netcdf-c/libdap2/dceconstraints.c

1178 lines
30 KiB
C
Raw Normal View History

2011-04-18 02:56:10 +08:00
/*********************************************************************
2018-12-07 05:21:03 +08:00
* Copyright 2018, UCAR/Unidata
2011-04-18 02:56:10 +08:00
* See netcdf/COPYRIGHT file for copying and redistribution conditions.
*********************************************************************/
#include "config.h"
#include <stddef.h>
2011-04-18 02:56:10 +08:00
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <assert.h>
Primary change: add dap4 support Specific changes: 1. Add dap4 code: libdap4 and dap4_test. Note that until the d4ts server problem is solved, dap4 is turned off. 2. Modify various files to support dap4 flags: configure.ac, Makefile.am, CMakeLists.txt, etc. 3. Add nc_test/test_common.sh. This centralizes the handling of the locations of various things in the build tree: e.g. where is ncgen.exe located. See nc_test/test_common.sh for details. 4. Modify .sh files to use test_common.sh 5. Obsolete separate oc2 by moving it to be part of netcdf-c. This means replacing code with netcdf-c equivalents. 5. Add --with-testserver to configure.ac to allow override of the servers to be used for --enable-dap-remote-tests. 6. There were multiple versions of nctypealignment code. Try to centralize in libdispatch/doffset.c and include/ncoffsets.h 7. Add a unit test for the ncuri code because of its complexity. 8. Move the findserver code out of libdispatch and into a separate, self contained program in ncdap_test and dap4_test. 9. Move the dispatch header files (nc{3,4}dispatch.h) to .../include because they are now shared by modules. 10. Revamp the handling of TOPSRCDIR and TOPBUILDDIR for shell scripts. 11. Make use of MREMAP if available 12. Misc. minor changes e.g. - #include <config.h> -> #include "config.h" - Add some no-install headers to /include - extern -> EXTERNL and vice versa as needed - misc header cleanup - clean up checking for misc. unix vs microsoft functions 13. Change copyright decls in some files to point to LICENSE file. 14. Add notes to RELEASENOTES.md
2017-03-09 08:01:10 +08:00
#include "dapincludes.h"
#include "dceparselex.h"
2011-04-18 02:56:10 +08:00
#undef DEBUG
2011-04-18 02:56:10 +08:00
#define LBRACE "{"
#define RBRACE "}"
static const char* opstrings[] = OPSTRINGS ;
2011-04-18 02:56:10 +08:00
static void ceallnodesr(DCEnode* node, NClist* allnodes, CEsort which);
static void dcedump(DCEnode* node, NCbytes* buf);
static void dcedumpraw(DCEnode* node, NCbytes* buf);
static void dcedumprawlist(NClist* list, NCbytes* buf);
2011-04-18 02:56:10 +08:00
#if 0 /*not currently used */
2011-04-18 02:56:10 +08:00
/* Parse incoming url constraints, if any,
to check for syntactic correctness
*/
int
dapparseconstraints(char* constraints, DCEconstraint* dapconstraint)
{
int ncstat = NC_NOERR;
char* errmsg;
assert(dapconstraint != NULL);
nclistclear(dapconstraint->projections);
nclistclear(dapconstraint->selections);
ncstat = dapceparse(constraints,dapconstraint,&errmsg);
if(ncstat) {
nclog(NCLOGWARN,"DAP constraint parse failure: %s",errmsg);
if(errmsg) free(errmsg);
nclistclear(dapconstraint->projections);
nclistclear(dapconstraint->selections);
}
#ifdef DEBUG
fprintf(stderr,"constraint: %s",dcetostring((DCEnode*)dapconstraint));
#endif
return ncstat;
}
#endif
2011-04-18 02:56:10 +08:00
2013-04-18 02:58:37 +08:00
#ifdef DEBUG1
static void
2013-04-24 04:18:16 +08:00
slicedump(const char* prefix, DCEslice* s)
2013-04-18 02:58:37 +08:00
{
#ifdef DCEVERBOSE
2013-04-24 04:18:16 +08:00
fprintf(stderr,"%s: %s\n",prefix,dcetostring((DCEnode*)s));
