#include <stdio.h>

#include <stdlib.h>

#include <math.h>

#include <string.h>

#include <ctype.h>

#include <assert.h>

#include <omp.h>

#include <unistd.h>

#include <sys/types.h>

#include <limits.h>

#include "nrutil.h"

#include "random.h"

#include "data_interface.h"

#include "initial.h"

#include "check_converg.h"

#include "mcmc.h"

#include "result_analysis.h"

#include "quantile.h"

#include "poly_geno.h"

#define MAXLINE 10000

#define MAXLEN 100

static void param_decomp(int, char **);

static void printinfo(char *, int, char **,SEQDATA);

static void inf_K_val(char *outfilename, int n_small, int n_large, SEQDATA *data,INIT initial);

static double mem_cal(SEQDATA data,INIT initial);

double siglevel=0.900;

int nloci=100;

int popnum=2;

int totalsize=100;

int ploid=2;

long updatenum=1000000;

long burnin=500000;

int thinning=10;

int ckrep=20;

int GR_flag=1;

int chainnum=2;

char *missingdata="-9";

char *datafilename=NULL;

char *outfilename=NULL;

char *initialfilename=NULL;

char *convgfilename=NULL;

int label=1;

int popdata=1;

int prior_flag=0; //which prior for selfing rates, uniform (0) or DPM (1) Dirichlet Process prior

double alpha_dpm=10;

int back_refl=1;

int type_freq=1; //indicate which way to calculate genotype frequency,expectation way or structure way

int nstep_check_empty_cluster=20;

int n_extra_col=0;

int markername_flag=0;

int print_iter=1; //print the updated information of each iteration, 1=yes, 0=no

int print_freq=0; //print the allele frequencies to output file, 1=yes,0=no

int n_small=1;

int n_large=0;

int inf_K=0; //infer the number of subpopulations (1) or not (0)

int distr_fmt=1; //output follows Distruct format (1) or not (0)

int autopoly=1; //auto-tetraploid=1; allo-tetraploid=0

int data_fmt=0; //data_fmt=0 means one haploid per line;data_fmt=1 means one individual per line;

double max_mem=1.0e9;

int mode=1; //indicate which mode is used:

//infer population substructure without admixture only (0)

//infer population substructure only (1);

//infer substructure and selfing rates for populations (2);

//infer substructure and selfing rates for individuals (3);

//infer substructure and inbreeding coefficients for populations (4);

//infer substructure and inbreeding coefficients for individuals (5);

/*

* This application estimates the selfing rates for subpopulations and

* classfies each individual into subpopulations given the sequence data.

*

* Synopsis:

* InStruct -d data_file -o output_file [-i initial_file]

* [-K population number] [-L loci number] [-N total individual number]

* [-p ploid] [-u iteration number] [-b burn-in number] [-m missingdata]

* [-t thinning] [-c chain number] [-s seed1 seed2 seed3] [-sl significance level]

* [-lb label] [-a popdata] [-g GR_flag] [-r ckrep] [-f prior_flag] [-v mode]

* [-h alpha_dpm] [-e back_refl] [-y type_freq] [-j nstep_check_empty_cluster]

* [-x extra_columns] [-w markername] [-cf convgfilename] [-pi print_iter]

* [-pf print_freq] [-ik inf_K] [-kv n_small n_large] [-df distr_fmt] [-ap autopoly]

* [-af data_fmt] [-mm max_mem]

*

* Parameters:

* -d data_file - name of data file

* -o output_file - name of output file

* -i initial_file - name of initial file

* -K popnum - subpopulation number

* -L locinum - totoal loci number

* -N totalsize - total individual number

* -p ploid - the number of haplotype in a genome

* -u update - MCMC iteration number

* -b burnin - MCMC burn-in number

* -t thinning - MCMC thinning interval length

* -c chainnum - the number of MCMC chains

* -s seed1 seed2 seed3 three integers to override the default seed selection

* -m an integer represents missing data

* -sl significant level for the confidence intervals in result_analysis.c

* -lb label boolean indicates whether data_file contains labels

* for individuals, 1=yes, 0=no

* -a popdata boolean indicates whether data_file contains a column

* about the original population information, 1=yes, 0=no

* -g GR_flag boolean indicates whether Gelman_Rudin statistic

* is used to check convergence,1=yes, 0=no

* -r ckrep integer indicates how many stored iterations after

* burn-in are used in convergence checking

* -f prior_flag boolean indicates which prior for selfing rates,

* uniform (0) or DPM (1) Dirichlet Process prior

* -y boolean indicate which way to calculate genotype frequency,

* expectation way (0) or structure way (1)

