/** * @file * * Infer as much as possible from the omode + path. * Rewrite the path to a canonical form. * * Copyright 2018 University Corporation for Atmospheric * Research/Unidata. See COPYRIGHT file for more info. */ #include "config.h" #include #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_SYS_TYPES_H #include #endif #include "ncdispatch.h" #include "ncpathmgr.h" #include "netcdf_mem.h" #include "fbits.h" #include "ncbytes.h" #include "nclist.h" #include "nclog.h" #include "ncrc.h" #ifdef ENABLE_BYTERANGE #include "nchttp.h" #ifdef ENABLE_S3_SDK #include "ncs3sdk.h" #endif #endif #ifndef nulldup #define nulldup(x) ((x)?strdup(x):(x)) #endif #undef DEBUG /* If Defined, then use only stdio for all magic number io; otherwise use stdio or mpio as required. */ #undef USE_STDIO /** Sort info for open/read/close of file when searching for magic numbers */ struct MagicFile { const char* path; struct NCURI* uri; int omode; NCmodel* model; long long filelen; int use_parallel; int iss3; void* parameters; /* !NULL if inmemory && !diskless */ FILE* fp; #ifdef USE_PARALLEL MPI_File fh; #endif #ifdef ENABLE_BYTERANGE char* curlurl; /* url to use with CURLOPT_SET_URL */ NC_HTTP_STATE* state; #ifdef ENABLE_S3_SDK NCS3INFO s3; void* s3client; char* errmsg; #endif #endif }; /** @internal Magic number for HDF5 files. To be consistent with * H5Fis_hdf5, use the complete HDF5 magic number */ static char HDF5_SIGNATURE[MAGIC_NUMBER_LEN] = "\211HDF\r\n\032\n"; #define modelcomplete(model) ((model)->impl != 0) #ifdef DEBUG static void dbgflush(void) { fflush(stdout); fflush(stderr); } static void fail(int err) { return; } static int check(int err) { if(err != NC_NOERR) fail(err); return err; } #else #define check(err) (err) #endif /* Define a table of "mode=" string values from which the implementation can be inferred. Note that only cases that can currently take URLs are included. */ static struct FORMATMODES { const char* tag; const int impl; /* NC_FORMATX_XXX value */ const int format; /* NC_FORMAT_XXX value */ } formatmodes[] = { {"dap2",NC_FORMATX_DAP2,NC_FORMAT_CLASSIC}, {"dap4",NC_FORMATX_DAP4,NC_FORMAT_NETCDF4}, {"netcdf-3",NC_FORMATX_NC3,0}, /* Might be e.g. cdf5 */ {"classic",NC_FORMATX_NC3,0}, /* ditto */ {"netcdf-4",NC_FORMATX_NC4,NC_FORMAT_NETCDF4}, {"enhanced",NC_FORMATX_NC4,NC_FORMAT_NETCDF4}, {"udf0",NC_FORMATX_UDF0,0}, {"udf1",NC_FORMATX_UDF1,0}, {"nczarr",NC_FORMATX_NCZARR,NC_FORMAT_NETCDF4}, {"zarr",NC_FORMATX_NCZARR,NC_FORMAT_NETCDF4}, {"bytes",NC_FORMATX_NC4,NC_FORMAT_NETCDF4}, /* temporary until 3 vs 4 is determined */ {NULL,0}, }; /* Replace top-level name with defkey=defvalue */ static const struct MACRODEF { char* name; char* defkey; char* defvalues[4]; } macrodefs[] = { {"zarr","mode",{"nczarr","zarr",NULL}}, {"dap2","mode",{"dap2",NULL}}, {"dap4","mode",{"dap4",NULL}}, {"s3","mode",{"s3","nczarr",NULL}}, {"bytes","mode",{"bytes",NULL}}, {"xarray","mode",{"zarr", NULL}}, {"noxarray","mode",{"nczarr", "noxarray", NULL}}, {"zarr","mode",{"nczarr","zarr", NULL}}, {NULL,NULL,{NULL}} }; /* Mode inferences: if mode contains key value, then add the inferred value; Warning: be careful how this list is constructed to avoid infinite inferences. In order to (mostly) avoid that consequence, any attempt to infer a value that is already present will be ignored. This effectively means that the inference graph must be a DAG and may not have cycles. You have been warned. */ static const struct MODEINFER { char* key; char* inference; } modeinferences[] = { {"zarr","nczarr"}, {"xarray","zarr"}, {"noxarray","nczarr"}, {"noxarray","zarr"}, {NULL,NULL} }; /* Mode negations: if mode contains key, then remove all occurrences of the inference and repeat */ static const struct MODEINFER modenegations[] = { {"bytes","nczarr"}, /* bytes negates (nc)zarr */ {"bytes","zarr"}, {"noxarray","xarray"}, {NULL,NULL} }; /* Map FORMATX to readability to get magic number */ static struct Readable { int impl; int