2013-04-18 02:58:37 +08:00
#else
size_t last = (s->first+s->length)-1;
fprintf(stderr,"%s: [%lu:%lu:%lu p=%lu l=%lu c=%lu]\n",
This PR adds EXPERIMENTAL support for accessing data in the cloud using a variant of the Zarr protocol and storage format. This enhancement is generically referred to as "NCZarr". The data model supported by NCZarr is netcdf-4 minus the user-defined types and the String type. In this sense it is similar to the CDF-5 data model. More detailed information about enabling and using NCZarr is described in the document NUG/nczarr.md and in a [Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in). WARNING: this code has had limited testing, so do use this version for production work. Also, performance improvements are ongoing. Note especially the following platform matrix of successful tests: Platform | Build System | S3 support ------------------------------------ Linux+gcc | Automake | yes Linux+gcc | CMake | yes Visual Studio | CMake | no Additionally, and as a consequence of the addition of NCZarr, major changes have been made to the Filter API. NOTE: NCZarr does not yet support filters, but these changes are enablers for that support in the future. Note that it is possible (probable?) that there will be some accidental reversions if the changes here did not correctly mimic the existing filter testing. In any case, previously filter ids and parameters were of type unsigned int. In order to support the more general zarr filter model, this was all converted to char*. The old HDF5-specific, unsigned int operations are still supported but they are wrappers around the new, char* based nc_filterx_XXX functions. This entailed at least the following changes: 1. Added the files libdispatch/dfilterx.c and include/ncfilter.h 2. Some filterx utilities have been moved to libdispatch/daux.c 3. A new entry, "filter_actions" was added to the NCDispatch table and the version bumped. 4. An overly complex set of structs was created to support funnelling all of the filterx operations thru a single dispatch "filter_actions" entry. 5. Move common code to from libhdf5 to libsrc4 so that it is accessible to nczarr. Changes directly related to Zarr: 1. Modified CMakeList.txt and configure.ac to support both C and C++ -- this is in support of S3 support via the awd-sdk libraries. 2. Define a size64_t type to support nczarr. 3. More reworking of libdispatch/dinfermodel.c to support zarr and to regularize the structure of the fragments section of a URL. Changes not directly related to Zarr: 1. Make client-side filter registration be conditional, with default off. 2. Hack include/nc4internal.h to make some flags added by Ed be unique: e.g. NC_CREAT, NC_INDEF, etc. 3. cleanup include/nchttp.h and libdispatch/dhttp.c. 4. Misc. changes to support compiling under Visual Studio including: * Better testing under windows for dirent.h and opendir and closedir. 5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags and to centralize error reporting. 6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them. 7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible. Changes Left TO-DO: 1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
prefix,s->first,s->stride,last,s->last,s->length,s->count);
2013-04-18 02:58:37 +08:00
#endif
}
#endif
2011-04-18 02:56:10 +08:00
2013-04-18 02:58:37 +08:00
/*
Compose slice s1 with slice s2 -> sr
Compose means that the s2 constraint is applied
to the output of the s1 constraint.
2013-04-18 02:58:37 +08:00
Logical derivation of s1 compose s2 = sr
We have three index sequences to deal with,
where the sequence is the range [0..last].
