Author: Andreas Tille Description: Several patches for arb that enable new formats and contain other needed fixes diff -ubrN readseq-1.orig/readseq.c readseq-1/readseq.c --- readseq-1.orig/readseq.c 1993-02-01 01:00:00.000000000 +0100 +++ readseq-1/readseq.c 2007-11-14 12:14:36.000000000 +0100 @@ -93,6 +93,10 @@ = fix bug for possible memory overrun when truncating seqs for Phylip or Paup formats (thanks Anthony Persechini) + 13 Sep 96 GSt + RL (Steger@biophys.uni-duesseldorf.de) + * real time in MSF format (Main); #include + + added VIE multi sequence file format + + added LinAll sequence file format */ @@ -169,8 +173,11 @@ #include +#include /* MSch */ #include #include +#include /* RL */ +#include #include "ureadseq.h" @@ -199,9 +206,11 @@ "16. ASN.1", "17. PAUP/NEXUS", "18. Pretty (out-only)", + "19. LinAll", + "20. Vienna", "" }; -#define kFormCount 30 +#define kFormCount 32 #define kMaxFormName 15 const struct formatTable { @@ -238,6 +247,8 @@ {"paup", kPAUP}, {"nexus", kPAUP}, {"pretty", kPretty}, + {"linall", kLINALL}, + {"vie", kVIE}, }; const char *kASN1headline = "Bioseq-set ::= {\nseq-set {\n"; @@ -415,7 +426,7 @@ fprintf( stderr, " %-20s %-20s\n", formats[i], formats[midi+i]); fprintf(stderr,"\nChoose an output format (name or #): \n"); - gets(sform); + fgets(sform, 127, stdin); outform = parseformat(sform); if (outform == kNoformat) outform = kPearson; return outform; @@ -708,8 +719,12 @@ #else #define Exit(a) exit(a) +#ifdef NCBI +Nlm_Int2 Nlm_Main(void) +#else main( int argc, char *argv[]) #endif +#endif { boolean closein = false; short ifile, nseq, atseq, format, err = 0, seqtype = kDNA, @@ -721,6 +736,14 @@ char stempstore[256], *stemp = stempstore; FILE *ftmp, *fin, *fout; long outindexmax= 0, noutindex= 0, *outindex = NULL; +time_t time_val; /* GSt + RL */ +size_t size_timestr = 50;/* GSt + RL */ +char timestr[50]; /* GSt + RL */ + +#ifdef NCBI +int argc; +char** argv; +#endif #define exit_main(err) { \ if (closeout) fclose(fout); \ @@ -739,6 +762,10 @@ resetGlobals(); +#if NCBI + argc = Nlm_GetArgc(); + argv = Nlm_GetArgv(); +#endif foo = stdout; progname = argv[0]; *oname = 0; @@ -764,7 +791,7 @@ quietly = (dopipe || (gotinputfile && (listonly || whichSeq != 0))); - if (verbose || (!quietly && !gotinputfile)) fprintf( stderr, title); + //if (verbose || (!quietly && !gotinputfile)) fprintf( stderr, "%s\n", title); ifile = 1; /* UI: Choose output */ @@ -1003,6 +1030,13 @@ else if (dolower) for (i = 0; i +#define __NO_CTYPE +#include /* MSch */ #include #include #define UREADSEQ_G #include "ureadseq.h" +/* changed according to original which is the same with the changed header (at) */ #pragma segment ureadseq @@ -66,7 +69,7 @@ # define Local static /* local functions */ #endif -#define kStartLength 500 +#define kStartLength 500000 /* 20Apr93 temp. bug fix */ const char *aminos = "ABCDEFGHIKLMNPQRSTVWXYZ*"; const char *primenuc = "ACGTU"; @@ -101,6 +104,9 @@ long linestart; char s[256], *sp; +#ifdef ARB + int (*isseqcharfirst8)(); /* Patch by o. strunk (ARB) to allow numbers in genbank sequences*/ +#endif int (*isseqchar)(); /* int (*isseqchar)(int c); << sgi cc hates (int c) */ }; @@ -150,9 +156,23 @@ Local void addseq(char *s, struct ReadSeqVars *V) { char *ptr; +#ifdef ARB + /* Patch by o. strunk (ARB) to allow numbers in genbank sequences */ + int count = 0; +#endif +#ifdef ARB + if (V->addit){ + for (;*s != 0;s++,count++) { + if (count < 9 && V->isseqcharfirst8) { + if (!