/* Function: esl_msafile_a2m_GuessAlphabet()
* Synopsis: Guess the alphabet of an open A2M MSA file.
*
* Purpose: Guess the alpbabet of the sequences in open
* A2M format MSA file <afp>.
*
* On a normal return, <*ret_type> is set to <eslDNA>,
* <eslRNA>, or <eslAMINO>, and <afp> is reset to its
* original position.
*
* Args: afp - open A2M format MSA file
* ret_type - RETURN: <eslDNA>, <eslRNA>, or <eslAMINO>
*
* Returns: <eslOK> on success.
* <eslENOALPHABET> if alphabet type can't be determined.
* In either case, <afp> is rewound to the position it
* started at.
*
* Throws: <eslEMEM> on allocation error.
* <eslESYS> on failures of fread() or other system calls
*/
int
esl_msafile_a2m_GuessAlphabet(ESL_MSAFILE *afp, int *ret_type)
{
int alphatype = eslUNKNOWN;
esl_pos_t anchor = -1;
int threshold[3] = { 500, 5000, 50000 }; /* we check after 500, 5000, 50000 residues; else we go to EOF */
int nsteps = 3;
int step = 0;
int nres = 0;
int x;
int64_t ct[26];
char *p;
esl_pos_t n, pos;
int status;
for (x = 0; x < 26; x++) ct[x] = 0;
anchor = esl_buffer_GetOffset(afp->bf);
if ((status = esl_buffer_SetAnchor(afp->bf, anchor)) != eslOK) { status = eslEINCONCEIVABLE; goto ERROR; } /* [eslINVAL] can't happen here */
while ( (status = esl_buffer_GetLine(afp->bf, &p, &n)) == eslOK)
{
while (n && isspace(*p)) { p++; n--; }
if (!n || *p == '>') continue;
for (pos = 0; pos < n; pos++)
if (isalpha(p[pos])) {
x = toupper(p[pos]) - 'A';
ct[x]++;
nres++;
}
/* try to stop early, checking after 500, 5000, and 50000 residues: */
if (step < nsteps && nres > threshold[step]) {
if ((status = esl_abc_GuessAlphabet(ct, &alphatype)) == eslOK) goto DONE; /* (eslENOALPHABET) */
step++;
}
}
if (status != eslEOF) goto ERROR; /* [eslEMEM,eslESYS,eslEINCONCEIVABLE] */
status = esl_abc_GuessAlphabet(ct, &alphatype); /* (eslENOALPHABET) */
/* deliberate flowthrough...*/
DONE:
esl_buffer_SetOffset(afp->bf, anchor); /* Rewind to where we were. */
esl_buffer_RaiseAnchor(afp->bf, anchor);
*ret_type = alphatype;
return status;
ERROR:
if (anchor != -1) {
esl_buffer_SetOffset(afp->bf, anchor);
esl_buffer_RaiseAnchor(afp->bf, anchor);
}
*ret_type = eslUNKNOWN;
return status;
}
/* Function: esl_msafile_psiblast_GuessAlphabet()
* Synopsis: Guess the alphabet of an open PSI-BLAST MSA file.
*
* Purpose: Guess the alpbabet of the sequences in open
* PSI-BLAST format MSA file <afp>.
*
* On a normal return, <*ret_type> is set to <eslDNA>,
* <eslRNA>, or <eslAMINO>, and <afp> is reset to its
* original position.
*
* Args: afp - open PSI-BLAST format MSA file
* ret_type - RETURN: <eslDNA>, <eslRNA>, or <eslAMINO>
*
* Returns: <eslOK> on success.
* <eslENOALPHABET> if alphabet type can't be determined.
* In either case, <afp> is rewound to the position it
* started at.
*/
int
esl_msafile_psiblast_GuessAlphabet(ESLX_MSAFILE *afp, int *ret_type)
{
int alphatype = eslUNKNOWN;
esl_pos_t anchor = -1;
int threshold[3] = { 500, 5000, 50000 }; /* we check after 500, 5000, 50000 residues; else we go to EOF */
int nsteps = 3;
int step = 0;
int nres = 0;
int x;
int64_t ct[26];
char *p, *tok;
esl_pos_t n, toklen, pos;
int status;
for (x = 0; x < 26; x++) ct[x] = 0;
anchor = esl_buffer_GetOffset(afp->bf);
if ((status = esl_buffer_SetAnchor(afp->bf, anchor)) != eslOK) { status = eslEINCONCEIVABLE; goto ERROR; } /* [eslINVAL] can't happen here */
while ( (status = esl_buffer_GetLine(afp->bf, &p, &n)) == eslOK)
{
if ((status = esl_memtok(&p, &n, " \t", &tok, &toklen)) != eslOK) continue; /* blank lines */
/* p now points to the rest of the sequence line, after a name */
/* count characters into ct[] array */
for (pos = 0; pos < n; pos++)
if (isalpha(p[pos])) {
x = toupper(p[pos]) - 'A';
ct[x]++;
nres++;
}
/* try to stop early, checking after 500, 5000, and 50000 residues: */
if (step < nsteps && nres > threshold[step]) {
if ((status = esl_abc_GuessAlphabet(ct, &alphatype)) == eslOK) goto DONE; /* (eslENOALPHABET) */
step++;
}
}
if (status != eslEOF) goto ERROR; /* [eslEMEM,eslESYS,eslEINCONCEIVABLE] */
status = esl_abc_GuessAlphabet(ct, &alphatype); /* (eslENOALPHABET) */
DONE:
esl_buffer_SetOffset(afp->bf, anchor); /* Rewind to where we were. */
esl_buffer_RaiseAnchor(afp->bf, anchor);
*ret_type = alphatype;
return status;
ERROR:
if (anchor != -1) {
esl_buffer_SetOffset(afp->bf, anchor);
esl_buffer_RaiseAnchor(afp->bf, anchor);
}
*ret_type = eslUNKNOWN;
return status;
}