1. the original data indices [0..N]
2. the indices resulting from applying slice s1;
this will be a subset of #1 with the sequence
[s1.first .. s1.last] where s1.last = s1.first + s1.length - 1
3. the indices resulting from applying s2:[0..s2.last] wrt
to output of s2
[s2.first .. s2.last]
We can convert #3 into index sequence wrt
#1 as follows
[s2.first .. s2.last] -> [map(s2.first)..map(s2.last)]
where map(int index) = s1.first + (s1.stride * index)
Note that map(i) is undefined if map(i) > s1.last
So, we can compute the result (sr) stride and first as follows
. sr.stride = s1.stride * s2.stride
. sr.first = map(s2.first) = s1.first * (s1.stride * s2.first)
This throws an exception if sr.first > s1.last
. compute the s1.last wrt original data
last1 = s1.first + (s1.length - 1)
. compute the s2.last wrt s2
last2 = s2.first + (s2.length - 1)
. compute candidate last wrt original index sequence
lastx = map(last2) = s1.first + (s1.stride * last2)
2013-04-18 02:58:37 +08:00
. It is possible that lastx is outside of the range (s1.first..s1.last)
so we take the min of last2 and lastx to ensure no overrun wrt s1
sr.last = min(last1,lastx)
. If we want to be pedantic, we need to reduce
sr.last so that it is on an exact multiple of sr.stride
starting at sr.first
delta = ((sr.last + (sr.stride-1)) / sr.stride) * sr.stride
sr.last = sr.first + delta
. compute result length using sr.last
sr.len = (sr.last + 1) - sr.first
2013-04-18 02:58:37 +08:00
Example 1:
0000000000111111111
0123456789012345678
xxxxxxxxx
2013-04-18 02:58:37 +08:00
0 1 2 3 4 s1=(f=1 st=2 len=9)
0 1 2 3 4 s2=(f=0 st=1 len=5)
xxxxxxxxxy
0 1 3 4 5 sr=(f=1 st=2 len=9..10)
2013-04-18 02:58:37 +08:00
Example 2:
2011-04-18 02:56:10 +08:00
0000000000111111111122222222223
0123456789012345678901234567890
xxxxxxxxxxxxxxxxxxxxxxxxx
0 1 2 3 4 5 6 7 8 _ _ _ s1=(f=1 st=3 len=25)
0 1 2 3 4 s2=(f=3 st=2 len=5)
xxxxxxxxxxxxxyyyyy
0 1 2 sr=(f=10 st=6 l=13..17)
2013-04-18 02:58:37 +08:00
Example 3:
2011-04-18 02:56:10 +08:00
0000000000111111
0123456789012345
xxxxxxxxx
0 1 2 3 4 _ _ s1=(f=1 st=2 len=9)
0 1 2 3 4 s2=(f=2 st=1 len=4)
xxxxxy ----------------------------
sr=(f=5 st=2 len=5..6)
2013-04-18 02:58:37 +08:00
Example 4:
2011-04-18 02:56:10 +08:00
0000000000111111111
0123456789012345678
xxxxxxxxx
0 1 2 3 4 _ _ _ _ s1=(f=1 st=2 len=9)
0 1 2 3 s2=(f=2 st=2 len=4)
xxxxxxyy
sr=(f=5 st=4 len=6..8)
2013-04-18 02:58:37 +08:00
Example 5:
00000000001
01234567890
xxx
2013-04-18 02:58:37 +08:00
012 s1=(f=0 st=1 l=3)
012 s2=(f=0 st=1 l=3)
xxx ----------------------------
012 sr=(f=0 st=1 l=3)
2011-04-18 02:56:10 +08:00
Example 6:
2011-04-18 02:56:10 +08:00
00000
01234
xx
2013-04-18 02:58:37 +08:00
01 s1=(f=0 st=1 l=2)
0 s2=(f=0 st=1 l=1)
x ----------------------------
2013-04-18 02:58:37 +08:00
sr=(f=0 st=1 l=1)
2011-04-18 02:56:10 +08:00
*/
#define MAP(s1,i) ((s1)->first + ((s1)->stride*(i)))
#define XMIN(x,y) ((x) < (y) ? (x) : (y))
#define XMAX(x,y) ((x) > (y) ? (x) : (y))