(V->isseqcharfirst8) (*s)) continue; + }else{ + if (!(V->isseqchar) (*s)) continue; + } +#else if (V->addit) while (*s != 0) { if ((V->isseqchar)(*s)) { +#endif if (V->seqlen >= V->maxseq) { V->maxseq += kStartLength; ptr = (char*) realloc(V->seq, V->maxseq+1); @@ -164,7 +184,9 @@ } V->seq[(V->seqlen)++] = *s; } +#ifndef ARB s++; +#endif } } @@ -324,6 +346,11 @@ Local void readGenBank(struct ReadSeqVars *V) { /*GenBank -- many seqs/file */ +#ifdef ARB + /* Patch by o. strunk (ARB) to allow numbers in genbank sequences */ + V->isseqchar = isSeqNumChar; + V->isseqcharfirst8 = isSeqChar; +#endif while (!V->allDone) { strcpy(V->seqid, (V->s)+12); while (! (feof(V->f) || strstr(V->s,"ORIGIN") == V->s)) @@ -337,9 +364,44 @@ } if (feof(V->f)) V->allDone = true; } +#ifdef ARB + V->isseqchar = isSeqChar; + V->isseqcharfirst8 = 0; +#endif } +Local boolean endVIE( boolean *addend, boolean *ungetend, struct ReadSeqVars *V) /* GSt + RL */ +{ + if (*V->s == '>') { /* start of next seq */ + *addend = false; + *ungetend= true; + return(true); + } + else + return(false); +} + + +Local void readVIE(struct ReadSeqVars *V) /* GSt + RL */ +{ + while (!V->allDone) { + strcpy(V->seqid, (V->s)+2); + readLoop(0, false, endVIE, V); + if (feof(V->f)) V->allDone = true; + } +/* + printf("readVIE: V->nseq = %d\n",V->nseq); + printf("readVIE: V->choice = %d\n",V->choice); + printf("readVIE: V->addit = %d\n",V->addit); + printf("readVIE: V->seqlen = %ld\n",V->seqlen); + printf("readVIE: V->seqid = %s\n",V->seqid); + printf("readVIE: V->s = %s\n",V->s); + printf("readVIE: V->seqid = %s\n",V->seqid); + printf("readVIE: V->s = %s\n<<s); +*/ +} + Local boolean endNBRF( boolean *addend, boolean *ungetend, struct ReadSeqVars *V) { char *a; @@ -449,6 +511,46 @@ } } +Local void readLINALL(struct ReadSeqVars *V) /* GSt */ +{ + /* SeqLen[I4] Label[Char*60] + Seq[Char*70 per line] + */ + int laenge; + int i; + + V->nseq++; /* but there is only a single sequence ? */ + // dprintf(("readLINALL: V->nseq = %d\n",V->nseq)); + /* V->addit = (V->choice > 0); */ /* what's that for ???? */ + // dprintf(("readLINALL: V->choice = %d\n",V->choice)); + // dprintf(("readLINALL: V->addit = %d\n",V->addit)); + // dprintf(("readLINALL: V->seqid = %s\n",V->seqid)); + // dprintf(("readLINALL: V->s = %s\n",V->s)); + /* if (V->addit) V->seqlen = 0; */ /* what's that for ???? */ + sscanf(V->s, "%4d", &laenge); /* seqlen is in 1st 4 chars of 1st line */ + // dprintf(("readLINALL: laenge = %d\n",laenge)); + // fflush(stdout); + strcpy(V->seqid, (V->s)+5); /* label starts after 5th char of 1st line */ + // dprintf(("readLINALL: V->seqid = %s\n",V->seqid)); + // fflush(stdout); + do { + V->done = feof(V->f); + getline(V); + if (!V->done) addseq((V->s), V); + } while ( !