mirror of
https://github.com/Unidata/netcdf-c.git
synced 2024-12-09 08:11:38 +08:00
fb40a72b45
re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
589 lines
14 KiB
C
589 lines
14 KiB
C
/*
|
|
* Copyright 2018, University Corporation for Atmospheric Research
|
|
* See netcdf/COPYRIGHT file for copying and redistribution conditions.
|
|
*/
|
|
|
|
#ifdef HAVE_CONFIG_H
|
|
#include "config.h"
|
|
#endif
|
|
|
|
#ifdef HAVE_UNISTD_H
|
|
#include <unistd.h>
|
|
#endif
|
|
#ifdef HAVE_GETOPT_H
|
|
#include <getopt.h>
|
|
#endif
|
|
|
|
#if defined(_WIN32) && !defined(__MINGW32__)
|
|
#include "XGetopt.h"
|
|
#endif
|
|
|
|
#include "zincludes.h"
|
|
#include "ncpathmgr.h"
|
|
#include "nclog.h"
|
|
#include "ncuri.h"
|
|
|
|
#undef DEBUG
|
|
|
|
#define DATANAME "data"
|
|
|
|
typedef enum Mapop {
|
|
MOP_NONE=0,
|
|
MOP_OBJDUMP=1,
|
|
MOP_CLEAR=2
|
|
} Mapop;
|
|
|
|
typedef enum OBJKIND {
|
|
OK_NONE=0,
|
|
OK_META=1,
|
|
OK_CHUNK=2,
|
|
OK_GROUP=3,
|
|
OK_IGNORE=4
|
|
} OBJKIND;
|
|
|
|
static struct Mops {
|
|
Mapop mapop;
|
|
const char* opname;
|
|
} mapops[] = {
|
|
{MOP_NONE,"none"},
|
|
{MOP_OBJDUMP,"objdump"},
|
|
{MOP_CLEAR,"clear"},
|
|
{MOP_NONE,NULL}
|
|
};
|
|
|
|
static struct Type {
|
|
const char* typename;
|
|
nc_type nctype;
|
|
int typesize;
|
|
const char format[16];
|
|
} types[] = {
|
|
{"ubyte",NC_UBYTE,1,"%u"},
|
|
{"byte",NC_BYTE,1,"%d"},
|
|
{"ushort",NC_USHORT,2,"%u"},
|
|
{"short",NC_SHORT,2,"%d"},
|
|
{"uint",NC_UINT,4,"%u"},
|
|
{"int",NC_INT,4,"%d"},
|
|
{"uint64",NC_UINT64,8,"%llu"},
|
|
{"int64",NC_INT64,8,"%lld"},
|
|
{"float",NC_FLOAT,4,"%f"},
|
|
{"double",NC_DOUBLE,8,"%lf"},
|
|
{"char",NC_CHAR,1,"'%c'"},
|
|
{"string",NC_STRING,sizeof(char*),"%*s"},
|
|
{NULL,NC_NAT,0,""}
|
|
};
|
|
|
|
/* Command line options */
|
|
struct Dumpptions {
|
|
int debug;
|
|
int meta_only;
|
|
Mapop mop;
|
|
char infile[4096];
|
|
NCZM_IMPL impl;
|
|
char* rootpath;
|
|
const struct Type* nctype;
|
|
char format[16];
|
|
int xflags;
|
|
# define XNOZMETADATA 1
|
|
int strlen;
|
|
} dumpoptions;
|
|
|
|
/* Forward */
|
|
static int objdump(void);
|
|
static NCZM_IMPL implfor(const char* path);
|
|
static void printcontent(size64_t len, const char* content, OBJKIND kind);
|
|
static int breadthfirst(NCZMAP* map, const char*, NClist* stack);
|
|
static char* rootpathfor(const char* path);
|
|
static OBJKIND keykind(const char* key);
|
|
static void sortlist(NClist* l);
|
|
static const char* filenamefor(const char* f0);
|
|
|
|
#define NCCHECK(expr) nccheck((expr),__LINE__)
|
|
static void nccheck(int stat, int line)
|
|
{
|
|
if(stat) {
|
|
fprintf(stderr,"%d: %s\n",line,nc_strerror(stat));
|
|
fflush(stderr);
|
|
exit(1);
|
|
}
|
|
}
|
|
|
|
static void
|
|
zmapusage(void)
|
|
{
|
|
fprintf(stderr,"usage: zmapio [-t <type>][-d][-v][-x] <file>\n");
|
|
exit(1);
|
|
}
|
|
|
|
static Mapop
|
|
decodeop(const char* name)
|
|
{
|
|
struct Mops* p = mapops;
|
|
for(;p->opname != NULL;p++) {
|
|
if(strcasecmp(p->opname,name)==0) return p->mapop;
|
|
}
|
|
return MOP_NONE;
|
|