2011-04-18 02:56:10 +08:00
int
dceslicecompose(DCEslice* s1, DCEslice* s2, DCEslice* result)
2011-04-18 02:56:10 +08:00
{
int err = NC_NOERR;
size_t lastx = 0;
DCEslice sr; /* For back compatibility so s1 and result can be same object */
2013-04-18 02:58:37 +08:00
#ifdef DEBUG1
2013-04-24 04:18:16 +08:00
slicedump("compose: s1",s1);
slicedump("compose: s2",s2);
2013-04-18 02:58:37 +08:00
#endif
sr.node.sort = CES_SLICE;
sr.stride = s1->stride * s2->stride;
sr.first = MAP(s1,s2->first);
if(sr.first > s1->last)
return NC_EINVALCOORDS;
lastx = MAP(s1,s2->last);
sr.last = XMIN(s1->last,lastx);
sr.length = (sr.last + 1) - sr.first;
sr.declsize = XMAX(s1->declsize,s2->declsize); /* use max declsize */
2013-04-18 02:58:37 +08:00
/* fill in other fields */
sr.count = (sr.length + (sr.stride - 1))/sr.stride;
2013-04-24 04:18:16 +08:00
*result = sr;
2013-04-18 02:58:37 +08:00
#ifdef DEBUG1
2013-04-24 04:18:16 +08:00
slicedump("compose: result",result);
2013-04-18 02:58:37 +08:00
#endif
2011-04-18 02:56:10 +08:00
return err;
}
/*
Given two projection lists, merge
src into dst taking
overlapping projections into acct.
Dst will be modified.
2011-04-18 02:56:10 +08:00
*/
2011-04-18 02:56:10 +08:00
int
2011-11-14 12:20:19 +08:00
dcemergeprojectionlists(NClist* dst, NClist* src)
2011-04-18 02:56:10 +08:00
{
size_t i;
2011-04-18 02:56:10 +08:00
NClist* cat = nclistnew();
int ncstat = NC_NOERR;
#ifdef DEBUG
fprintf(stderr,"dapmergeprojection: dst = %s\n",dcetostring((DCEnode*)dst));
fprintf(stderr,"dapmergeprojection: src = %s\n",dcetostring((DCEnode*)src));
#endif
/* get dst concat clone(src) */
nclistsetalloc(cat,nclistlength(dst)+nclistlength(src));
for(i=0;i<nclistlength(dst);i++) {
DCEprojection* p = (DCEprojection*)nclistget(dst,i);
nclistpush(cat,(void*)p);
2011-04-18 02:56:10 +08:00
}
for(i=0;i<nclistlength(src);i++) {
DCEprojection* p = (DCEprojection*)nclistget(src,i);
nclistpush(cat,(void*)dceclone((DCEnode*)p));
2011-04-18 02:56:10 +08:00
}
nclistclear(dst);
/* Repeatedly pull elements from the concat,
merge with all duplicates, and stick into
the dst
*/
while(nclistlength(cat) > 0) {
DCEprojection* target = (DCEprojection*)nclistremove(cat,0);
if(target == NULL) continue;
if(target->discrim != CES_VAR) continue;
for(i=0;i<nclistlength(cat);i++) {
DCEprojection* p2 = (DCEprojection*)nclistget(cat,i);
if(p2 == NULL) continue;
if(p2->discrim != CES_VAR) continue;
if(dcesamepath(target->var->segments,
p2->var->segments)!=0) continue;
/* This entry matches our current target; merge */
2011-11-14 12:20:19 +08:00
ncstat = dcemergeprojections(target,p2);
2011-04-18 02:56:10 +08:00
/* null out this merged entry and release it */
nclistset(cat,i,(void*)NULL);
2011-04-18 02:56:10 +08:00
dcefree((DCEnode*)p2);
}
/* Capture the clone */
nclistpush(dst,(void*)target);
2011-04-18 02:56:10 +08:00
}
nclistfree(cat);
return ncstat;
}
2012-02-04 05:31:50 +08:00
/* Modify merged projection to include "addition" projection */
