/********************************************************************* * Copyright 1993, UCAR/Unidata * See netcdf/COPYRIGHT file for copying and redistribution conditions. * $Id: load.c,v 1.35 2009/11/17 18:15:08 dmh Exp $ *********************************************************************/ #include "config.h" #include #include #include #include #include #include #include "generic.h" #include "ncgen.h" #include "genlib.h" extern int netcdf_flag; extern int c_flag; extern int fortran_flag; #define MIN(a,b) (((a) < (b)) ? (a) : (b)) #define MAX(a,b) (((a) > (b)) ? (a) : (b)) #define fpr (void) fprintf /* * Remove trailing zeros (after decimal point) but not trailing decimal * point from ss, a string representation of a floating-point number that * might include an exponent part. */ static void tztrim( char *ss /* returned string representing dd */ ) { char *cp, *ep; cp = ss; if (*cp == '-') cp++; while(isdigit((int)*cp) || *cp == '.') cp++; if (*--cp == '.') return; ep = cp+1; while (*cp == '0') cp--; cp++; if (cp == ep) return; while (*ep) *cp++ = *ep++; *cp = '\0'; return; } /* generate C to put netCDF record from in-memory data */ static void gen_load_c( void *rec_start ) { int idim, ival; char *val_string = NULL; char *charvalp = NULL; short *shortvalp = NULL; int *intvalp = NULL; float *floatvalp = NULL; double *doublevalp = NULL; char stmnt[C_MAX_STMNT]; size_t stmnt_len; char s2[C_MAX_STMNT] = {'\0'}; if (!vars[varnum].has_data) return; cline(""); sprintf(stmnt, " {\t\t\t/* store %s */", vars[varnum].name); cline(stmnt); if (vars[varnum].ndims > 0) { if (vars[varnum].dims[0] == rec_dim) { sprintf(stmnt, " static size_t %s_start[RANK_%s];", vars[varnum].lname, vars[varnum].lname); cline(stmnt); sprintf(stmnt, " static size_t %s_count[RANK_%s];", vars[varnum].lname, vars[varnum].lname); cline(stmnt); } /* load variable with data values using static initialization */ sprintf(stmnt, " static %s %s[] = {", ncctype(vars[varnum].type), vars[varnum].lname); stmnt_len = strlen(stmnt); switch (vars[varnum].type) { case NC_CHAR: val_string = cstrstr((char *) rec_start, var_len); sprintf(s2, "%s", val_string); strlcat(stmnt, s2, C_MAX_STMNT); free(val_string); break; default: switch (vars[varnum].type) { case NC_BYTE: charvalp = (char *) rec_start; break; case NC_SHORT: shortvalp = (short *) rec_start; break; case NC_INT: intvalp = (int *) rec_start; break; case NC_FLOAT: floatvalp = (float *) rec_start; break; case NC_DOUBLE: doublevalp = (double *) rec_start; break; default: break; } for (ival = 0; ival < var_len-1; ival++) { switch (vars[varnum].type) { case NC_BYTE: assert(charvalp != NULL); sprintf(s2, "%d, ", *charvalp++); break; case NC_SHORT: assert(shortvalp != NULL); sprintf(s2, "%d, ", *shortvalp++); break; case NC_INT: assert(intvalp != NULL); sprintf(s2, "%ld, ", (long)*intvalp++); break; case NC_FLOAT: assert(floatvalp != NULL); sprintf(s2, "%.8g, ", *floatvalp++); break; case NC_DOUBLE: assert(doublevalp != NULL); sprintf(s2, "%#.16g", *doublevalp++); tztrim(s2); strcat(s2, ", "); break; default: break; } stmnt_len += strlen(s2); if (stmnt_len < C_MAX_STMNT) strcat(stmnt, s2); else { cline(stmnt); strcpy(stmnt,s2); stmnt_len = strlen(stmnt); } } for (;ival < var_len; ival++) { switch (vars[varnum].type) { case NC_BYTE: assert(charvalp != NULL); sprintf(s2, "%d", *charvalp); break; case NC_SHORT: assert(shortvalp != NULL); sprintf(s2, "%d", *shortvalp); break; case NC_INT: assert(intvalp != NULL); sprintf(s2, "%ld", (long)*intvalp); break; case NC_FLOAT: assert(floatvalp != NULL); sprintf(s2, "%.8g", *floatvalp); break; case NC_DOUBLE: assert(doublevalp != NULL); sprintf(s2, "%#.16g", *doublevalp++); tztrim(s2); break; default: break; } stmnt_len += strlen(s2); if (stmnt_len < C_MAX_STMNT) strcat(stmnt, s2); else { cline(stmnt); strcpy(stmnt,s2); stmnt_len = strlen(stmnt); } } break; } strcat(stmnt,"};"); cline(stmnt); if (vars[varnum].dims[0] == rec_dim) { sprintf(stmnt, " %s_len = %lu; /* number of records of %s data */", dims[rec_dim].lname, (unsigned long)vars[varnum].nrecs, /* number of recs for this variable */ vars[varnum].name); cline(stmnt); for (idim = 0; idim < vars[varnum].ndims; idim++) { sprintf(stmnt, " %s_start[%d] = 