/* Copyright 2002 by Jeffrey Chang. All rights reserved. * This code is part of the Biopython distribution and governed by its * license. Please see the LICENSE file that should have been included * as part of this package. * * cpairwise2module.c * Created 30 Sep 2001 * * Optimized C routines that complement pairwise2.py. */ #include "Python.h" #include "csupport.h" #define _PRECISION 1000 #define rint(x) (int)((x)*_PRECISION+0.5) /* Functions in this module. */ double calc_affine_penalty(int length, double open, double extend, int penalize_extend_when_opening) { double penalty; if(length <= 0) return 0.0; penalty = open + extend * length; if(!penalize_extend_when_opening) penalty -= extend; return penalty; } struct IndexList { int num_used; int num_allocated; int *indexes; /* Array of ints. Even ints are rows, odd ones are cols. */ }; static void IndexList_init(struct IndexList *il) { /* il->num_used = 0; il->num_allocated = 0; il->indexes = NULL; */ memset((void *)il, 0, sizeof(struct IndexList)); } static void IndexList_free(struct IndexList *il) { if(il->indexes) free(il->indexes); IndexList_init(il); } static void IndexList_clear(struct IndexList *il) { il->num_used = 0; } static int IndexList_contains(struct IndexList *il, const int row, const int col) { int i; int stop; stop = il->num_used*2; for(i=0; iindexes[i] == row) && (il->indexes[i+1] == col)) return 1; } return 0; } static int IndexList__verify_free_index(struct IndexList *il, int num_needed) { int *indexes; int num_to_allocate; int num_bytes; if(il->num_allocated >= num_needed) return 1; /* Nearly all the cases are for list of length 1 or 2. Empirically, the code seems to run fastest when I allocate the exact amount for these. */ if(num_needed <= 2) num_to_allocate = num_needed; else num_to_allocate = num_needed*2; num_bytes = num_to_allocate*sizeof(int)*2; if(!(indexes = realloc((void *)il->indexes, num_bytes))) { PyErr_SetString(PyExc_MemoryError, "Out of memory"); return 0; } il->indexes = indexes; il->num_allocated = num_to_allocate; return 1; } static void IndexList_append(struct IndexList *il, const int row, const int col) { int i; if(!IndexList__verify_free_index(il, il->num_used+1)) return; i=il->num_used*2; il->indexes[i] = row; il->indexes[i+1] = col; il->num_used += 1; } static void IndexList_extend(struct IndexList *il1, struct IndexList *il2) { int i1, i2; int stop; if(!IndexList__verify_free_index(il1, il1->num_used+il2->num_used)) return; stop=il2->num_used * 2; for(i1=il1->num_used*2, i2=0; i2indexes[i1] = il2->indexes[i2]; il1->indexes[i1+1] = il2->indexes[i2+1]; } il1->num_used += il2->num_used; } /* static void IndexList_copy(struct IndexList *il1, struct IndexList *il2) { IndexList_clear(il1); IndexList_extend(il1, il2); } */ /* static void IndexList_assign(struct IndexList *il1, struct IndexList *il2) { if(il1->indexes) free(il1->indexes); il1->indexes = il2->indexes; il1->num_used = il2->num_used; il1->num_allocated = il2->num_allocated; } */ double _get_match_score(PyObject *py_sequenceA, PyObject *py_sequenceB, PyObject *py_match_fn, int i, int j, char *sequenceA, char *sequenceB, int use_sequence_cstring, double match, double mismatch, int use_match_mismatch_scores) { PyObject *py_A=NULL, *py_B=NULL; PyObject *py_arglist=NULL, *py_result=NULL; double score = 0; if(use_sequence_cstring && use_match_mismatch_scores) { score = (sequenceA[i] == sequenceB[j]) ? match : mismatch; return score; } /* Calculate the match score. */ if(!