(V->done) && (V->seqlen)<=laenge ); + V->seqlen = laenge; /* only laenge chars are relevant for V->seq */ + // dprintf(("readLINALL: V->s = %s\n",V->s)); + /* if (V->choice == kListSequences) addinfo(V->seqid, V); */ /* what's that for ???? */ + // dprintf(("readLINALL: V->seqid = %s\n",V->seqid)); + // dprintf(("readLINALL: V->seqlen = %ld\n",V->seqlen)); + // dprintf(("readLINALL: V->seq =>")); + // for ( i=0; iseqlen; i++ ) dprintf(("%c",V->seq[i])); + // dprintf(("<\n")); + // dprintf(("readLINALL: V->s = %s\n<<s)); + V->allDone = true; + +} + Local boolean endFitch( boolean *addend, boolean *ungetend, struct ReadSeqVars *V) @@ -956,6 +1058,8 @@ case kZuker : readZuker(V); break; case kOlsen : readOlsen(V); break; case kMSF : readMSF(V); break; + case kLINALL: readLINALL(V); break; + case kVIE : readVIE(V); break; case kPAUP : { boolean done= false; @@ -1049,6 +1153,9 @@ V.err = 0; V.nseq = 0; V.isseqchar = isSeqChar; +#ifdef ARB + V.isseqcharfirst8 = 0; +#endif if (V.choice == kListSequences) ; /* leave as is */ else if (V.choice <= 0) V.choice = 1; /* default ?? */ V.addit = (V.choice > 0); @@ -1092,6 +1199,9 @@ V.err = 0; V.nseq = 0; V.isseqchar = isSeqChar; +#ifdef ARB + V.isseqcharfirst8 = 0; +#endif if (V.choice == kListSequences) ; /* leave as is */ else if (V.choice <= 0) V.choice = 1; /* default ?? */ V.addit = (V.choice > 0); @@ -1152,6 +1262,7 @@ boolean foundDNA= false, foundIG= false, foundStrider= false, foundGB= false, foundPIR= false, foundEMBL= false, foundNBRF= false, foundPearson= false, foundFitch= false, foundPhylip= false, foundZuker= false, + foundLINALL= false, foundVIE= false, gotolsen= false, gotpaup = false, gotasn1 = false, gotuw= false, gotMSF= false, isfitch= false, isphylip= false, done= false; short format= kUnknown; @@ -1159,6 +1270,8 @@ char sp[256]; long linestart=0; int maxlines2check=500; + int linallSeqLen; + char linallHeader[60]; #define ReadOneLine(sp) \ { done |= (feof(fseq)); \ @@ -1225,8 +1338,9 @@ foundPIR= true; else if (*sp == '>') { - if (sp[3] == ';') foundNBRF= true; - else foundPearson= true; + if (sp[3] == ';') foundNBRF= true; /* {foundNBRF= true; printf("foundNBRF\n");} */ + else if (sp[1] == ' ') foundVIE= true; /* {foundVIE= true; printf("foundVIE\n");} */ + else foundPearson= true; /* {foundPearson= true; printf("foundPearson\n");} */ } else if (strstr(sp,"ID ") == sp) @@ -1239,9 +1353,16 @@ else { if (nlines - *skiplines == 1) { - int ispp= 0, ilen= 0; - sscanf( sp, "%d%d", &ispp, &ilen); - if (ispp > 0 && ilen > 0) isphylip= true; + int ispp= 0, ilen= 0, icnt=0; + char junkstr[120]; + memset(junkstr,0,120); + icnt= sscanf( sp, "%d%d%c", &ispp, &ilen, junkstr); + if (icnt == 2 && ispp > 0 && ilen > 0) { + isphylip= true; + } + else if (icnt==3 && ispp > 0 && ilen > 0 && *junkstr == ' ') { + isphylip= true; + } } else if (isphylip && nlines - *skiplines == 2) { int tseq; @@ -1257,6 +1378,65 @@ } if (isfitch & (splen > 20)) foundFitch= true; +#ifdef DEBUG_LINALL + dprintf(("Check for LINALL\n")); + dprintf(("\tstrtol(sp,NULL,0) = %d\n",strtol(sp,NULL,0))); + dprintf(("\tnlines = %d\n",nlines)); + dprintf(("\tisphylip = %d\n",isphylip)); + dprintf(("\tfoundPhylip = %d\n",foundPhylip)); + dprintf(("\tisfitch = %d\n",isfitch)); + dprintf(("\tfoundFitch = %d\n",foundFitch)); +#endif + + /* + * This format detection was highly bogus ... + * Lesson 1: always initialize variables (in case the conversion fails...) + * Lesson 2: strings are passed to sscanf _without_ & (string variable is a pointer already) + * Lesson 3: forget to check return codes from syscalls: you lose. + */ + + if (nlines==1) { + int rv; +#ifdef DEBUG_LINALL + int i, sane=1; + char *spp; + + /* + * possible sanity check, for losers (see above) + */ + for (spp=sp, i=0; i<4; i++, spp++) + if (!