readable; } readable[] = { {NC_FORMATX_NC3,1}, {NC_FORMATX_NC_HDF5,1}, {NC_FORMATX_NC_HDF4,1}, {NC_FORMATX_PNETCDF,1}, {NC_FORMATX_DAP2,0}, {NC_FORMATX_DAP4,0}, {NC_FORMATX_UDF0,1}, {NC_FORMATX_UDF1,1}, {NC_FORMATX_NCZARR,0}, /* eventually make readable */ {0,0}, }; /* Define the known URL protocols and their interpretation */ static struct NCPROTOCOLLIST { const char* protocol; const char* substitute; const char* fragments; /* arbitrary fragment arguments */ } ncprotolist[] = { {"http",NULL,NULL}, {"https",NULL,NULL}, {"file",NULL,NULL}, {"dods","http","mode=dap2"}, {"dap4","http","mode=dap4"}, {"s3","s3","mode=s3"}, {NULL,NULL,NULL} /* Terminate search */ }; /* Forward */ static int NC_omodeinfer(int useparallel, int omode, NCmodel*); static int check_file_type(const char *path, int omode, int use_parallel, void *parameters, NCmodel* model, NCURI* uri); static int processuri(const char* path, NCURI** urip, NClist* fraglist); static int processmacros(NClist** fraglistp); static char* envvlist2string(NClist* pairs, const char*); static void set_default_mode(int* cmodep); static int parseonchar(const char* s, int ch, NClist* segments); static int mergelist(NClist** valuesp); static int openmagic(struct MagicFile* file); static int readmagic(struct MagicFile* file, long pos, char* magic); static int closemagic(struct MagicFile* file); static int NC_interpret_magic_number(char* magic, NCmodel* model); #ifdef DEBUG static void printmagic(const char* tag, char* magic,struct MagicFile*); static void printlist(NClist* list, const char* tag); #endif static int isreadable(NCURI*,NCmodel*); static char* list2string(NClist*); static int parsepair(const char* pair, char** keyp, char** valuep); static NClist* parsemode(const char* modeval); static const char* getmodekey(const NClist* envv); static int replacemode(NClist* envv, const char* newval); static void infernext(NClist* current, NClist* next); static int negateone(const char* mode, NClist* modes); static void cleanstringlist(NClist* strs, int caseinsensitive); /* If the path looks like a URL, then parse it, reformat it. */ static int processuri(const char* path, NCURI** urip, NClist* fraglenv) { int stat = NC_NOERR; int found = 0; NClist* tmp = NULL; struct NCPROTOCOLLIST* protolist; NCURI* uri = NULL; size_t pathlen = strlen(path); char* str = NULL; const char** ufrags; const char** p; if(path == NULL || pathlen == 0) {stat = NC_EURL; goto done;} /* Defaults */ if(urip) *urip = NULL; ncuriparse(path,&uri); if(uri == NULL) goto done; /* not url */ /* Look up the protocol */ for(found=0,protolist=ncprotolist;protolist->protocol;protolist++) { if(strcmp(uri->protocol,protolist->protocol) == 0) { found = 1; break; } } if(!found) {stat = NC_EINVAL; goto done;} /* unrecognized URL form */ /* process the corresponding fragments for that protocol */ if(protolist->fragments != NULL) { int i; tmp = nclistnew(); if((stat = parseonchar(protolist->fragments,'&',tmp))) goto done; for(i=0;isubstitute) ncurisetprotocol(uri,protolist->substitute); /* capture the fragments of the url */ ufrags = ncurifragmentparams(uri); if(ufrags != NULL) { for(p=ufrags;*p;p+=2) { const char* key = p[0]; const char* value = p[1]; nclistpush(fraglenv,nulldup(key)); value = (value==NULL?"":value); nclistpush(fraglenv,strdup(value)); } } if(urip) { *urip = uri; uri = NULL; } done: nclistfreeall(tmp); nullfree(str); if(uri != NULL) ncurifree(uri); return check(stat); } /* Split a key=value pair */ static int parsepair(const char* pair, char** keyp, char** valuep) { const char* p; char* key = NULL; char* value = NULL; if(pair == NULL) return NC_EINVAL; /* empty pair */ if(pair[0] == '\0' || pair[0] == '=') return NC_EINVAL; /* no key */ p = strchr(pair,'='); if(p == NULL) { value = NULL; key = strdup(pair); } else { ptrdiff_t len = (p-pair); if((key = malloc(len+1))==NULL) return NC_ENOMEM; memcpy(key,pair,len); key[len] = '\0'; if(p[1] == '\0') value = NULL; else value = strdup(p+1); } if(keyp) {*keyp = key; key = NULL;}; if(valuep) {*valuep = value; value = NULL;}; nullfree(key); nullfree(value); return NC_NOERR; } #if 0 static int parseurlmode(const char* modestr, NClist* list) { int stat = NC_NOERR; const char* p = NULL; const char* endp = NULL; if(modestr == NULL || *modestr == '\0') goto done; /* Split modestr at the commas or EOL */ p = modestr; for(;;) { char* s; ptrdiff_t slen; endp = strchr(p,','); if(endp == NULL) endp = p + strlen(p); slen = (endp - p); if((s = malloc(slen+1)) == NULL) {stat = NC_ENOMEM; goto done;} memcpy(s,p,slen); s[slen] = '\0'; nclistpush(list,s); if(*endp == '\0') break; p = endp+1; } done: return check(stat); } #endif /* Split a string at a given char */ static int parseonchar(const char* s, int ch, NClist* segments) { int stat = NC_NOERR; const char* p = NULL; const char* endp = NULL; if(s == NULL || *s == '\0') goto done; p = s; for(;;) { char* q; ptrdiff_t slen; endp = strchr(p,ch); if(endp == NULL) endp = p + strlen(p); slen = (endp - p); if((q = malloc(slen+1)) == NULL) {stat = NC_ENOMEM; goto done;} memcpy(q,p,slen); q[slen] = '\0'; nclistpush(segments,q); if(*endp == '\0') break; p = endp+1; } done: return check(stat); } /* Convert a key,value envv pairlist into a delimited string*/ static char* envvlist2string(NClist* envv, const char* delim) { int i; NCbytes* buf = NULL; char* result = NULL; if(envv == NULL || nclistlength(envv) == 0) return NULL; buf = ncbytesnew(); for(i=0;i 0) ncbytescat(buf,"&"); ncbytescat(buf,key); if(val != NULL && val[0] != '\0') { ncbytescat(buf,"="); ncbytescat(buf,val); } } result = ncbytesextract(buf); ncbytesfree(buf); return result; } /* Given a mode= argument, fill in the impl */ static int processmodearg(const char* arg, NCmodel* model) { int stat = NC_NOERR; struct FORMATMODES* format = formatmodes; for(;format->tag;format++) { if(strcmp(format->tag,arg)==0) { model->impl = format->impl; if(format->format != 0) model->format = format->format; } } return check(stat); } /* Given an envv fragment list, do macro replacement */ static int processmacros(NClist** fraglenvp) { int stat = NC_NOERR; const struct MACRODEF* macros = NULL; NClist* fraglenv = NULL; NClist* expanded = NULL; if(fraglenvp == NULL || nclistlength(*fraglenvp) == 0) goto done; fraglenv = *fraglenvp; expanded = nclistnew(); while(nclistlength(fraglenv) > 0) { int found = 0; char* key = NULL; char* value = NULL; key = nclistremove(fraglenv,0); /* remove from changing front */ value = nclistremove(fraglenv,0); /* remove from changing front */ if(strlen(value) == 0) { /* must be a singleton */ for(macros=macrodefs;macros->name;macros++) { if(strcmp(macros->name,key)==0) { char* const * p; nclistpush(expanded,strdup(macros->defkey)); for(p=macros->defvalues;*p;p++) nclistpush(expanded,strdup(*p)); found = 1; break; } } } if(!found) {/* pass thru */ nclistpush(expanded,strdup(key)); nclistpush(expanded,strdup(value)); } nullfree(key); nullfree(value); } *fraglenvp = expanded; expanded = NULL; done: nclistfreeall(expanded); nclistfreeall(fraglenv); return check(stat); } /* Process mode flag inferences */ static int processinferences(NClist* fraglenv) { int stat = NC_NOERR; const char* modeval = NULL; NClist* newmodes = nclistnew(); NClist* currentmodes = NULL; NClist* nextmodes = nclistnew(); int i; char* newmodeval = NULL; if(fraglenv == NULL || nclistlength(fraglenv) == 0) goto done; /* Get "mode" entry */ if((modeval = getmodekey(fraglenv))==NULL) goto done; /* Get the mode as list */ currentmodes = parsemode(modeval); #ifdef DEBUG printlist(currentmodes,"processinferences: initial mode list"); #endif /* Do what amounts to breadth first inferencing down the inference DAG. */ for(;;) { NClist* tmp = NULL; /* Compute the next set of inferred modes */ #ifdef DEBUG printlist(currentmodes,"processinferences: current mode list"); #endif infernext(currentmodes,nextmodes); #ifdef DEBUG printlist(nextmodes,"processinferences: next mode list"); #endif /* move current modes into list of newmodes */ for(i=0;ikey;tests++) { int i; if(strcasecmp(tests->key,mode)==0) { /* Find and