mirror of https://github.com/ecmwf/eccodes.git
Improvement on file spliting
Allow an input file to be split into each individual message (if nchunk=-1)
This commit is contained in:
parent
741e62d234
commit
fea41048bd
|
@ -13,16 +13,19 @@
|
|||
* Split an input file (GRIB, BUFR etc) into chunks of roughly the same size.
|
||||
* The output files are named input_01, input_02 etc. This is much faster than grib_copy/bufr_copy
|
||||
*
|
||||
* 2019-07-26 W.Qu Allow an input file to be split into each individual message (if nchunk=-1)
|
||||
*
|
||||
*/
|
||||
|
||||
#include "grib_api_internal.h"
|
||||
#include <assert.h>
|
||||
|
||||
static int verbose = 0;
|
||||
static const char* OUTPUT_FILENAME_FORMAT = "%s_%02d"; /* x_01, x_02 etc */
|
||||
static const char* OUTPUT_FILENAME_FORMAT = "%s_%03d"; /* x_001, x_002 etc */
|
||||
static void usage(const char* prog)
|
||||
{
|
||||
printf("Usage: %s [-v] nchunks infile\n",prog);
|
||||
printf("nchunks=-1, split infile into individual grib/bufr message\n");
|
||||
exit(1);
|
||||
}
|
||||
|
||||
|
@ -30,8 +33,7 @@ static int split_file(FILE* in, const char* filename, const int nchunks, unsigne
|
|||
{
|
||||
void* mesg=NULL;
|
||||
FILE* out;
|
||||
off_t insize=0;
|
||||
size_t size=0,read_size=0,chunk_size;
|
||||
size_t size=0,read_size=0,insize=0,chunk_size, msg_size=0, num_msg=0;
|
||||
off_t offset=0;
|
||||
int err=GRIB_SUCCESS;
|
||||
int i;
|
||||
|
@ -45,11 +47,13 @@ static int split_file(FILE* in, const char* filename, const int nchunks, unsigne
|
|||
|
||||
fseeko(in, 0, SEEK_END);
|
||||
insize = ftello(in);
|
||||
if (insize==-1)
|
||||
return 1;
|
||||
fseeko(in, 0, SEEK_SET);
|
||||
if(nchunks == -1){
|
||||
chunk_size = size;
|
||||
}else{
|
||||
assert(nchunks > 0);
|
||||
chunk_size=insize/nchunks;
|
||||
}
|
||||
|
||||
i=1;
|
||||
sprintf(ofilename, OUTPUT_FILENAME_FORMAT, filename, i);
|
||||
|
@ -62,8 +66,10 @@ static int split_file(FILE* in, const char* filename, const int nchunks, unsigne
|
|||
|
||||
while ( err!=GRIB_END_OF_FILE ) {
|
||||
mesg=wmo_read_any_from_file_malloc(in, 0, &size, &offset, &err);
|
||||
if (mesg!=NULL && err==0) {
|
||||
if (fwrite(mesg,1,size,out)!=size) {
|
||||
num_msg++;
|
||||
/*printf("=1=%d\t%d\t%d\n",*count,size,insize);*/
|
||||
if ( mesg!=NULL && err==0 ) {
|
||||
if (fwrite(mesg,1,size,out)!=size ) {
|
||||
perror(ofilename);
|
||||
free(ofilename);
|
||||
fclose(out);
|
||||
|
@ -71,11 +77,13 @@ static int split_file(FILE* in, const char* filename, const int nchunks, unsigne
|
|||
}
|
||||
grib_context_free(c,mesg);
|
||||
read_size+=size;
|
||||
if (read_size>chunk_size) {
|
||||
if (verbose) printf("Wrote output file %s\n", ofilename);
|
||||
msg_size+=size;
|
||||
if (read_size>chunk_size && msg_size < insize) {
|
||||
if (verbose) printf("Wrote output file %s (%d msgs)\n", ofilename, num_msg);
|
||||
fclose(out);
|
||||
i++;
|
||||
/* Start writing to the next file */
|
||||
/*printf("=2=%d\t%d\n",*count,msg_size);*/
|
||||
sprintf(ofilename, OUTPUT_FILENAME_FORMAT, filename, i);
|
||||
out=fopen(ofilename,"w");
|
||||
if (!out) {
|
||||
|
@ -84,11 +92,12 @@ static int split_file(FILE* in, const char* filename, const int nchunks, unsigne
|
|||
return GRIB_IO_PROBLEM;
|
||||
}
|
||||
read_size=0;
|
||||
num_msg=0;
|
||||
}
|
||||
(*count)++;
|
||||
}
|
||||
}
|
||||
if (verbose) printf("Wrote output file %s\n", ofilename);
|
||||
if (verbose) printf("Wrote output file %s (%d msgs)\n", ofilename,num_msg-1);
|
||||
fclose(out);
|
||||
free(ofilename);
|
||||
|
||||
|
@ -117,8 +126,8 @@ int main(int argc,char* argv[])
|
|||
|
||||
/* add some error checking */
|
||||
nchunks=atoi(argv[i]);
|
||||
if (nchunks<1) {
|
||||
fprintf(stderr,"ERROR: Invalid number %d. Please specify a positive integer.\n", nchunks);
|
||||
if (nchunks<1 && nchunks!=-1) {
|
||||
fprintf(stderr,"ERROR: Invalid number %d. Please specify a positive integer. or -1 for spliting each message\n", nchunks);
|
||||
return 1;
|
||||
}
|
||||
|
||||
|
|
Loading…
Reference in New Issue