(isspace(*spp) || isdigit(*spp))) { + dprintf(("bogus linall format header: %s\n", sp)); + sane=0; + break; + } +#endif + + linallSeqLen = 0; + *linallHeader = '\0'; + rv = sscanf( sp, "%d %s", &linallSeqLen, linallHeader); + +#ifdef DEBUG_LINALL + dprintf(("\tsscanf rval = %d\n",rv)); + dprintf(("\tlinallSeqLen = %d\n",linallSeqLen)); + dprintf(("\tlinallHeader = %s\n",linallHeader)); + dprintf(("\tlinallHeader = %d\n",strlen(linallHeader))); +#endif + + if (rv > 0 && linallSeqLen>0 && + strlen(linallHeader)>0 && + !(isphylip || foundPhylip)) { + /* !(isphylip || foundPhylip || isfitch || foundFitch)) { */ + foundLINALL= true; /* The 1st line contains the seqlength (4 digits), a blank, and a label (up to 60 char). */ + /* The following lines contain the sequence with 70 chars per line. */ +#ifdef DEBUG_LINALL + dprintf(("debug: foundLINALL: %ld<\n",strtol(sp,NULL,0))); + dprintf(("debug: sp:%s<\n\n",sp)); +#endif + } + + } + /* kRNA && kDNA are fairly certain...*/ switch (getseqtype( sp, splen)) { case kOtherSeq: otherlines++; break; @@ -1302,6 +1482,16 @@ done= true; } + else if (foundLINALL) { + format= kLINALL; + done= true; + } + + else if (foundVIE) { + format= kVIE; + done= true; + } + else if ((dnalines > 1) || done || (nlines > maxlines2check)) { /* decide on most likely format */ /* multichar idents: */ @@ -1785,6 +1975,27 @@ linesout += 2; break; + case kLINALL: /* GSt */ + fprintf(outf,"%4d %-60s\n",seqlen,seqname); + strcpy(endstr,"\n"); + linesout++; + width = 70; + tab = 0; + spacer = 0; + nameleft = false; + nameright = false; + numleft = false; + numright = false; + break; + + case kVIE: /* GSt + RL */ + if ( strchr(seqname,' ') != NULL ) seqname[strchr(seqname,' ')-seqname] = '\0'; /* no blanks in label line */ + fprintf(outf,"> %-s\n", seqname); + linesout++; + fprintf(outf,"%s\n\n",seq); /* complete sequence in one line; additional blank line before next sequence */ + return linesout; /* thus, do nothing else */ + break; + default : case kZuker: /* don't attempt Zuker's ftn format */ case kPearson: @@ -1841,7 +2052,8 @@ s[l++] = ' '; if (!baseonlynum) ibase++; else if (0==strchr(nocountsymbols,seq[i])) ibase++; - s[l++] = seq[i++]; + if (outform==kLINALL) { s[l++] = to_upper(seq[i]); i++; } /* GSt */ + else s[l++] = seq[i++] ; } if (l1 == width || i == seqlen) { diff -ubrN readseq-1.orig/ureadseq.h readseq-1/ureadseq.h --- readseq-1.orig/ureadseq.h 1992-12-30 01:00:00.000000000 +0100 +++ readseq-1/ureadseq.h 2007-11-14 12:14:36.000000000 +0100 @@ -66,8 +66,10 @@ #define kASN1 16 #define kPAUP 17 #define kPretty 18 +#define kLINALL 19 +#define kVIE 20 -#define kMaxFormat 18 +#define kMaxFormat 20 #define kMinFormat 1 #define kNoformat -1 /* format not tested */ #define kUnknown 0 /* format not determinable */ @@ -100,7 +102,7 @@ p.noleaves= p.domatch= p.degap= false;\ p.matchchar='.';\ p.gapchar='-';\ - p.namewidth=8;\ + p.namewidth=10;\ p.numwidth=5;\ p.interline=1;\ p.spacer=10;\