(py_A = PySequence_GetItem(py_sequenceA, i))) goto _get_match_score_cleanup; if(!(py_B = PySequence_GetItem(py_sequenceB, j))) goto _get_match_score_cleanup; if(!(py_arglist = Py_BuildValue("(OO)", py_A, py_B))) goto _get_match_score_cleanup; if(!(py_result = PyEval_CallObject(py_match_fn, py_arglist))) goto _get_match_score_cleanup; score = PyNumber_AsDouble(py_result); _get_match_score_cleanup: if(py_A) { Py_DECREF(py_A); } if(py_B) { Py_DECREF(py_B); } if(py_arglist) { Py_DECREF(py_arglist); } if(py_result) { Py_DECREF(py_result); } return score; } /* This function is a more-or-less straightforward port of the * equivalent function in pairwise2. Please see there for algorithm * documentation. */ static PyObject *cpairwise2__make_score_matrix_fast( PyObject *self, PyObject *args) { int i; int row, col; PyObject *py_sequenceA, *py_sequenceB, *py_match_fn; char *sequenceA=NULL, *sequenceB=NULL; int use_sequence_cstring; double open_A, extend_A, open_B, extend_B; int penalize_extend_when_opening, penalize_end_gaps; int align_globally, score_only; double first_A_gap, first_B_gap; double match, mismatch; int use_match_mismatch_scores; int lenA, lenB; double *score_matrix = (double *)NULL; struct IndexList *trace_matrix = (struct IndexList *)NULL; PyObject *py_score_matrix=NULL, *py_trace_matrix=NULL; double *row_cache_score = (double *)NULL, *col_cache_score = (double *)NULL; struct IndexList *row_cache_index = (struct IndexList *)NULL, *col_cache_index = (struct IndexList *)NULL; PyObject *py_retval = NULL; if(!PyArg_ParseTuple(args, "OOOddddiiii", &py_sequenceA, &py_sequenceB, &py_match_fn, &open_A, &extend_A, &open_B, &extend_B, &penalize_extend_when_opening, &penalize_end_gaps, &align_globally, &score_only)) return NULL; if(!PySequence_Check(py_sequenceA) || !PySequence_Check(py_sequenceB)) { PyErr_SetString(PyExc_TypeError, "py_sequenceA and py_sequenceB should be sequences."); return NULL; } /* Optimize for the common case. Check to see if py_sequenceA and py_sequenceB are strings. If they are, use the c string representation. */ use_sequence_cstring = 0; if(PyString_Check(py_sequenceA) && PyString_Check(py_sequenceB)) { sequenceA = PyString_AS_STRING(py_sequenceA); sequenceB = PyString_AS_STRING(py_sequenceB); use_sequence_cstring = 1; } if(!PyCallable_Check(py_match_fn)) { PyErr_SetString(PyExc_TypeError, "py_match_fn must be callable."); return NULL; } /* Optimize for the common case. Check to see if py_match_fn is an identity_match. If so, pull out the match and mismatch member variables and calculate the scores myself. */ match = mismatch = 0; use_match_mismatch_scores = 0; if(PyInstance_Check(py_match_fn)) { PyObject *py_match=NULL, *py_mismatch=NULL; if(!(py_match = PyObject_GetAttrString(py_match_fn, "match"))) goto cleanup_after_py_match_fn; match = PyNumber_AsDouble(py_match); if(PyErr_Occurred()) goto cleanup_after_py_match_fn; if(!(py_mismatch = PyObject_GetAttrString(py_match_fn, "mismatch"))) goto cleanup_after_py_match_fn; mismatch = PyNumber_AsDouble(py_mismatch); if(PyErr_Occurred()) goto cleanup_after_py_match_fn; use_match_mismatch_scores = 1; cleanup_after_py_match_fn: if(PyErr_Occurred()) PyErr_Clear(); if(py_match) { Py_DECREF(py_match); } if(py_mismatch) { Py_DECREF(py_mismatch); } } /* Cache some commonly used gap penalties */ first_A_gap = calc_affine_penalty(1, open_A, extend_A, penalize_extend_when_opening); first_B_gap = calc_affine_penalty(1, open_B, extend_B, penalize_extend_when_opening); /* Allocate matrices for storing the results and initalize them. */ lenA = PySequence_Length(py_sequenceA); lenB = PySequence_Length(py_sequenceB); score_matrix = (double *)malloc(lenA*lenB*sizeof(*score_matrix)); trace_matrix = (struct IndexList *)malloc(lenA*lenB*sizeof(*trace_matrix)); if(!score_matrix || !trace_matrix) { PyErr_SetString(PyExc_MemoryError, "Out of memory"); goto _cleanup_make_score_matrix_fast; } for(i=0; i 1) { row_score = row_cache_score[row-1]; } else { row_score = nogap_score-1; /* Make sure it's not best score */ } if(row > 1) { col_score = col_cache_score[col-1]; } else { col_score = nogap_score-1; /* Make sure it's not best score */ } best_score = (row_score > col_score) ? row_score : col_score; if(nogap_score > best_score) best_score = nogap_score; best_score_rint = rint(best_score); /* Set the score and traceback matrices. */ score = best_score + _get_match_score(py_sequenceA, py_sequenceB, py_match_fn, row, col, sequenceA, sequenceB, use_sequence_cstring, match, mismatch, use_match_mismatch_scores); if(PyErr_Occurred()) goto _cleanup_make_score_matrix_fast; if(!align_globally && score < 0) score_matrix[row*lenB+col] = 0; else score_matrix[row*lenB+col] = score; il = &trace_matrix[row*lenB+col]; if(best_score_rint == rint(nogap_score)) { IndexList_append(il, row-1, col-1); } if(best_score_rint == rint(row_score)) { IndexList_extend(il, &row_cache_index[row-1]); } if(best_score_rint == rint(col_score)) { IndexList_extend(il, &col_cache_index[col-1]); } /* Update the cached column scores. */ open_score = score_matrix[(row-1)*lenB+col-1] + first_B_gap; extend_score = col_cache_score[col-1] + extend_B; open_score_rint = rint(open_score); extend_score_rint = rint(extend_score); if(open_score_rint > extend_score_rint) { col_cache_score[col-1] = open_score; IndexList_clear(&col_cache_index[col-1]); IndexList_append(&col_cache_index[col-1], row-1, col-1); } else if(extend_score_rint > open_score_rint) { col_cache_score[col-1] = extend_score; } else { col_cache_score[col-1] = open_score; if(!IndexList_contains(&col_cache_index[col-1], row-1, col-1)) IndexList_append(&col_cache_index[col-1], row-1, col-1); } /* Update the cached row scores. */ open_score = score_matrix[(row-1)*lenB+col-1] + first_A_gap; extend_score = row_cache_score[row-1] + extend_A; open_score_rint = rint(open_score); extend_score_rint = rint(extend_score); if(open_score_rint > extend_score_rint) { row_cache_score[row-1] = open_score; IndexList_clear(&row_cache_index[row-1]); IndexList_append(&row_cache_index[row-1], row-1, col-1); } else if(extend_score_rint > open_score_rint) { row_cache_score[row-1] = extend_score; } else { row_cache_score[row-1] = open_score; if(!IndexList_contains(&row_cache_index[row-1], row-1, col-1)) IndexList_append(&row_cache_index[row-1], row-1, col-1); } } } /* Save the score and traceback matrices into real python objects. */ if(!(py_score_matrix = PyList_New(lenA))) goto _cleanup_make_score_matrix_fast; if(!(py_trace_matrix = PyList_New(lenA))) goto _cleanup_make_score_matrix_fast; for(row=0; rownum_used))) goto _cleanup_make_score_matrix_fast; for(i=0; inum_used; i++) { PyObject *py_index=NULL; int row = il->indexes[i*2], col = il->indexes[i*2+1]; if(!(py_index = Py_BuildValue("(ii)", row, col))) goto _cleanup_make_score_matrix_fast; PyList_SET_ITEM(py_indexlist, i, py_index); } } PyList_SET_ITEM(py_trace_row, col, py_indexlist); } } py_retval = Py_BuildValue("(OO)", py_score_matrix, py_trace_matrix); _cleanup_make_score_matrix_fast: if(score_matrix) free(score_matrix); if(trace_matrix) { for(i=0; i