| 1 |
#ifndef GFASEQGET_H |
| 2 |
#define GFASEQGET_H |
| 3 |
|
| 4 |
#include "GBase.h" |
| 5 |
#include "GList.hh" |
| 6 |
|
| 7 |
|
| 8 |
#define MAX_FASUBSEQ 0x20000000 |
| 9 |
//max 512MB sequence data held in memory at a time |
| 10 |
|
| 11 |
class GSubSeq { |
| 12 |
public: |
| 13 |
uint sqstart; //1-based coord of subseq start on sequence |
| 14 |
uint sqlen; //length of subseq loaded |
| 15 |
char* sq; //actual subsequence data will be stored here |
| 16 |
// (with end-of-line characters removed) |
| 17 |
|
| 18 |
/*char* xseq; //the exposed pointer to the last requested subsequence start |
| 19 |
off_t xstart; //the coordinate start for the last requested subseq |
| 20 |
off_t xlen; //the last requested subseq len*/ |
| 21 |
GSubSeq() { |
| 22 |
sqstart=0; |
| 23 |
sqlen=0; |
| 24 |
sq=NULL; |
| 25 |
/* xseq=NULL; |
| 26 |
xstart=0; |
| 27 |
xlen=0;*/ |
| 28 |
} |
| 29 |
~GSubSeq() { |
| 30 |
GFREE(sq); |
| 31 |
} |
| 32 |
// genomic, 1-based coordinates: |
| 33 |
void setup(uint sstart, int slen, int sovl=0, int qfrom=0, int qto=0); |
| 34 |
//check for overlap with previous window and realloc/extend appropriately |
| 35 |
//returns offset from seq that corresponds to sstart |
| 36 |
// the window will keep extending until MAX_FASUBSEQ is reached |
| 37 |
}; |
| 38 |
|
| 39 |
class GFaSeqGet { |
| 40 |
char* fname; |
| 41 |
FILE* fh; |
| 42 |
//raw offset in the file where the sequence actually starts: |
| 43 |
off_t fseqstart; |
| 44 |
int linelen; //length of each sequence line (assumed fixed) |
| 45 |
char lendlen; //length of end-of-line characters between lines |
| 46 |
//(assumed fixed) |
| 47 |
char lendch; //end-of-line signal character (can only be '\n' or '\r') |
| 48 |
GSubSeq* lastsub; |
| 49 |
void initialParse(off_t fofs=0, bool checkall=true); |
| 50 |
const char* loadsubseq(uint cstart, int& clen); |
| 51 |
void finit(const char* fn, off_t fofs, bool validate); |
| 52 |
public: |
| 53 |
GFaSeqGet() { |
| 54 |
fseqstart=0; |
| 55 |
linelen=0; |
| 56 |
lendch='\0'; |
| 57 |
fname=NULL; |
| 58 |
lastsub=NULL; |
| 59 |
} |
| 60 |
GFaSeqGet(const char* fn, off_t fofs, bool validate=false) { |
| 61 |
finit(fn,fofs,validate); |
| 62 |
} |
| 63 |
GFaSeqGet(const char* fn, bool validate=false) { |
| 64 |
finit(fn,0,validate); |
| 65 |
} |
| 66 |
/* |
| 67 |
GFaSeqGet(bool readAll, const char* fn, off_t fofs=0); |
| 68 |
GFaSeqGet(bool readAll, FILE* f, off_t fofs=0); |
| 69 |
*/ |
| 70 |
GFaSeqGet(FILE* f, off_t fofs=0, bool validate=false); |
| 71 |
~GFaSeqGet() { |
| 72 |
if (fname!=NULL) { |
| 73 |
GFREE(fname); |
| 74 |
fclose(fh); |
| 75 |
} |
| 76 |
delete lastsub; |
| 77 |
} |
| 78 |
const char* subseq(uint cstart, int& clen); |
| 79 |
const char* getRange(uint cstart, uint cend) { |
| 80 |
if (cstart>cend) { swap(cstart, cend); } |
| 81 |
int clen=cend-cstart+1; |
| 82 |
//int rdlen=clen; |
| 83 |
return subseq(cstart, clen); |
| 84 |
} |
| 85 |
//caller is responsible for deallocating copyRange() return string |
| 86 |
char* copyRange(uint cstart, uint cend, bool revCmpl=false, bool upCase=false); |
| 87 |
|
| 88 |
void loadall() { |
| 89 |
int clen=MAX_FASUBSEQ; |
| 90 |
subseq(1, clen); |
| 91 |
} |
| 92 |
void load(uint cstart, uint cend) { |
| 93 |
//cache as much as possible |
| 94 |
int clen=cend-cstart+1; |
| 95 |
subseq(cstart, clen); |
| 96 |
} |
| 97 |
int getsublen() { return lastsub!=NULL ? lastsub->sqlen : 0 ; } |
| 98 |
off_t getseqofs() { return fseqstart; } |
| 99 |
int getlinelen() { return linelen; } |
| 100 |
int getlendlen() { return lendlen; } |
| 101 |
//reads a subsequence starting at genomic coordinate cstart (1-based) |
| 102 |
}; |
| 103 |
|
| 104 |
|
| 105 |
#endif |