* -e boolean indicate which proposal method for selfing rates,

* adaptive independence sampler(0) or back-reflection (1)

* -h alpha_dpm the scaling parameter "alpha" in Dirichlet Process Mixture model

* -j nstep_check_empty_cluster the number of iterations after burn-in

* that will be used to determine the existence of empty clusters

* -x extra_columns integer indicates the number of extra columns

* in data file besides label and popdata coloumns

* -w markernames boolean indicates existence of marker name

* line at the beginning of data file:

* no marker name line (0) or marker name line exist (1)

* -pi indicates whether to print the updated information of each iteration,

* etc. log-Likelihood. 1=yes, 0=no

* -cf convgfilename -name of the file storing updated values

* for convergence assessment

* -pf print_freq indicates whether to print the result of allele

* frequencies to output file, 1=yes, 0=no

* -ik inf_K indicates whether inferring the number of subpopulations or not

* -kv n_small n_large indicates the lower and upper boundary for value of K

* -df distr_fmt indicates whether to use the Distruct format for output (1) or not (0)

* -ap autopoly indicates whether the species is autopolyploid (1) or allopolyploid (0)

* -af data_fmt=0 means one haploid per line;data_fmt=1 means one individual per line;

* -mm

* -v mode integer indicate which mode is used:

* infer population substructure without admixture only (0);

* the rest options infer population structure with admixture:

* infer population substructure only (1);

* infer substructure and selfing rates for populations (2);

* infer substructure and selfing rates for individuals (3);

* infer substructure and inbreeding coefficients for populations (4);

* infer substructure and inbreeding coefficients for individuals (5);

*

* Example:

* ./InStruct -d example.str -o exampleout.txt -N 50 -L 50 -lb 1 -a 1 -w 1 -mm 3.0e9 -v 0 -ik 1 -kv 1 7 -c 1 -u 10000 -b 1250 -t 25 -r 250 -cf exampleconv.txt

* ./InStruct -d example.str -o exampleout.txt -N 50 -L 50 -lb 1 -a 1 -w 1 -mm 3.0e9 -v 0 -ik 1 -kv 1 7 -c 1 -u 10000 -b 1250 -t 25 -g 0

*/

int main(int argc, char **argv)

{

SEQDATA seqdata;

INIT initial;

CHAIN chain;

CONVG cvg;

int chn;

param_decomp(argc,argv);

seqdata=read_data(datafilename,ploid,totalsize,popnum,nloci,missingdata,label,popdata,siglevel,back_refl,type_freq,nstep_check_empty_cluster,prior_flag,mode,n_extra_col,markername_flag,alpha_dpm,print_iter,print_freq,inf_K,distr_fmt,autopoly,data_fmt,max_mem);

if(inf_K==0)

{

initial=read_init(initialfilename,chainnum,popnum,updatenum,burnin,thinning);

}

else{

initial=read_init(initialfilename,chainnum,n_large,updatenum,burnin,thinning);

}

if(mem_cal(seqdata,initial)>seqdata.max_mem)

{ nrerror("Your request of memory exceeds the maximum memory allowed! Please change the parameter max_mem");}

printinfo(outfilename,argc,argv,seqdata);

if(inf_K==1)

{ // make inference of K

inf_K_val(outfilename,n_small,n_large,&seqdata,initial);

}

else{

if(GR_flag==1) allocate_convg(seqdata,&cvg,chainnum,ckrep,convgfilename);

for(chn=0;chn<chainnum;chn++)

{

chain=mcmc_updating(seqdata,initial,chn,&cvg);

if(chain.flag_empty_cluster==1)

{

free_chain(&chain,seqdata);

chn--;

continue;

}

chain_stat(outfilename,chain,seqdata,chn);

free_chain(&chain,seqdata);

}

if(GR_flag==1)

{

chain_converg(outfilename,&cvg);

free_convg(&cvg);