2011-11-14 12:20:19 +08:00
int
2012-02-04 05:31:50 +08:00
dcemergeprojections(DCEprojection* merged, DCEprojection* addition)
2011-04-18 02:56:10 +08:00
{
int ncstat = NC_NOERR;
size_t i,j;
2011-04-18 02:56:10 +08:00
2012-02-04 05:31:50 +08:00
ASSERT((merged->discrim == CES_VAR && addition->discrim == CES_VAR));
ASSERT((nclistlength(merged->var->segments) == nclistlength(addition->var->segments)));
for(i=0;i<nclistlength(merged->var->segments);i++) {
DCEsegment* mergedseg = (DCEsegment*)nclistget(merged->var->segments,i);
DCEsegment* addedseg = (DCEsegment*)nclistget(addition->var->segments,i);
/* If one segment has larger rank, then copy the extra slices unchanged */
for(j=0;j<addedseg->rank;j++) {
if(j < mergedseg->rank)
dceslicecompose(mergedseg->slices+j,addedseg->slices+j,mergedseg->slices+j);
2012-02-04 05:31:50 +08:00
else
mergedseg->slices[j] = addedseg->slices[j];
2011-04-18 02:56:10 +08:00
}
2012-02-04 05:31:50 +08:00
if(addedseg->rank > mergedseg->rank)
mergedseg->rank = addedseg->rank;
2011-04-18 02:56:10 +08:00
}
return ncstat;
}
2011-11-14 12:20:19 +08:00
/* Convert a DCEprojection instance into a string
2011-04-18 02:56:10 +08:00
that can be used with the url
*/
char*
dcebuildprojectionstring(NClist* projections)
2011-04-18 02:56:10 +08:00
{
char* pstring;
NCbytes* buf = ncbytesnew();
dcelisttobuffer(projections,buf,",");
pstring = ncbytesdup(buf);
ncbytesfree(buf);
return pstring;
}
char*
dcebuildselectionstring(NClist* selections)
2011-04-18 02:56:10 +08:00
{
NCbytes* buf = ncbytesnew();
char* sstring;
dcelisttobuffer(selections,buf,",");
sstring = ncbytesdup(buf);
ncbytesfree(buf);
return sstring;
}
char*
dcebuildconstraintstring(DCEconstraint* constraints)
2011-04-18 02:56:10 +08:00
{
NCbytes* buf = ncbytesnew();
char* result = NULL;
dcetobuffer((DCEnode*)constraints,buf);
result = ncbytesdup(buf);
ncbytesfree(buf);
return result;
}
DCEnode*
dceclone(DCEnode* node)
{
DCEnode* result = NULL;
result = (DCEnode*)dcecreate(node->sort);
if(result == NULL) goto done;
switch (node->sort) {
case CES_SLICE: {
DCEslice* clone = (DCEslice*)result;
DCEslice* orig = (DCEslice*)node;
*clone = *orig;
} break;
case CES_SEGMENT: {
DCEsegment* clone = (DCEsegment*)result;
DCEsegment* orig = (DCEsegment*)node;
*clone = *orig;
clone->name = nulldup(orig->name);
if(orig->rank > 0)
memcpy(clone->slices,orig->slices,orig->rank*sizeof(DCEslice));
} break;
case CES_VAR: {
DCEvar* clone = (DCEvar*)result;
DCEvar* orig = (DCEvar*)node;
*clone = *orig;
clone->segments = dceclonelist(clone->segments);
} break;
case CES_FCN: {
DCEfcn* clone = (DCEfcn*)result;
DCEfcn* orig = (DCEfcn*)node;
*clone = *orig;
clone->name = nulldup(orig->name);
clone->args = dceclonelist(orig->args);
} break;
case CES_CONST: {
DCEconstant* clone = (DCEconstant*)result;
DCEconstant* orig = (DCEconstant*)node;
*clone = *orig;
if(clone->discrim == CES_STR)
clone->text = nulldup(clone->text);
} break;