remove all instances of the inference value */ for(i=nclistlength(newmodes)-1;i>=0;i--) { char* candidate = nclistget(newmodes,i); if(strcasecmp(candidate,tests->inference)==0) { nclistremove(newmodes,i); nullfree(candidate); changed = 1; } } } } return changed; } static void infernext(NClist* current, NClist* next) { int i; for(i=0;ikey;tests++) { if(strcasecmp(tests->key,cur)==0) { /* Append the inferred mode unless dup */ if(!nclistmatch(next,tests->inference,1)) nclistpush(next,strdup(tests->inference)); } } } } /* Given a list of strings, remove nulls and duplicates */ static int mergelist(NClist** valuesp) { int i,j; int stat = NC_NOERR; NClist* values = *valuesp; NClist* allvalues = nclistnew(); NClist* newvalues = nclistnew(); char* value = NULL; for(i=0;i 0) { value = nclistremove(allvalues,0); if(strlen(value) == 0) { nullfree(value); value = NULL; } else { for(j=0;jimpl = NC_FORMATX_UDF0; } else { model->impl = NC_FORMATX_UDF1; } if(fIsSet(cmode,NC_64BIT_OFFSET)) { model->format = NC_FORMAT_64BIT_OFFSET; } else if(fIsSet(cmode,NC_64BIT_DATA)) { model->format = NC_FORMAT_64BIT_DATA; } else if(fIsSet(cmode,NC_NETCDF4)) { if(fIsSet(cmode,NC_CLASSIC_MODEL)) model->format = NC_FORMAT_NETCDF4_CLASSIC; else model->format = NC_FORMAT_NETCDF4; } if(! model->format) model->format = NC_FORMAT_CLASSIC; goto done; } if(fIsSet(cmode,NC_64BIT_OFFSET)) { model->impl = NC_FORMATX_NC3; model->format = NC_FORMAT_64BIT_OFFSET; goto done; } if(fIsSet(cmode,NC_64BIT_DATA)) { model->impl = NC_FORMATX_NC3; model->format = NC_FORMAT_64BIT_DATA; goto done; } if(fIsSet(cmode,NC_NETCDF4)) { model->impl = NC_FORMATX_NC4; if(fIsSet(cmode,NC_CLASSIC_MODEL)) model->format = NC_FORMAT_NETCDF4_CLASSIC; else model->format = NC_FORMAT_NETCDF4; goto done; } /* Default to classic model */ model->format = NC_FORMAT_CLASSIC; model->impl = NC_FORMATX_NC3; done: /* Apply parallel flag */ if(useparallel) { if(model->impl == NC_FORMATX_NC3) model->impl = NC_FORMATX_PNETCDF; } return check(stat); } /* If the mode flags do not necessarily specify the format, then default it by adding in appropriate flags. */ static void set_default_mode(int* modep) { int mode = *modep; int dfaltformat; dfaltformat = nc_get_default_format(); switch (dfaltformat) { case NC_FORMAT_64BIT_OFFSET: mode |= NC_64BIT_OFFSET; break; case NC_FORMAT_64BIT_DATA: mode |= NC_64BIT_DATA; break; case NC_FORMAT_NETCDF4: mode |= NC_NETCDF4; break; case NC_FORMAT_NETCDF4_CLASSIC: mode |= (NC_NETCDF4|NC_CLASSIC_MODEL); break; case NC_FORMAT_CLASSIC: /* fall thru */ default: break; /* default to classic */ } *modep = mode; /* final result */ } /**************************************************/ /* Infer model for this dataset using some combination of cmode, path, and reading the dataset. See the documentation in docs/internal.dox. @param path @param omode @param iscreate @param useparallel @param params @param model @param newpathp */ int NC_infermodel(const char* path, int* omodep, int iscreate, int useparallel, void* params, NCmodel* model, char** newpathp) { int i,stat = NC_NOERR; NCURI* uri = NULL; int omode = *omodep; NClist* fraglenv = nclistnew(); NClist* modeargs = nclistnew(); char* sfrag = NULL; const char* modeval = NULL; char* abspath = NULL; /* Phase 1: 1. convert special protocols to http|https 2. begin collecting fragments */ if((stat = processuri(path, &uri, fraglenv))) goto done; if(uri != NULL) { #ifdef DEBUG printlist(fraglenv,"processuri"); #endif /* Phase 2: Expand macros and add to fraglenv */ if((stat = processmacros(&fraglenv))) goto done; #ifdef DEBUG printlist(fraglenv,"processmacros"); #endif /* Cleanup the fragment list */ if((stat = cleanfragments(&fraglenv))) goto done; /* Phase 2a: Expand mode inferences and add to fraglenv */ if((stat = processinferences(fraglenv))) goto done; #ifdef DEBUG printlist(fraglenv,"processinferences"); #endif /* Phase 3: coalesce duplicate fragment keys and remove duplicate values */ if((stat = cleanfragments(&fraglenv))) goto done; #ifdef DEBUG