}

}

fprintf(stdout,"THE JOB IS SUCCESSFULLY FINISHED\n");

return(0);

}

double mem_cal(SEQDATA data,INIT initial)

{

double temp=0;

int K=0;

if(inf_K==0) { K=data.popnum;}

else{ if(inf_K==1) { K=n_large;}}

if(data.print_freq==1) temp+=(double)(8*K*data.locinum*data.allelenum_max);

temp+=(double)(8+8*data.totalsize);

switch(data.mode)

{

case 2: temp+=(double)(K*8);break;

case 3: temp+=(double)(data.totalsize*8);break;

case 4: temp+=(double)(K*8);break;

case 5: temp+=(double)(data.totalsize*8); break;

}

temp+=(double)(data.totalsize*4);

temp+=(double)(8*data.totalsize*K);

temp*=(double)((initial.update-initial.burnin)/initial.thinning);

fprintf(stdout,"The memory required for this run is %f \n",temp);

fprintf(stdout,"The maximum memory allowed is %f \n",data.max_mem);

return(temp);

}

void param_decomp(int argc, char ** argv)

/*

* Function param_decomp decomposites the commandline arguments

*/

{

int i,j;

long *seeds,seednum=3;

char *msg;

seeds=lvector(0,seednum-1);

msg="Synopsis:\n\tInStruct -d data_file -o output_file [-i initial_file] [-K population number] [-L loci number] [-N total individual number] [-p ploid] [-u iteration number] [-b burn-in number] [-m missingdata] [-t thinning] [-c chain number] [-s seed1 seed2 seed3] [-sl significance level] [-lb label] [-a popdata] [-g GR_flag] [-r ckrep] [-f prior_flag] [-v mode] [-h alpha_dpm] [-e back_refl] [-y type_freq] [-j nstep_check_empty_cluster] [-x extra_columns] [-w markername] [-cf convgfilename] [-pi print_iter] [-pf print_freq] [-ik inf_K] [-kv n_small n_large] [-df distr_fmt] [-ap autopoly] [-af data_fmt] [-mm max_mem]\n";

if(argc==2&&strcmp(argv[1],"-h")==0) /*print help message*/

{

fprintf(stdout,"%s",msg);

exit(1);

}

else{

if(argc<5) /* partition the commandline arguments*/

{ nrerror("Too few arguments in the command line!");}

else{

for(i=1;i<argc;i++)

{

if(strcmp(argv[i],"-d")==0)

{ /*-d means to assign data file name*/

datafilename=argv[i+1];

continue;

}

if(strcmp(argv[i],"-o")==0)

{ /*-o means to assign output file name*/

outfilename=argv[i+1];

continue;

}

if(strcmp(argv[i],"-i")==0)

{ /*-i means to assign initial file name*/

initialfilename=argv[i+1];

continue;

}

if(strcmp(argv[i],"-cf")==0)

{ /*-cf means to assign initial file name*/

convgfilename=argv[i+1];

continue;

}

if(strcmp(argv[i],"-L")==0)

{ /*-L means to reassign the loci number a new value*/

nloci=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-N")==0)

{ /*-N means to reset the number of total individuals*/

totalsize=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-K")==0)

{ /*-K means to reassign population number a new value*/

popnum=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-p")==0)

{ /*-p means to reset the number of haplotype in a genome*/

ploid=atoi(argv[i+1]); /*for diploid, ploid=2*/

continue;

}

if(strcmp(argv[i],"-u")==0)

{ /*-u means to reset the number of update steps of MCMC*/

updatenum=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-b")==0)

{ /*-b means to reassign burnin number a new value*/

burnin=atoi(argv[i+1]);

if(burnin==0)

{ nrerror("Burn-in should not be zero!");}

continue;

}

if(strcmp(argv[i],"-t")==0)

{ /*-t means to reassign thinning number a new value */

thinning=atoi(argv[i+1]); /*thinning is to take iterations at an even interval*/

continue; /*which can reduces the autocorrelation between iterations*/

} /*and thinning can also reduces the memory needed*/

if(strcmp(argv[i],"-c")==0)

{ /*-c means to reassign thinning number a new value*/

chainnum=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-m")==0)