case CES_VALUE: {
DCEvalue* clone = (DCEvalue*)result;
DCEvalue* orig = (DCEvalue*)node;
*clone = *orig;
switch (clone->discrim) {
case CES_CONST:
clone->constant = (DCEconstant*)dceclone((DCEnode*)orig->constant); break;
case CES_VAR:
clone->var = (DCEvar*)dceclone((DCEnode*)orig->var); break;
case CES_FCN:
clone->fcn = (DCEfcn*)dceclone((DCEnode*)orig->fcn); break;
default: assert(0);
}
} break;
case CES_PROJECT: {
DCEprojection* clone = (DCEprojection*)result;
DCEprojection* orig = (DCEprojection*)node;
*clone = *orig;
switch (orig->discrim) {
case CES_VAR:
clone->var = (DCEvar*)dceclone((DCEnode*)orig->var); break;
case CES_FCN:
clone->fcn = (DCEfcn*)dceclone((DCEnode*)orig->fcn); break;
default: assert(0);
}
} break;
case CES_SELECT: {
DCEselection* clone = (DCEselection*)result;
DCEselection* orig = (DCEselection*)node;
*clone = *orig;
clone->lhs = (DCEvalue*)dceclone((DCEnode*)orig->lhs);
clone->rhs = dceclonelist(orig->rhs);
} break;
case CES_CONSTRAINT: {
DCEconstraint* clone = (DCEconstraint*)result;
DCEconstraint* orig = (DCEconstraint*)node;
*clone = *orig;
clone->projections = dceclonelist(orig->projections);
clone->selections = dceclonelist(orig->selections);
} break;
default:
assert(0);
}
done:
return result;
}
NClist*
dceclonelist(NClist* list)
{
size_t i;
2011-04-18 02:56:10 +08:00
NClist* clone;
if(list == NULL) return NULL;
clone = nclistnew();
for(i=0;i<nclistlength(list);i++) {
DCEnode* node = (DCEnode*)nclistget(list,i);
DCEnode* newnode = dceclone((DCEnode*)node);
nclistpush(clone,(void*)newnode);
2011-04-18 02:56:10 +08:00
}
return clone;
}
void
dcefree(DCEnode* node)
{
if(node == NULL) return;
switch (node->sort) {
case CES_VAR: {
DCEvar* target = (DCEvar*)node;
dcefreelist(target->segments);
} break;
case CES_FCN: {
DCEfcn* target = (DCEfcn*)node;
dcefreelist(target->args);
nullfree(target->name);
} break;
case CES_CONST: {
DCEconstant* target = (DCEconstant*)node;
if(target->discrim == CES_STR)
nullfree(target->text);
} break;
case CES_VALUE: {
DCEvalue* target = (DCEvalue*)node;
switch(target->discrim) {
case CES_CONST: dcefree((DCEnode*)target->constant); break;
case CES_VAR: dcefree((DCEnode*)target->var); break;
case CES_FCN: dcefree((DCEnode*)target->fcn); break;
default: assert(0);
}
} break;
case CES_PROJECT: {
DCEprojection* target = (DCEprojection*)node;
switch (target->discrim) {
case CES_VAR: dcefree((DCEnode*)target->var); break;
case CES_FCN: dcefree((DCEnode*)target->fcn); break;
default: assert(0);
}
} break;
case CES_SELECT: {
DCEselection* target = (DCEselection*)node;
dcefreelist(target->rhs);
dcefree((DCEnode*)target->lhs);
} break;
case CES_CONSTRAINT: {
DCEconstraint* target = (DCEconstraint*)node;
dcefreelist(target->projections);
dcefreelist(target->selections);
} break;
case CES_SEGMENT: {
DCEsegment* target = (DCEsegment*)node;
target->rank = 0;
nullfree(target->name);
} break;
case CES_SLICE: {
} break;
default:
assert(0);
}