}
|
|
|
|
static const struct Type*
|
|
decodetype(const char* name, int* strlenp)
|
|
{
|
|
struct Type* p = types;
|
|
|
|
if(strncmp(name,"string/",strlen("string/"))==0) {
|
|
*strlenp = atoi(name+strlen("string/"));
|
|
name = "string";
|
|
}
|
|
|
|
for(;p->typename != NULL;p++) {
|
|
if(strcasecmp(p->typename,name)==0) return p;
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
int
|
|
main(int argc, char** argv)
|
|
{
|
|
int stat = NC_NOERR;
|
|
int c;
|
|
char* p;
|
|
|
|
nc_initialize();
|
|
|
|
/* Init options */
|
|
memset((void*)&dumpoptions,0,sizeof(dumpoptions));
|
|
|
|
while ((c = getopt(argc, argv, "dhvx:t:F:T:X:")) != EOF) {
|
|
switch(c) {
|
|
case 'd':
|
|
dumpoptions.debug = 1;
|
|
break;
|
|
case 'h':
|
|
dumpoptions.meta_only = 1;
|
|
break;
|
|
case 't':
|
|
dumpoptions.nctype = decodetype(optarg,&dumpoptions.strlen);
|
|
if(dumpoptions.nctype == NULL) zmapusage();
|
|
break;
|
|
case 'x':
|
|
dumpoptions.mop = decodeop(optarg);
|
|
if(dumpoptions.mop == MOP_NONE) zmapusage();
|
|
break;
|
|
case 'v':
|
|
zmapusage();
|
|
goto done;
|
|
case 'F':
|
|
strcpy(dumpoptions.format,optarg);
|
|
break;
|
|
case 'T':
|
|
nctracelevel(atoi(optarg));
|
|
break;
|
|
case 'X':
|
|
for(p=optarg;*p;p++) {
|
|
switch (*p) {
|
|
case 'm': dumpoptions.xflags |= XNOZMETADATA; break;
|
|
default: fprintf(stderr,"Unknown -X argument: %c",*p); break;
|
|
}
|
|
};
|
|
break;
|
|
case '?':
|
|
fprintf(stderr,"unknown option\n");
|
|
goto fail;
|
|
}
|
|
}
|
|
|
|
/* Default the kind */
|
|
if(dumpoptions.nctype == NULL) {
|
|
dumpoptions.nctype = &types[0];
|
|
fprintf(stderr,"Default type: %s\n",dumpoptions.nctype->typename);
|
|
}
|
|
|
|
/* get file argument */
|
|
argc -= optind;
|
|
argv += optind;
|
|
|
|
if (argc > 1) {
|
|
fprintf(stderr, "zmapio: only one input file argument permitted\n");
|
|
goto fail;
|
|
}
|
|
if (argc == 0) {
|
|
fprintf(stderr, "zmapio: no input file specified\n");
|
|
goto fail;
|
|
}
|
|
|
|
{
|
|
char* p = NC_shellUnescape(argv[0]);
|
|
strcpy(dumpoptions.infile,filenamefor(p));
|
|
if(p) free(p);
|
|
}
|
|
|
|
if((dumpoptions.impl = implfor(dumpoptions.infile))== NCZM_UNDEF)
|
|
zmapusage();
|
|
|
|
if((dumpoptions.rootpath = rootpathfor(dumpoptions.infile))== NULL)
|
|
zmapusage();
|
|
|
|
switch (dumpoptions.mop) {
|
|
default:
|
|
fprintf(stderr,"Default action: objdump\n");
|
|
/* fall thru */
|
|
case MOP_OBJDUMP:
|
|
if((stat = objdump())) goto done;
|
|
break;
|
|
}
|
|
|
|
done:
|
|
/* Reclaim dumpoptions */
|
|
nullfree(dumpoptions.rootpath);
|
|
nc_finalize();
|
|
if(stat)
|
|
fprintf(stderr,"fail: %s\n",nc_strerror(stat));
|
|
return (stat ? 1 : 0);
|
|
fail:
|
|
stat = NC_EINVAL;
|
|
goto done;
|
|
}
|
|
|
|
static NCZM_IMPL
|
|
implfor(const char* path)
|
|
{
|
|
NCURI* uri = NULL;
|
|
const char* mode = NULL;
|
|
NClist* segments = nclistnew();
|
|
int i;
|
|
NCZM_IMPL impl = NCZM_UNDEF;
|
|
|
|
ncuriparse(path,&uri);
|
|
if(uri == NULL) goto done;
|
|
mode = ncurifragmentlookup(uri,"mode");
|
|
if(mode == NULL) goto done;
|
|