{ /*-m means to reset the number that represents missing data*/

missingdata=argv[i+1];

continue;

}

if(strcmp(argv[i],"-lb")==0)

{ /*-lb indicates whether data_file contains labels for individuals, 1=yes, 0=no*/

label=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-a")==0)

{ /*-a indicates whether data_file contains a column about the original population information, 1=yes, 0=no*/

popdata=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-g")==0)

{ /*-g indicates whether the Gelman_Rudin statistic is used to check convergence,1=yes, 0=no*/

GR_flag=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-f")==0)

{ /*-f indicates which prior is used for selfing rates, 0=uniform,1=normal,2=DPM*/

prior_flag=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-v")==0)

{ /*-v indicates whether selfing rates are wrt. pop (0) or individuals (1)*/

mode=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-r")==0)

{ /*-r indicates how many stored iterations after burn-in are used in convergence checking*/

ckrep=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-e")==0)

{ /*-e indicates which proposal method for selfing rates, adaptive independence sampler(0) or back-reflection (1)*/

back_refl=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-y")==0)

{ /*-y indicates which way to calculate genotype frequency, expectation way (0) or structure way (1)*/

type_freq=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-x")==0)

{ /*-x indicates the number of extra columns in data file*/

n_extra_col=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-pi")==0)

{ /*-pi indicates whether to print the information of each iteration along MCMC running*/

print_iter=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-ap")==0)

{ /*-ap indicates whether the species is autopolyploid (1) or allopolyploid (0) */

autopoly=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-pf")==0)

{ /*-pf indicates whether to print the result of allele frequencies to output file*/

print_freq=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-w")==0)

{ /*-w indicates existence of marker name line*/

markername_flag=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-af")==0)

{ /*-af indicates which format of input file is used*/

data_fmt=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-mm")==0)

{ /*-mm indicates maximum memory allowed*/

max_mem=atof(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-ik")==0)

{ /*-ik indicates whether inferring the number of subpopulations or not*/

inf_K=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-kv")==0)

{ /*-kv indicates the lower and upper boundary for value of K*/

n_small=atoi(argv[i+1]);

n_large=atoi(argv[i+2]);

continue;

}

if(strcmp(argv[i],"-df")==0)

{ /*-df indicates whether to use the Distruct format for output (1) or not (0)*/

distr_fmt=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-sl")==0)

{ /*-sl means to reset the significance level*/

siglevel=atof(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-h")==0)

{ /*-h means to reset the spread alpha in Dirichlet Process Mixture model*/

alpha_dpm=atof(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-j")==0)

{ /*-j means to reset the number of iterations after burn-in that will be used to determine the existence of empty clusters*/

nstep_check_empty_cluster=atoi(argv[i+1]);

continue;

}

if(strcmp(argv[i],"-s")==0)

{

for(j=0;j<seednum;j++) /*-s means to reset seeds for the random number generator*/

{

seeds[j]=atoi(argv[i+j+1]);

}

setseeds(seeds[0],seeds[1],seeds[2]);

continue;

}

}

}

}

if(ckrep>((updatenum-burnin)/thinning))

{

nrerror("The number of iterations for convergence assessment is greater than the total number of retained iterations from MCMC.");

}

if(nstep_check_empty_cluster>((updatenum-burnin)/thinning))

{

nrerror("The number of iterations for checking the existence of empty cluster is greater than the total number of retained iterations from MCMC.");

}

free_lvector(seeds,0,seednum-1);

}

void printinfo(char *outfilename,int argc, char **argv,SEQDATA data)

/*

* Print the basic information of the running condition

*/

{

int i;

FILE *outfile;

if((outfile=fopen(outfilename,"w"))==NULL)