0;", vars[varnum].lname, idim); cline(stmnt); } for (idim = 0; idim < vars[varnum].ndims; idim++) { sprintf(stmnt, " %s_count[%d] = %s_len;", vars[varnum].lname, idim, dims[vars[varnum].dims[idim]].lname); cline(stmnt); } } if (vars[varnum].dims[0] == rec_dim) { sprintf(stmnt, " stat = nc_put_vara_%s(ncid, %s_id, %s_start, %s_count, %s);", ncstype(vars[varnum].type), vars[varnum].lname, vars[varnum].lname, vars[varnum].lname, vars[varnum].lname); } else { /* non-record variables */ sprintf(stmnt, " stat = nc_put_var_%s(ncid, %s_id, %s);", ncstype(vars[varnum].type), vars[varnum].lname, vars[varnum].lname); } cline(stmnt); } else { /* scalar variables */ /* load variable with data values using static initialization */ sprintf(stmnt, " static %s %s = ", ncctype(vars[varnum].type), vars[varnum].lname); switch (vars[varnum].type) { case NC_CHAR: val_string = cstrstr((char *) rec_start, var_len); val_string[strlen(val_string)-1] = '\0'; sprintf(s2, "'%s'", &val_string[1]); free(val_string); break; case NC_BYTE: charvalp = (char *) rec_start; sprintf(s2, "%d", *charvalp); break; case NC_SHORT: shortvalp = (short *) rec_start; sprintf(s2, "%d", *shortvalp); break; case NC_INT: intvalp = (int *) rec_start; sprintf(s2, "%ld", (long)*intvalp); break; case NC_FLOAT: floatvalp = (float *) rec_start; sprintf(s2, "%.8g", *floatvalp); break; case NC_DOUBLE: doublevalp = (double *) rec_start; sprintf(s2, "%#.16g", *doublevalp++); tztrim(s2); break; default: break; } strlcat(stmnt, s2, C_MAX_STMNT); strlcat(stmnt,";", C_MAX_STMNT); cline(stmnt); sprintf(stmnt, " stat = nc_put_var_%s(ncid, %s_id, &%s);", ncstype(vars[varnum].type), vars[varnum].lname, vars[varnum].lname); cline(stmnt); } cline(" check_err(stat,__LINE__,__FILE__);"); cline(" }"); } /* * Add to a partial Fortran statement, checking if it's too long. If it is too * long, output the first part of it as a single statement with continuation * characters and start a new (probably invalid) statement with the remainder. * This will cause a Fortran compiler error, but at least all the information * will be available. */ static void fstrcat( char *s, /* source string of stement being built */ const char *t, /* string to be appended to source */ size_t *slenp /* pointer to length of source string */ ) { *slenp += strlen(t); if (*slenp >= FORT_MAX_STMNT) { derror("FORTRAN statement too long: %s",s); fline(s); strncpy(s, t, FORT_MAX_STMNT); *slenp = strlen(s); } else { /* Suppress a coverity-related issue without actually ignoring it in the coverity dashboard. */ /* coverity[unsigned_compare] */ strncat(s, t, MAX(0,MIN(strlen(t),strlen(s)-(strlen(t))))); } } /* * Create Fortran data statement to initialize numeric variable with * values. */ static void f_var_init( int varnum, /* which variable */ void *rec_start /* start of data */ ) { char *val_string; char *charvalp; short *shortvalp; int *intvalp; float *floatvalp; double *doublevalp; char stmnt[FORT_MAX_STMNT]; size_t stmnt_len; char s2[FORT_MAX_STMNT]; int ival; /* load variable with data values */ sprintf(stmnt, "data %s /",vars[varnum].lname); stmnt_len = strlen(stmnt); switch (vars[varnum].type) { case NC_BYTE: charvalp = (char *) rec_start; for (ival = 0; ival < var_len-1; ival++) { val_string = fstring(NC_BYTE,(void *)charvalp++,0); sprintf(s2, "%s, ", val_string); fstrcat(stmnt, s2, &stmnt_len); free(val_string); } val_string = fstring(NC_BYTE,(void *)charvalp++,0); fstrcat(stmnt, val_string, &stmnt_len); free(val_string); break; case NC_SHORT: shortvalp = (short *) rec_start; for (ival = 0; ival < var_len-1; ival++) { sprintf(s2, "%d, ", *shortvalp++); fstrcat(stmnt, s2, &stmnt_len); } sprintf(s2, "%d", *shortvalp); fstrcat(stmnt, s2, &stmnt_len); break; case NC_INT: intvalp = (int *) rec_start; for (ival = 0; ival < var_len-1; ival++) { sprintf(s2, "%ld, ", (long)*intvalp++); fstrcat(stmnt, s2, &stmnt_len); } sprintf(s2, "%ld", (long)*intvalp); fstrcat(stmnt, s2, &stmnt_len); break; case NC_FLOAT: floatvalp = (float *) rec_start; for (ival = 0; ival < var_len-1; ival++) { sprintf(s2, "%.8g, ", *floatvalp++); fstrcat(stmnt, s2, &stmnt_len); } sprintf(s2, "%.8g", *floatvalp); fstrcat(stmnt, s2, &stmnt_len); break; case