/* final action */
free(node);
}
void
dcefreelist(NClist* list)
{
size_t i;
2011-04-18 02:56:10 +08:00
if(list == NULL) return;
for(i=0;i<nclistlength(list);i++) {
DCEnode* node = (DCEnode*)nclistget(list,i);
dcefree((DCEnode*)node);
}
nclistfree(list);
}
char*
dcetostring(DCEnode* node)
{
char* s;
NCbytes* buf = ncbytesnew();
dcetobuffer(node,buf);
s = ncbytesextract(buf);
ncbytesfree(buf);
return s;
}
/* Mostly for debugging */
char*
dcerawtostring(void* node)
{
char* s;
NCbytes* buf = ncbytesnew();
dcedumpraw((DCEnode*)node,buf);
s = ncbytesextract(buf);
ncbytesfree(buf);
return s;
}
char*
dcerawlisttostring(NClist* list)
{
char* s;
NCbytes* buf = ncbytesnew();
dcedumprawlist(list,buf);
s = ncbytesextract(buf);
ncbytesfree(buf);
return s;
}
2011-04-18 02:56:10 +08:00
void
dcetobuffer(DCEnode* node, NCbytes* buf)
{
dcedump(node,buf);
}
static void
dcedump(DCEnode* node, NCbytes* buf)
2011-04-18 02:56:10 +08:00
{
char tmp[1024];
if(buf == NULL) return;
2011-11-14 12:20:19 +08:00
if(node == NULL) {ncbytescat(buf,"<null>"); return;}
2011-04-18 02:56:10 +08:00
switch (node->sort) {
case CES_SLICE: {
DCEslice* slice = (DCEslice*)node;
size_t last = (slice->first+slice->length)-1;
if(slice->count == 1) {
snprintf(tmp,sizeof(tmp),"[%lu]",
(unsigned long)slice->first);
2011-04-18 02:56:10 +08:00
} else if(slice->stride == 1) {
snprintf(tmp,sizeof(tmp),"[%lu:%lu]",
2011-04-18 02:56:10 +08:00
(unsigned long)slice->first,
(unsigned long)last);
2011-04-18 02:56:10 +08:00
} else {
snprintf(tmp,sizeof(tmp),"[%lu:%lu:%lu]",
2011-04-18 02:56:10 +08:00
(unsigned long)slice->first,
(unsigned long)slice->stride,
(unsigned long)last);
2011-04-18 02:56:10 +08:00
}
ncbytescat(buf,tmp);
} break;
case CES_SEGMENT: {
DCEsegment* segment = (DCEsegment*)node;
char* name = (segment->name?segment->name:"<unknown>");
int rank = segment->rank;
int i;
name = nulldup(name);
ncbytescat(buf,name);
nullfree(name);
if(!dceiswholesegment(segment)) {
2011-11-14 12:20:19 +08:00
for(i=0;i<rank;i++) {
2011-04-18 02:56:10 +08:00
DCEslice* slice = segment->slices+i;
dcetobuffer((DCEnode*)slice,buf);
}
}
} break;
case CES_VAR: {
DCEvar* var = (DCEvar*)node;
dcelisttobuffer(var->segments,buf,".");
} break;
case CES_FCN: {
DCEfcn* fcn = (DCEfcn*)node;
ncbytescat(buf,fcn->name);
ncbytescat(buf,"(");
dcelisttobuffer(fcn->args,buf,",");
ncbytescat(buf,")");
} break;
case CES_CONST: {
DCEconstant* value = (DCEconstant*)node;
switch (value->discrim) {
case CES_STR:
ncbytescat(buf,value->text);
break;
case CES_INT:
snprintf(tmp,sizeof(tmp),"%lld",value->intvalue);
ncbytescat(buf,tmp);
break;
case CES_FLOAT:
snprintf(tmp,sizeof(tmp),"%g",value->floatvalue);
ncbytescat(buf,tmp);
break;
default: assert(0);
}
} break;
case CES_VALUE: {
DCEvalue* value = (DCEvalue*)node;
switch (value->discrim) {
case CES_CONST:
dcetobuffer((DCEnode*)value->constant,buf);
break;
case CES_VAR:
dcetobuffer((DCEnode*)value->var,buf);
break;
case CES_FCN:
