *b)
+int bcf_gl2pl(bcf1_t *b)
 {
        char *p;
-       int i;
+       int i, n_smpl = b->n_smpl;
        bcf_ginfo_t *g;
        float *d0;
        uint8_t *d1;
@@ -108,3 +109,113 @@ int bcf_gl2pl(int n_smpl, bcf1_t *b)
        }
        return 0;
 }
+/* FIXME: this function will fail given AB:GTX:GT. BCFtools never
+ * produces such FMT, but others may do. */
+int bcf_fix_gt(bcf1_t *b)
+{
+       char *s;
+       int i;
+       uint32_t tmp;
+       bcf_ginfo_t gt;
+       // check the presence of the GT FMT
+       if ((s = strstr(b->fmt, ":GT")) == 0) return 0; // no GT or GT is already the first
+       if (s[3] != '\0' && s[3] != ':') return 0; // :GTX in fact
+       tmp = bcf_str2int("GT", 2);
+       for (i = 0; i < b->n_gi; ++i)
+               if (b->gi[i].fmt == tmp) break;
+       if (i == b->n_gi) return 0; // no GT in b->gi; probably a bug...
+       gt = b->gi[i];
+       // move GT to the first
+       for (; i > 0; --i) b->gi[i] = b->gi[i-1];
+       b->gi[0] = gt;
+       memmove(b->fmt + 3, b->fmt, s + 1 - b->fmt);
+       b->fmt[0] = 'G'; b->fmt[1] = 'T'; b->fmt[2] = ':';
+       return 0;
+}
+
+static void *locate_field(const bcf1_t *b, const char *fmt, int l)
+{
+       int i;
+       uint32_t tmp;
+       tmp = bcf_str2int(fmt, l);
+       for (i = 0; i < b->n_gi; ++i)
+               if (b->gi[i].fmt == tmp) break;
+       return i == b->n_gi? 0 : b->gi[i].data;
+}
+
+int bcf_anno_max(bcf1_t *b)
+{
+       int k, max_gq, max_sp, n_het;
+       kstring_t str;
+       uint8_t *gt, *gq, *sp;
+       max_gq = max_sp = n_het = 0;
+       gt = locate_field(b, "GT", 2);
+       if (gt == 0) return -1;
+       gq = locate_field(b, "GQ", 2);
+       sp = locate_field(b, "SP", 2);
+       if (sp)
+               for (k = 0; k < b->n_smpl; ++k)
+                       if (gt[k]&0x3f)
+                               max_sp = max_sp > (int)sp[k]? max_sp : sp[k];
+       if (gq)
+               for (k = 0; k < b->n_smpl; ++k)
+                       if (gt[k]&0x3f)
+                               max_gq = max_gq > (int)gq[k]? max_gq : gq[k];
+       for (k = 0; k < b->n_smpl; ++k) {
+               int a1, a2;
+               a1 = gt[k]&7; a2 = gt[k]>>3&7;
+               if ((!a1 && a2) || (!a2 && a1)) { // a het
+                       if (gq == 0) ++n_het;
+                       else if (gq[k] >= 20) ++n_het;
+               }
+       }
+       if (n_het) max_sp -= (int)(4.343 * log(n_het) + .499);
+       if (max_sp < 0) max_sp = 0;
+       memset(&str, 0, sizeof(kstring_t));
+       if (*b->info) kputc(';', &str);
+       ksprintf(&str, "MXSP=%d;MXGQ=%d", max_sp, max_gq);
+       bcf_append_info(b, str.s, str.l);
+       free(str.s);
+       return 0;
+}
+
+bcf_hdr_t *bcf_hdr_subsam(const bcf_hdr_t *h0, int n, char *const* samples, int *list)
+{
+       int i, ret, j;
+       khint_t k;
+       bcf_hdr_t *h;
+       khash_t(str2id) *hash;
+       kstring_t s;
+       s.l = s.m = 0; s.s = 0;
+       hash = kh_init(str2id);
+       for (i = 0; i < n; ++i)
+               k = kh_put(str2id, hash, samples[i], &ret);
+       for (i = j = 0; i < h0->n_smpl; ++i) {
+               if (kh_get(str2id, hash, h0->sns[i]) != kh_end(hash)) {
+                       list[j++] = i;
+                       kputs(h0->sns[i], &s); kputc('\0', &s);
+               }
+       }
+       if (j < n) fprintf(stderr, "<%s> %d samples in the list but not in BCF.", __func__, n - j);
+       kh_destroy(str2id, hash);
+       h = calloc(1, sizeof(bcf_hdr_t));
+       *h = *h0;
+       h->ns = 0; h->sns = 0;
+       h->name = malloc(h->l_nm); memcpy(h->name, h0->name, h->l_nm);
+       h->txt = calloc(1, h->l_txt + 1); memcpy(h->txt, h0->txt, h->l_txt);
+       h->l_smpl = s.l; h->sname = s.s;
+       bcf_hdr_sync(h);
+       return h;
+}
+
+int bcf_subsam(int n_smpl, int *list, bcf1_t *b) // list MUST BE sorted
+{
+       int i, j;
+       for (j = 0; j < b->n_gi; ++j) {
+               bcf_ginfo_t *gi = b->gi + j;
+               for (i = 0; i < n_smpl; ++i)
+                       memcpy((uint8_t*)gi->data + i * gi->len, (uint8_t*)gi->data + list[i] * gi->len, gi->len);
+       }
+       b->n_smpl = n_smpl;
+       return 0;
+}