/* split on commas */
|
|
NCCHECK(nczm_split_delim(mode,',',segments));
|
|
for(i=0;i<nclistlength(segments);i++) {
|
|
const char* value = nclistget(segments,i);
|
|
if(strcmp(value,"file")==0) {impl = NCZM_FILE; goto done;}
|
|
if(strcmp(value,"zip")==0) {impl = NCZM_ZIP; goto done;}
|
|
if(strcmp(value,"s3")==0) {impl = NCZM_S3; goto done;}
|
|
}
|
|
done:
|
|
ncurifree(uri);
|
|
nclistfreeall(segments);
|
|
return impl;
|
|
}
|
|
|
|
static char*
|
|
rootpathfor(const char* path)
|
|
{
|
|
int stat = NC_NOERR;
|
|
NCURI* uri = NULL;
|
|
char* rootpath = NULL;
|
|
NClist* segments = nclistnew();
|
|
|
|
ncuriparse(path,&uri);
|
|
if(uri == NULL) goto done;
|
|
switch (dumpoptions.impl) {
|
|
case NCZM_FILE:
|
|
case NCZM_ZIP:
|
|
rootpath = strdup("/"); /*constant*/
|
|
break;
|
|
#ifdef ENABLE_S3
|
|
case NCZM_S3: {
|
|
char* p = NULL;
|
|
/* Split the path part */
|
|
if((stat = nczm_split(uri->path,segments))) goto done;
|
|
/* remove the bucket name */
|
|
p = (char*)nclistremove(segments,0);
|
|
nullfree(p); p = NULL;
|
|
/* Put it back together */
|
|
if((stat = nczm_join(segments,&rootpath))) goto done;
|
|
} break;
|
|
#endif
|
|
default:
|
|
stat = NC_EINVAL;
|
|
goto done;
|
|
}
|
|
|
|
done:
|
|
nclistfreeall(segments); segments = NULL;
|
|
ncurifree(uri); uri = NULL;
|
|
if(stat)
|
|
{nullfree(rootpath); rootpath = NULL;}
|
|
return rootpath;
|
|
}
|
|
|
|
static int
|
|
objdump(void)
|
|
{
|
|
int stat = NC_NOERR;
|
|
NCZMAP* map = NULL;
|
|
NClist* stack = nclistnew();
|
|
char* obj = NULL;
|
|
char* content = NULL;
|
|
int depth;
|
|
|
|
if((stat=nczmap_open(dumpoptions.impl, dumpoptions.infile, NC_NOCLOBBER, 0, NULL, &map)))
|
|
goto done;
|
|
|
|
/* Depth first walk all the groups to get all keys */
|
|
if((stat = breadthfirst(map,"/",stack))) goto done;
|
|
|
|
if(dumpoptions.debug) {
|
|
int i;
|
|
fprintf(stderr,"stack:\n");
|
|
for(i=0;i<nclistlength(stack);i++)
|
|
fprintf(stderr,"[%d] %s\n",i,(char*)nclistget(stack,i));
|
|
}
|
|
for(depth=0;nclistlength(stack) > 0;depth++) {
|
|
size64_t len = 0;
|
|
OBJKIND kind = 0;
|
|
int hascontent = 0;
|
|
nullfree(content); content = NULL;
|
|
nullfree(obj); obj = NULL;
|
|
obj = nclistremove(stack,0); /* zero pos is always top of stack */
|
|
kind = keykind(obj);
|
|
/* Now print info for this obj key */
|
|
switch (stat=nczmap_len(map,obj,&len)) {
|
|
case NC_NOERR: hascontent = 1; break;
|
|
case NC_EEMPTY: /* fall thru */ /* this is not a content bearing key */
|
|
case NC_EACCESS: hascontent = 0; len = 0; stat = NC_NOERR; break;
|
|
default: goto done;
|
|
}
|
|
if(!hascontent) continue; /* ignore it */
|
|
if(len > 0) {
|
|
size_t padlen = (len+dumpoptions.nctype->typesize);
|
|
content = calloc(1,padlen+1);
|
|
if((stat=nczmap_read(map,obj,0,len,content))) goto done;
|
|
content[len] = '\0';
|
|
}
|
|
if(hascontent) {
|
|
if(len > 0) {
|
|
assert(content != NULL);
|
|
if(kind == OK_CHUNK) {
|
|
len = ceildiv(len,dumpoptions.nctype->typesize);
|
|
}
|
|
printf("[%d] %s : (%llu)",depth,obj,len);
|
|