printlist(fraglenv,"cleanfragments"); #endif /* Phase 4: Rebuild the url fragment and rebuilt the url */ sfrag = envvlist2string(fraglenv,"&"); nclistfreeall(fraglenv); fraglenv = NULL; #ifdef DEBUG fprintf(stderr,"frag final: %s\n",sfrag); #endif ncurisetfragments(uri,sfrag); nullfree(sfrag); sfrag = NULL; /* If s3, then rebuild the url */ if(NC_iss3(uri)) { NCURI* newuri = NULL; if((stat = NC_s3urlrebuild(uri,&newuri,NULL,NULL))) goto done; ncurifree(uri); uri = newuri; } else if(strcmp(uri->protocol,"file")==0) { /* convert path to absolute */ char* canon = NULL; abspath = NCpathabsolute(uri->path); if((stat = NCpathcanonical(abspath,&canon))) goto done; nullfree(abspath); abspath = canon; canon = NULL; if((stat = ncurisetpath(uri,abspath))) goto done; } /* rebuild the path */ if(newpathp) { *newpathp = ncuribuild(uri,NULL,NULL,NCURIALL); #ifdef DEBUG fprintf(stderr,"newpath=|%s|\n",*newpathp); fflush(stderr); #endif } /* Phase 5: Process the mode key to see if we can tell the formatx */ modeval = ncurifragmentlookup(uri,"mode"); if(modeval != NULL) { if((stat = parseonchar(modeval,',',modeargs))) goto done; for(i=0;iimpl = NC_FORMATX_DAP2; model->format = NC_FORMAT_NC3; } } else {/* Not URL */ if(newpathp) *newpathp = NULL; } /* Phase 8: mode inference from mode flags */ /* The modeargs did not give us a model (probably not a URL). So look at the combination of mode flags and the useparallel flag */ if(!modelcomplete(model)) { if((stat = NC_omodeinfer(useparallel,omode,model))) goto done; } /* Phase 9: Infer from file content, if possible; this has highest precedence, so it may override previous decisions. Note that we do this last because we need previously determined model info to guess if this file is readable. */ if(!iscreate && isreadable(uri,model)) { /* Ok, we need to try to read the file */ if((stat = check_file_type(path, omode, useparallel, params, model, uri))) goto done; } /* Need a decision */ if(!modelcomplete(model)) {stat = NC_ENOTNC; goto done;} /* Force flag consistency */ switch (model->impl) { case NC_FORMATX_NC4: case NC_FORMATX_NC_HDF4: case NC_FORMATX_DAP4: case NC_FORMATX_NCZARR: omode |= NC_NETCDF4; if(model->format == NC_FORMAT_NETCDF4_CLASSIC) omode |= NC_CLASSIC_MODEL; break; case NC_FORMATX_NC3: omode &= ~NC_NETCDF4; /* must be netcdf-3 (CDF-1, CDF-2, CDF-5) */ if(model->format == NC_FORMAT_64BIT_OFFSET) omode |= NC_64BIT_OFFSET; else if(model->format == NC_FORMAT_64BIT_DATA) omode |= NC_64BIT_DATA; break; case NC_FORMATX_PNETCDF: omode &= ~NC_NETCDF4; /* must be netcdf-3 (CDF-1, CDF-2, CDF-5) */ if(model->format == NC_FORMAT_64BIT_OFFSET) omode |= NC_64BIT_OFFSET; else if(model->format == NC_FORMAT_64BIT_DATA) omode |= NC_64BIT_DATA; break; case NC_FORMATX_DAP2: omode &= ~(NC_NETCDF4|NC_64BIT_OFFSET|NC_64BIT_DATA|NC_CLASSIC_MODEL); break; case NC_FORMATX_UDF0: case NC_FORMATX_UDF1: if(model->format == NC_FORMAT_64BIT_OFFSET) omode |= NC_64BIT_OFFSET; else if(model->format == NC_FORMAT_64BIT_DATA) omode |= NC_64BIT_DATA; else if(model->format == NC_FORMAT_NETCDF4) omode |= NC_NETCDF4; else if(model->format == NC_FORMAT_NETCDF4_CLASSIC) omode |= NC_NETCDF4|NC_CLASSIC_MODEL; break; default: {stat = NC_ENOTNC; goto done;} } done: nullfree(sfrag); nullfree(abspath); ncurifree(uri); nclistfreeall(modeargs); nclistfreeall(fraglenv); *omodep = omode; /* in/out */ return check(stat); } static int isreadable(NCURI* uri, NCmodel* model) { int canread = 0; struct Readable* r; /* Step 1: Look up the implementation */ for(r=readable;r->impl;r++) { if(model->impl == r->impl) {canread = r->readable; break;} } /* Step 2: check for bytes mode */ if(!canread && NC_testmode(uri,"bytes") && (model->impl == NC_FORMATX_NC4 || model->impl == NC_FORMATX_NC_HDF5)) canread = 1; return canread; } #if 0 static char* emptyify(char* s) { if(s == NULL) s = strdup(""); return strdup(s); } static const char* nullify(const char* s) { if(s != NULL && strlen(s) == 0) return NULL; return s; } #endif /**************************************************/ /**************************************************/ /** * Provide a hidden interface to allow utilities * to check if a given path name is really a url. * If not, put null in basenamep, else put basename of the url path * minus any extension into basenamep; caller frees. * Return 1 if it looks like a url, 0 otherwise. */ int nc__testurl(const char* path0, char** basenamep) { NCURI* uri = NULL; int ok = 0; char* path = NULL; if(!ncuriparse(path0,&uri)) { char* p; char* q; path = strdup(uri->path); if(path == NULL||strlen(path)==0) goto done; p = strrchr(path, '/'); if(p == NULL) p = path; else p++; q = strrchr(p,'.'); if(q != NULL) *q = '\0'; if(strlen(p) == 0) goto done; if(basenamep) *basenamep = strdup(p); ok = 1; } done: ncurifree(uri); nullfree(path); return ok; } /**************************************************/ /* Envv list utilities */ static const char* getmodekey(const NClist* envv) { int i; /* Get "mode" entry */ for(i=0;i 0) ncbytescat(buf,","); ncbytescat(buf,m); } result = ncbytesextract(buf); ncbytesfree(buf); if(result == NULL) result = strdup(""); return result; } #if 0 /* Given a comma separated string, remove duplicates; mostly used to cleanup mode list */ static char* cleancommalist(const char* commalist, int caseinsensitive) { NClist* tmp = nclistnew(); char* newlist = NULL; if(commalist == NULL || strlen(commalist)==0) return nulldup(commalist); (void)parseonchar(commalist,',',tmp);/* split on commas */ cleanstringlist(tmp,caseinsensitive); newlist = list2string(tmp); nclistfreeall(tmp); return newlist; } #endif /* Given a list of strings, remove nulls and duplicated */ static void cleanstringlist(NClist* strs, int caseinsensitive) { int i,j; if(nclistlength(strs) == 0) return; /* Remove nulls */ for(i=nclistlength(strs)-1;i>=0;i--) { if(nclistget(strs,i)==NULL) nclistremove(strs,i); } /* Remove duplicates*/ for(i=0;ii;j--) { int match; const char* candidate = nclistget(strs,j); if(caseinsensitive) match = (strcasecmp(value,candidate) == 0); else match = (strcmp(value,candidate) == 0); if(match) {char* dup = nclistremove(strs,j); nullfree(dup);} } } } /**************************************************/ /** * @internal Given an existing file, figure out its format and return * that format value (NC_FORMATX_XXX) in model arg. Assume any path * conversion was already performed at a higher level. * * @param path File name. * @param flags * @param use_parallel * @param parameters * @param model Pointer that gets the model to use for the dispatch table. * @param version Pointer that gets version of the file. * * @return ::NC_NOERR No error. * @author Dennis Heimbigner */ static int check_file_type(const char *path, int omode, int use_parallel, void *parameters, NCmodel* model, NCURI* uri) { char magic[NC_MAX_MAGIC_NUMBER_LEN]; int status = NC_NOERR; struct MagicFile magicinfo; #ifdef _WIN32 NC* nc = NULL; #endif memset((void*)&magicinfo,0,sizeof(magicinfo)); #ifdef _WIN32 /* including MINGW */ /* Windows does not handle well multiple handles to the same file. So if file is already open/created, then find it and just get the model from that. */ if((nc = find_in_NCList_by_name(path)) != NULL) { int format = 0; /* Get the model from this NC */ if((status = nc_inq_format_extended(nc->ext_ncid,&format,NULL))) goto done; model->impl = format; if((status = nc_inq_format(nc->ext_ncid,&format))) goto done; model->format = format; goto done; } #endif magicinfo.path = path; /* do not free */ magicinfo.uri = uri; /* do not free */ magicinfo.omode = omode; magicinfo.model = model; /* do not free */ magicinfo.parameters = parameters; /* do not free */ #ifdef USE_STDIO magicinfo.use_parallel = 0; #else magicinfo.use_parallel = use_parallel; #endif if((status = openmagic(&magicinfo))) goto done; /* Verify we have a large enough file */ if(magicinfo.filelen < (unsigned long long)MAGIC_NUMBER_LEN) {status = NC_ENOTNC; goto done;} if((status = readmagic(&magicinfo,0L,magic)) != NC_NOERR) { status = NC_ENOTNC; goto