{ nrerror("Cannot open output file!");}

//print to output file

fprintf(outfile,"\n");

for(i=0;i<MAXLEN;i++)

fprintf(outfile,"=");

fprintf(outfile,"\n\tInStruct by Gao, Williamson and Bustamante (2007)\n");

fprintf(outfile,"\t\t Code by Hong Gao\n");

fprintf(outfile,"\t\tVersion 1.0 (May. 2007)\n");

for(i=0;i<MAXLEN;i++)

fprintf(outfile,"=");

fprintf(outfile,"\n\n\n\nCommand line arguments:\n ");

for(i=0;i<argc;i++)

fprintf(outfile,"%s ",argv[i]);

fprintf(outfile,"\n\n");

fprintf(outfile,"Data File: %s\n",datafilename);

if(initialfilename!=NULL) fprintf(outfile,"Initial File: %s\n",initialfilename);

fprintf(outfile,"Output File: %s\n\n",outfilename);

fprintf(outfile,"\nRun parameters:\n");

fprintf(outfile," Chain Number=%d\n",chainnum);

fprintf(outfile," MCMC Iterations Number=%ld\n",updatenum);

fprintf(outfile," Burn-in=%ld\n",burnin);

fprintf(outfile," Thinning=%d\n",thinning);

fprintf(outfile," Ploid=%d\n",data.ploid);

if(data.ploid>2)

{ if(data.autopoly==1) fprintf(outfile,"Autopolyploid assumed\n");

else{ if(data.autopoly==0) fprintf(outfile,"Allopolyploid assumed\n");}

}

fprintf(outfile," Missing Data=%s\n",data.missingdata);

fprintf(outfile," Population size=%d\n",data.totalsize);

fprintf(outfile," Number of loci=%d\n",data.locinum);

fprintf(outfile," Population number assumed=%d\n",data.popnum);

fprintf(outfile," Significance level for Posterior Credible Interval=%f\n",data.siglevel);

fprintf(outfile," Mode = ");

if(data.ploid==2)

{

switch(data.mode)

{

case 0: fprintf(outfile,"Make inference of population structure only without admixture.\n");break;

case 1: fprintf(outfile,"Make inference of population structure only with admixture.\n");break;

case 2: fprintf(outfile,"Make inference of population structure and the selfing rates for subpopulations.\n");break;

case 3: fprintf(outfile,"Make inference of population structure and the selfing rates for individuals.\n");break;

case 4: fprintf(outfile,"Make inference of population structure and the inbreeding coefficients for subpopulations.\n");break;

case 5: fprintf(outfile,"Make inference of population structure and the inbreeding coefficients for individuals.\n");break;

}

}

else{

if(data.ploid==4)

{ fprintf(outfile,"Make inference of population structure and the selfing rates for subpopulations.\n");}

}

if(data.inf_K==1)

{ fprintf(outfile,"\nMake inference of the number of subpopulations.\n");}

if(data.mode==3||data.mode==5)

{

if(data.prior_flag==0) fprintf(outfile,"The Uniform prior is used for selfing rates.\n\n");

if(data.prior_flag==1) fprintf(outfile,"The Dirichlet Process prior is used for selfing rates and the scaling parameter is %f.\n\n",alpha_dpm);

}

switch(data.back_refl)

{

case 0: fprintf(outfile,"The proposal method for selfing rates is adaptive independence sampler.\n");break;

case 1: fprintf(outfile,"The proposal method for selfing rates is back-reflection.\n"); break;

}

if(data.print_freq==1)

{ fprintf(outfile,"The posterior allele frequencies will also be summarized and written to output file.\n");}

if(GR_flag==1) fprintf(outfile,"The %d stored iteration results after burn-in will be used to calculate the GR statistic.\n",ckrep);

if(data.distr_fmt==1)

{ fprintf(outfile,"The output of Q are generated in the Distruct format.\n");}

fclose(outfile);

}

void inf_K_val(char *outfilename, int n_small, int n_large, SEQDATA *data, INIT initial)

{

//flag=0;

int parallelism_enabled = 1; //0=no, not 0 = yes

int pid = getpid();

#pragma omp parallel if(parallelism_enabled)

{

int K,chn,K_best,K_num;//flag,,temp;

double **dic,*val_K;

CONVG cvg;

FILE *outfile;

CHAIN chain;

if(n_large<1||n_small<1||n_small>n_large)

{

n_small=1;

n_large=(int)pow((double)data->totalsize,0.3)+1;

fprintf(stdout,"The range of value for K is not correct! Change to default value (%d - %d)!\n",n_small,n_large);

}

K_num=n_large-n_small+1;

dic=dmatrix(0,K_num-1,0,initial.chainnum-1);

val_K=dvector(0,K_num-1);

/*//flag=0;

int parallelism_enabled = 1; //0=no, not 0 = yes

#pragma omp parallel if(parallelism_enabled)