NC_DOUBLE: doublevalp = (double *) rec_start; for (ival = 0; ival < var_len-1; ival++) { sprintf(s2, "%#.16g", *doublevalp++); tztrim(s2); expe2d(s2); /* change 'e' to 'd' in exponent */ fstrcat(s2, ", ", &stmnt_len); fstrcat(stmnt, s2, &stmnt_len); } sprintf(s2, "%#.16g", *doublevalp++); tztrim(s2); expe2d(s2); fstrcat(stmnt, s2, &stmnt_len); break; default: derror("fstrstr: bad type"); break; } fstrcat(stmnt, "/", &stmnt_len); /* For record variables, store data statement for later use; otherwise, just print it. */ if (vars[varnum].ndims > 0 && vars[varnum].dims[0] == rec_dim) { char *dup_stmnt = emalloc(strlen(stmnt)+1); strcpy(dup_stmnt, stmnt); /* ULTRIX missing strdup */ vars[varnum].data_stmnt = dup_stmnt; } else { fline(stmnt); } } /* make Fortran to put record */ static void gen_load_fortran( void *rec_start ) { char stmnt[FORT_MAX_STMNT]; struct vars *v = &vars[varnum]; if (!v->has_data) return; if (v->ndims == 0 || v->dims[0] != rec_dim) { sprintf(stmnt, "* store %s", v->name); fline(stmnt); } /* generate code to initialize variable with values found in CDL input */ if (v->type != NC_CHAR) { f_var_init(varnum, rec_start); } else { v->data_stmnt = fstrstr(rec_start, valnum); } if (v->ndims >0 && v->dims[0] == rec_dim) { return; } if (v->type != NC_CHAR) { sprintf(stmnt, "iret = nf_put_var_%s(ncid, %s_id, %s)", nfftype(v->type), v->lname, v->lname); } else { char *char_expr = fstrstr(rec_start, valnum); if(strlen("iret = nf_put_var_(ncid, _id, )") + strlen(nfftype(v->type)) + strlen(v->lname) + strlen(char_expr) > FORT_MAX_STMNT) { derror("FORTRAN statement to assign values to %s too long!", v->lname); exit(9); } sprintf(stmnt, "iret = nf_put_var_%s(ncid, %s_id, %s)", nfftype(v->type), v->lname, char_expr); free(char_expr); } fline(stmnt); fline("call check_err(iret)"); } /* invoke netcdf calls (or generate C or Fortran code) to load netcdf variable * from in-memory data. Assumes following global variables set from yacc * parser: * int varnum - number of variable to be loaded. * struct vars[varnum] - structure containing info on variable, specifically * name, type, ndims, dims, fill_value, has_data * int rec_dim - id of record dimension, or -1 if none * struct dims[] - structure containing name and size of dimensions. */ int put_variable( void *rec_start /* points to data to be loaded */ ) { if (netcdf_flag) load_netcdf(rec_start); /* put variable values */ if (c_flag) /* create C code to put values */ gen_load_c(rec_start); if (fortran_flag) /* create Fortran code to put values */ gen_load_fortran(rec_start); return 0; } /* write out variable's data from in-memory structure */ void load_netcdf( void *rec_start ) { int idim; int stat = NC_NOERR; size_t start[NC_MAX_VAR_DIMS]; size_t count[NC_MAX_VAR_DIMS]; char *charvalp; short *shortvalp; int *intvalp; float *floatvalp; double *doublevalp; /* load values into variable */ switch (vars[varnum].type) { case NC_CHAR: case NC_BYTE: charvalp = (char *) rec_start; break; case NC_SHORT: shortvalp = (short *) rec_start; break; case NC_INT: intvalp = (int *) rec_start; break; case NC_FLOAT: floatvalp = (float *) rec_start; break; case NC_DOUBLE: doublevalp = (double *) rec_start; break; default: break; } if (vars[varnum].ndims > 0) { /* initialize start to upper left corner (0,0,0,...) */ start[0] = 0; if (vars[varnum].dims[0] == rec_dim) { count[0] = vars[varnum].nrecs; } else { count[0] = dims[vars[varnum].dims[0]].size; } } for (idim = 1; idim < vars[varnum].ndims; idim++) { start[idim] = 0; count[idim] = dims[vars[varnum].dims[idim]].size; } switch (vars[varnum].type) { case NC_BYTE: stat = nc_put_vara_schar(ncid, varnum, start, count, (signed char *)charvalp); break; case NC_CHAR: stat = nc_put_vara_text(ncid, varnum, start, count, charvalp); break; case NC_SHORT: stat = nc_put_vara_short(ncid, varnum, start, count, shortvalp); break; case NC_INT: stat = nc_put_vara_int(ncid, varnum, start, count, intvalp); break; case NC_FLOAT: stat = nc_put_vara_float(ncid, varnum, start, count, floatvalp); break; case NC_DOUBLE: stat = nc_put_vara_double(ncid, varnum, start, count, doublevalp); break; default: break; } check_err(stat); }