done; } /* Look at the magic number */ if(NC_interpret_magic_number(magic,model) == NC_NOERR && model->format != 0) { if (use_parallel && (model->format == NC_FORMAT_NC3 || model->impl == NC_FORMATX_NC3)) /* this is called from nc_open_par() and file is classic */ model->impl = NC_FORMATX_PNETCDF; goto done; /* found something */ } /* Remaining case when implementation is an HDF5 file; search forward at starting at 512 and doubling to see if we have HDF5 magic number */ { long pos = 512L; for(;;) { if((pos+MAGIC_NUMBER_LEN) > magicinfo.filelen) {status = NC_ENOTNC; goto done;} if((status = readmagic(&magicinfo,pos,magic)) != NC_NOERR) {status = NC_ENOTNC; goto done; } NC_interpret_magic_number(magic,model); if(model->impl == NC_FORMATX_NC4) break; /* double and try again */ pos = 2*pos; } } done: closemagic(&magicinfo); return check(status); } /** \internal \ingroup datasets Provide open, read and close for use when searching for magic numbers */ static int openmagic(struct MagicFile* file) { int status = NC_NOERR; if(fIsSet(file->omode,NC_INMEMORY)) { /* Get its length */ NC_memio* meminfo = (NC_memio*)file->parameters; assert(meminfo != NULL); file->filelen = (long long)meminfo->size; #ifdef ENABLE_BYTERANGE } else if(file->uri != NULL) { #ifdef ENABLE_S3_SDK /* If this is an S3 URL, then handle specially */ if(NC_iss3(file->uri)) { if((status = NC_s3urlprocess(file->uri,&file->s3))) goto done; if((file->s3client = NC_s3sdkcreateclient(&file->s3))==NULL) {status = NC_EURL; goto done;} if((status = NC_s3sdkinfo(file->s3client,file->s3.bucket,file->s3.rootkey,&file->filelen,&file->errmsg))) goto done; file->iss3 = 1; } else #endif { /* Construct a URL minus any fragment */ file->curlurl = ncuribuild(file->uri,NULL,NULL,NCURISVC); /* Open the curl handle */ if((status=nc_http_init(&file->state))) goto done; if((status=nc_http_size(file->state,file->curlurl,&file->filelen))) goto done; } #endif /*BYTERANGE*/ } else { #ifdef USE_PARALLEL if (file->use_parallel) { int retval; MPI_Offset size; assert(file->parameters != NULL); if((retval = MPI_File_open(((NC_MPI_INFO*)file->parameters)->comm, (char*)file->path,MPI_MODE_RDONLY, ((NC_MPI_INFO*)file->parameters)->info, &file->fh)) != MPI_SUCCESS) { #ifdef MPI_ERR_NO_SUCH_FILE int errorclass; MPI_Error_class(retval, &errorclass); if (errorclass == MPI_ERR_NO_SUCH_FILE) #ifdef NC_ENOENT status = NC_ENOENT; #else status = errno; #endif else #endif status = NC_EPARINIT; file->fh = MPI_FILE_NULL; goto done; } /* Get its length */ if((retval=MPI_File_get_size(file->fh, &size)) != MPI_SUCCESS) {status = NC_EPARINIT; goto done;} file->filelen = (long long)size; } else #endif /* USE_PARALLEL */ { if (file->path == NULL || strlen(file->path) == 0) {status = NC_EINVAL; goto done;} file->fp = NCfopen(file->path, "r"); if(file->fp == NULL) {status = errno; goto done;} /* Get its length */ { int fd = fileno(file->fp); #ifdef _WIN32 __int64 len64 = _filelengthi64(fd); if(len64 < 0) {status = errno; goto done;} file->filelen = (long long)len64; #else off_t size; size = lseek(fd, 0, SEEK_END); if(size == -1) {status = errno; goto done;} file->filelen = (long long)size; #endif } rewind(file->fp); } } done: return check(status); } static int readmagic(struct MagicFile* file, long pos, char* magic) { int status = NC_NOERR; NCbytes* buf = ncbytesnew(); memset(magic,0,MAGIC_NUMBER_LEN); if(fIsSet(file->omode,NC_INMEMORY)) { char* mempos; NC_memio* meminfo = (NC_memio*)file->parameters; if((pos + MAGIC_NUMBER_LEN) > meminfo->size) {status = NC_EINMEMORY; goto done;} mempos = ((char*)meminfo->memory) + pos; memcpy((void*)magic,mempos,MAGIC_NUMBER_LEN); #ifdef DEBUG printmagic("XXX: readmagic",magic,file); #endif #ifdef ENABLE_BYTERANGE } else if(file->uri != NULL) { fileoffset_t start = (size_t)pos; fileoffset_t count = MAGIC_NUMBER_LEN; #ifdef ENABLE_S3_SDK if(file->iss3) { if((status = NC_s3sdkread(file->s3client,file->s3.bucket,file->s3.rootkey,start,count,(void*)magic,&file->errmsg))) {goto done;} } else #endif { status = nc_http_read(file->state, file->curlurl, start, count, buf); if (status == NC_NOERR) { if (ncbyteslength(buf) != count) status = NC_EINVAL; else memcpy(magic, ncbytescontents(buf), count); } } #endif } else { #ifdef USE_PARALLEL if (file->use_parallel) { MPI_Status mstatus; int retval; if((retval = MPI_File_read_at_all(file->fh, pos, magic, MAGIC_NUMBER_LEN, MPI_CHAR, &mstatus)) != MPI_SUCCESS) {status = NC_EPARINIT; goto done;} } else #endif /* USE_PARALLEL */ { /* Ordinary read */ long i; i = fseek(file->fp, pos, SEEK_SET); if (i < 0) { status = errno; goto done; } ncbytessetlength(buf, 0); if ((status = NC_readfileF(file->fp, buf, MAGIC_NUMBER_LEN))) goto done; memcpy(magic, ncbytescontents(buf), MAGIC_NUMBER_LEN); } } done: ncbytesfree(buf); if(file && file->fp) clearerr(file->fp); return check(status); } /** * Close the file opened to check for magic number. * * @param file pointer to the MagicFile struct for this open file. * @returns NC_NOERR for success * @returns NC_EPARINIT if there was a problem closing file with MPI * (parallel builds only). * @author Dennis Heimbigner */ static int closemagic(struct MagicFile* file) { int status = NC_NOERR; if(fIsSet(file->omode,NC_INMEMORY)) { /* noop */ #ifdef ENABLE_BYTERANGE } else if(file->uri != NULL) { #ifdef ENABLE_S3_SDK if(file->iss3) { NC_s3sdkclose(file->s3client, &file->s3, 0, &file->errmsg); NC_s3clear(&file->s3); nullfree(file->errmsg); } else #endif { status = nc_http_close(file->state); nullfree(file->curlurl); } #endif } else { #ifdef USE_PARALLEL if (file->use_parallel) { int retval; if(file->fh != MPI_FILE_NULL && (retval = MPI_File_close(&file->fh)) != MPI_SUCCESS) {status = NC_EPARINIT; return status;} } else #endif { if(file->fp) fclose(file->fp); } } return status; } /*! Interpret the magic number found in the header of a netCDF file. This function interprets the magic number/string contained in the header of a netCDF file and sets the appropriate NC_FORMATX flags. @param[in] magic Pointer to a character array with the magic number block. @param[out] model Pointer to an integer to hold the corresponding netCDF type. @param[out] version Pointer to an integer to hold the corresponding netCDF version. @returns NC_NOERR if a legitimate file type found @returns NC_ENOTNC otherwise \internal \ingroup datasets */ static int NC_interpret_magic_number(char* magic, NCmodel* model) { int status = NC_NOERR; int tmpimpl = 0; /* Look at the magic number */ if(model->impl == NC_FORMATX_UDF0 || model->impl == NC_FORMATX_UDF1) tmpimpl = model->impl; /* Use the complete magic number string for HDF5 */ if(memcmp(magic,HDF5_SIGNATURE,sizeof(HDF5_SIGNATURE))==0) { model->impl = NC_FORMATX_NC4; model->format = NC_FORMAT_NETCDF4; goto done; } if(magic[0] == '\016' && magic[1] == '\003' && magic[2] == '\023' && magic[3] == '\001') { model->impl = NC_FORMATX_NC_HDF4; model->format = NC_FORMAT_NETCDF4; goto done; } if(magic[0] == 'C' && magic[1] == 'D' && magic[2] == 'F') { if(magic[3] == '\001') { model->impl = NC_FORMATX_NC3; model->format = NC_FORMAT_CLASSIC; goto done; } if(magic[3] == '\002') { model->impl = NC_FORMATX_NC3; model->format = NC_FORMAT_64BIT_OFFSET; goto done; } if(magic[3] == '\005') { model->impl = NC_FORMATX_NC3; model->format = NC_FORMAT_64BIT_DATA; goto done; } } /* No match */ if (!tmpimpl) status = NC_ENOTNC; goto done; done: /* if model->impl was UDF0 or UDF1 on entry, make it so on exit */ if(tmpimpl) model->impl = tmpimpl; /* if this is a UDF magic_number update the model->impl */ if (strlen(UDF0_magic_number) && !strncmp(UDF0_magic_number, magic, strlen(UDF0_magic_number))) { model->impl = NC_FORMATX_UDF0; status = NC_NOERR; } if (strlen(UDF1_magic_number) && !strncmp(UDF1_magic_number, magic, strlen(UDF1_magic_number))) { model->impl = NC_FORMATX_UDF1; status = NC_NOERR; } return check(status); } #ifdef DEBUG static void printmagic(const char* tag, char* magic, struct MagicFile* f) { int i; fprintf(stderr,"%s: ispar=%d magic=",tag,f->use_parallel); for(i=0;i