{*/

//initialize private variables

char foutfilename[strlen(outfilename)];

#pragma omp for

for(K=n_small;K<=n_large;K++)

{

//modify outfilename so each K produces its own output

strcpy(foutfilename,outfilename); //start with original outfilename

//int K to char pK

int ndigits = floor(log10(K)) + 1;

char pK[ndigits]; //char array to contain int, sized to length of K

sprintf(pK, "%d", K); //convert integer K into a char array

//int pid to char pid

char ppid[(sizeof(int)*CHAR_BIT-1)/3 + 3];

sprintf(ppid, "%d", pid); //convert integer pid into a char array

char pp[2] = ".";

strncat(foutfilename, pp, 1); //add terminal "."

strncat(foutfilename, ppid, strlen(ppid)); //add process id

strncat(foutfilename, pp, 1); //add "."

strncat(foutfilename, pK, strlen(pK)); //add "K"

int tid = omp_get_thread_num();

printf("t%d,K%d,pK%s,pp%s,out(%s),fout(%s),pid(%d)\n",tid,K,pK,pp,outfilename,foutfilename,pid);

//getchar();

data->popnum=K;

printf("t%d,data.popnum%d\n",tid,(*data).popnum);

if((outfile=fopen(foutfilename,"a+"))==NULL)

{ nrerror("Cannot open output file!");}

fprintf(outfile,"\n\nThe current K is %d\n",K);

fclose(outfile);

if(GR_flag==1) allocate_convg(*data,&cvg,chainnum,ckrep,convgfilename);

for(chn=0;chn<initial.chainnum;chn++)

{

chain=mcmc_updating((*data),initial,chn,&cvg);

dic[K-n_small][chn]=chain_stat(foutfilename,chain,*data,chn);

free_chain(&chain,(*data));

}

if(GR_flag==1)

{

//temp=

chain_converg(foutfilename,&cvg);

free_convg(&cvg);

}

//if(temp==1) //bad convergence or only one chain

//{ flag=1;}

}

} //end omp parallel if

//COMBINE SEPARATE OUTPUT FILES

printf("combining files.\n");

char foutfilename[strlen(outfilename)];

int K;

for(K=n_small;K<=n_large;K++)

{

//modify outfilename so each K produces its own output

strcpy(foutfilename,outfilename); //start with original outfilename

//int K to char pK

int ndigits = floor(log10(K)) + 1;

char pK[ndigits]; //char array to contain int, sized to length of K

sprintf(pK, "%d", K); //convert integer K into a char array

//int pid to char pid

char ppid[(sizeof(int)*CHAR_BIT-1)/3 + 3];

sprintf(ppid, "%d", pid); //convert integer pid into a char array

char pp[2] = ".";

strncat(foutfilename, pp, 1); //add terminal "."

strncat(foutfilename, ppid, strlen(ppid)); //add process id

strncat(foutfilename, pp, 1); //add "."

strncat(foutfilename, pK, strlen(pK)); //add "K"

//open the base file (head) and the file to append (tail)

FILE *head = fopen(outfilename, "ab");

FILE *tail = fopen(foutfilename, "rb");

if (!head || !tail) abort();

//append

char buf[1024];

size_t n;

while ((n = fread(buf, 1, sizeof buf, tail)) > 0)

if (fwrite(buf, 1, n, head) != n) abort();

if (ferror(tail)) abort();

//close

fclose(head);

fclose(tail);

//delete the file just appended (to do)

}

/*

//CALCULATE BEST K USING DIC

//if(flag==1) //use the minimum DIC

//{

for(K=0;K<K_num;K++)

{

val_K[K]=dic[K][find_min(dic[K],initial.chainnum)];

}

K_best=find_min(val_K,K_num)+n_small;

//}

if((outfile=fopen(outfilename,"a+"))==NULL)

{ nrerror("Cannot open output file!");}

fprintf(outfile,"\n\nThe range of value for K is (%d - %d)!\n",n_small,n_large);

fprintf(outfile,"The optimal K is %d\n",K_best);

fclose(outfile);

free_dmatrix(dic,0,K_num-1,0,initial.chainnum-1);

free_dvector(val_K,0,K_num-1);

*/

}