if(kind == OK_CHUNK && dumpoptions.nctype->nctype != NC_STRING)
|
|
printf(" (%s)",dumpoptions.nctype->typename);
|
|
printf(" |");
|
|
switch(kind) {
|
|
case OK_GROUP:
|
|
case OK_META:
|
|
printcontent(len,content,kind);
|
|
break;
|
|
case OK_CHUNK:
|
|
if(dumpoptions.meta_only)
|
|
printf("...");
|
|
else
|
|
printcontent(len,content,kind);
|
|
break;
|
|
default: break;
|
|
}
|
|
printf("|\n");
|
|
} else {
|
|
printf("[%d] %s : (%llu) ||\n",depth,obj,len);
|
|
}
|
|
} else {
|
|
printf("[%d] %s\n",depth,obj);
|
|
}
|
|
}
|
|
done:
|
|
nullfree(obj);
|
|
nullfree(content);
|
|
nczmap_close(map,0);
|
|
nclistfreeall(stack);
|
|
return stat;
|
|
}
|
|
|
|
/* Depth first walk all the groups to get all keys */
|
|
static int
|
|
breadthfirstR(NCZMAP* map, NCbytes* prefix, NClist* stack)
|
|
{
|
|
int stat = NC_NOERR;
|
|
NClist* nextlevel = nclistnew();
|
|
size_t mark;
|
|
const char* content;
|
|
int isroot = 0;
|
|
|
|
content = ncbytescontents(prefix);
|
|
if(content[0] == '/' && content[1] == '\0') isroot = 1;
|
|
if((stat=nczmap_search(map,content,nextlevel))) goto done;
|
|
/* Sort nextlevel */
|
|
sortlist(nextlevel);
|
|
/* Push new names onto the stack and recurse */
|
|
mark = ncbyteslength(prefix); /* save this position */
|
|
while(nclistlength(nextlevel) > 0) {
|
|
char* subkey = nclistremove(nextlevel,0);
|
|
if(!isroot) ncbytescat(prefix,"/");
|
|
ncbytescat(prefix,subkey);
|
|
nullfree(subkey);
|
|
nclistpush(stack,ncbytesdup(prefix));
|
|
if((stat = breadthfirstR(map,prefix,stack))) goto done;
|
|
ncbytessetlength(prefix,mark); ncbytesnull(prefix);
|
|
}
|
|
done:
|
|
nclistfreeall(nextlevel);
|
|
return stat;
|
|
}
|
|
|
|
/* Depth first walk all the groups to get all keys */
|
|
static int
|
|
breadthfirst(NCZMAP* map, const char* key, NClist* stack)
|
|
{
|
|
int stat = NC_NOERR;
|
|
NCbytes* prefix = ncbytesnew();
|
|
|
|
if(key == NULL || key[0] == '\0')
|
|
key = "/";
|
|
ncbytescat(prefix,key);
|
|
if(strlen(key) > 1 && key[strlen(key)-1]=='/') {
|
|
ncbytessetlength(prefix,ncbyteslength(prefix)-1); /* remove trailing '/' */
|
|
ncbytesnull(prefix);
|
|
}
|
|
stat = breadthfirstR(map,prefix,stack);
|
|
ncbytesfree(prefix);
|
|
return stat;
|
|
}
|
|
|
|
static char hex[16] = "0123456789abcdef";
|
|
|
|
static void
|
|
printcontent(size64_t len, const char* content, OBJKIND kind)
|
|
{
|
|
size64_t i, count;
|
|
unsigned int c0,c1;
|
|
|
|
const char* format = NULL;
|
|
int strlen = 1;
|
|
|
|
format = dumpoptions.nctype->format;
|
|
if(dumpoptions.format[0] != '\0')
|
|
format = dumpoptions.format;
|
|
strlen = dumpoptions.strlen;
|
|
count = len;
|
|
|
|
#ifdef DEBUG
|
|
printf("debug: len=%d strlen=%d count=%d\n",(int)len,(int)strlen,(int)count); fflush(stdout);
|
|
#endif
|
|
|
|
for(i=0;i<count;i++) {
|
|
/* If kind is chunk, then len is # of values, not # of bytes */
|
|
switch(kind) {
|
|
case OK_CHUNK:
|
|
if(i > 0) printf(", ");
|
|
switch(dumpoptions.nctype->nctype) {
|
|
case NC_BYTE: printf(format,((char*)content)[i]); break;
|
|
case NC_SHORT: printf(format,((short*)content)[i]); break;
|
|
case NC_INT: printf(format,((int*)content)[i]); break;
|
|
case NC_INT64: printf(format,((long long*)content)[i]); break;
|
|
case NC_UBYTE: printf(format,((unsigned char*)content)[i]); break;
|
|
case NC_USHORT: printf(format,((unsigned short*)content)[i]); break;
|
|
case NC_UINT: printf(format,((unsigned int*)content)[i]); break;
|
|
case NC_UINT64: printf(format,((unsigned long long*)content)[i]); break;
|
|
case NC_FLOAT: printf(format,((float*)content)[i]); break;
|
|
case NC_DOUBLE: printf(format,((double*)content)[i]); break;
|
|
case NC_CHAR: printf(format,((char*)content)[i]); break;
|
|
case NC_STRING: printf(format,(int)strlen,((char*)(&content[i*strlen]))); break;
|
|
default: abort();
|
|
}
|
|
break;
|
|
case OK_META:
|
|
printf("%c",content[i]);
|
|
break;
|
|
default:
|
|
c1 = (unsigned char)(content[i]);
|
|
c0 = c1 & 0xf;
|
|
c1 = (c1 >> 4);
|
|
c0 = hex[c0];
|
|
c1 = hex[c1];
|
|
printf("%c%c",(char)c1,(char)c0);
|
|
}
|
|
}
|
|
}
|
|
|
|
static char chunkchars[] = ".0123456789";
|
|
|
|
static OBJKIND
|
|
keykind(const char* key)
|
|
{
|
|
OBJKIND kind = OK_NONE;
|
|
char* suffix = NULL;
|
|
if(nczm_divide_at(key,-1,NULL,&suffix) == NC_NOERR) {
|
|
if(suffix) {
|
|
if(suffix[0] != '/')
|
|
kind = OK_NONE;
|
|
else if(suffix[1] == '.') {
|
|
if(strcmp(&suffix[1],".zmetadata")==0 && (dumpoptions.xflags & XNOZMETADATA))
|
|
kind = OK_IGNORE;
|
|
else
|
|
kind = OK_META;
|
|
} else if(suffix[strlen(suffix)-1] == '/')
|
|
kind = OK_GROUP;
|
|
else {
|
|
char* p = suffix+1;
|
|
for(;*p;p++) {
|
|
if(strchr(chunkchars,*p) == NULL) break;
|
|
}
|
|
kind = OK_CHUNK;
|
|
}
|
|
}
|
|
}
|
|
nullfree(suffix);
|
|
return kind;
|
|
}
|
|
|
|
/* bubble sort a list of strings */
|
|
static void
|
|
sortlist(NClist* l)
|
|
{
|
|
size_t i, switched;
|
|
|
|
if(nclistlength(l) <= 1) return;
|
|
do {
|
|
switched = 0;
|
|
for(i=0;i<nclistlength(l)-1;i++) {
|
|
char* ith = nclistget(l,i);
|
|
char* ith1 = nclistget(l,i+1);
|
|
if(strcmp(ith,ith1) > 0) {
|
|
nclistset(l,i,ith1);
|
|
nclistset(l,i+1,ith);
|
|
switched = 1;
|
|
}
|
|
}
|
|
} while(switched);
|
|
#if 0
|
|
for(i=0;i<nclistlength(l);i++)
|
|
fprintf(stderr,"sorted: [%d] %s\n",i,(const char*)nclistget(l,i));
|
|
#endif
|
|
}
|
|
|
|
static const char* urlexts[] = {"nzf", "zip", "nz4", NULL};
|
|
|
|
static const char*
|
|
filenamefor(const char* f0)
|
|
{
|
|
static char result[4096];
|
|
const char** extp;
|
|
char* p;
|
|
NCURI* uri = NULL;
|
|
|
|
strcpy(result,f0); /* default */
|
|
ncuriparse(f0,&uri);
|
|
if(uri == NULL) {
|
|
/* Not a URL */
|
|
p = strrchr(f0,'.'); /* look at the extension, if any */
|
|
if(p == NULL) goto done; /* No extension */
|
|
p++;
|
|
for(extp=urlexts;*extp;extp++) {
|
|
if(strcmp(p,*extp)==0) break;
|
|
}
|
|
if(*extp == NULL) goto done; /* not found */
|
|
/* Assemble the url */
|
|
strcpy(result,"file://");
|
|
strcat(result,f0); /* core path */
|
|
strcat(result,"#mode=nczarr,");
|
|
strcat(result,*extp);
|
|
}
|
|
done:
|
|
ncurifree(uri);
|
|
return result;
|
|
}
|