dcetobuffer((DCEnode*)value->fcn,buf);
break;
default: assert(0);
}
} break;
case CES_PROJECT: {
DCEprojection* target = (DCEprojection*)node;
switch (target->discrim) {
case CES_VAR:
dcetobuffer((DCEnode*)target->var,buf);
break;
case CES_FCN: dcetobuffer((DCEnode*)target->fcn,buf); break;
default: assert(0);
}
} break;
case CES_SELECT: {
DCEselection* sel = (DCEselection*)node;
dcetobuffer((DCEnode*)sel->lhs,buf);
if(sel->operator == CES_NIL) break;
ncbytescat(buf,opstrings[(int)sel->operator]);
if(nclistlength(sel->rhs) > 1)
ncbytescat(buf,"{");
dcelisttobuffer(sel->rhs,buf,",");
if(nclistlength(sel->rhs) > 1)
ncbytescat(buf,"}");
} break;
case CES_CONSTRAINT: {
DCEconstraint* con = (DCEconstraint*)node;
if(con->projections != NULL && nclistlength(con->projections) > 0) {
dcelisttobuffer(con->projections,buf,",");
}
if(con->selections != NULL && nclistlength(con->selections) > 0) {
ncbytescat(buf,"&"); /* because & is really a prefix */
dcelisttobuffer(con->selections,buf,"&");
}
} break;
case CES_NIL: {
ncbytescat(buf,"<nil>");
} break;
default:
assert(0);
}
}
char*
dcelisttostring(NClist* list, char* sep)
{
char* s;
NCbytes* buf = ncbytesnew();
dcelisttobuffer(list,buf,sep);
s = ncbytesextract(buf);
ncbytesfree(buf);
return s;
}
void
dcelisttobuffer(NClist* list, NCbytes* buf, char* sep)
{
size_t i;
2011-04-18 02:56:10 +08:00
if(list == NULL || buf == NULL) return;
if(sep == NULL) sep = ",";
for(i=0;i<nclistlength(list);i++) {
DCEnode* node = (DCEnode*)nclistget(list,i);
2011-11-14 12:20:19 +08:00
if(node == NULL) continue;
2011-04-18 02:56:10 +08:00
if(i>0) ncbytescat(buf,sep);
dcetobuffer((DCEnode*)node,buf);
}
}
/* Collect all nodes within a specified constraint tree */
/* Caller frees result */
NClist*
dceallnodes(DCEnode* node, CEsort which)
{
NClist* allnodes = nclistnew();
ceallnodesr(node,allnodes,which);
return allnodes;
}
static void
ceallnodesr(DCEnode* node, NClist* allnodes, CEsort which)
{
size_t i;
2011-04-18 02:56:10 +08:00
if(node == NULL) return;
if(nclistcontains(allnodes,(void*)node)) return;
2011-04-18 02:56:10 +08:00
if(which == CES_NIL || node->sort == which)
nclistpush(allnodes,(void*)node);
2011-04-18 02:56:10 +08:00
switch(node->sort) {
case CES_FCN: {
DCEfcn* fcn = (DCEfcn*)node;
for(i=0;i<nclistlength(fcn->args);i++) {
ceallnodesr((DCEnode*)nclistget(fcn->args,i),allnodes,which);
}
} break;
case CES_VAR: {
DCEvar* var = (DCEvar*)node;
for(i=0;i<nclistlength(var->segments);i++) {
ceallnodesr((DCEnode*)nclistget(var->segments,i),allnodes,which);
}
} break;
case CES_VALUE: {
DCEvalue* value = (DCEvalue*)node;
if(value->discrim == CES_VAR)
ceallnodesr((DCEnode*)value->var,allnodes,which);
else if(value->discrim == CES_FCN)
ceallnodesr((DCEnode*)value->fcn,allnodes,which);
else
ceallnodesr((DCEnode*)value->constant,allnodes,which);
} break;
case CES_SELECT: {
DCEselection* selection = (DCEselection*)node;
ceallnodesr((DCEnode*)selection->lhs,allnodes,which);
for(i=0;i<nclistlength(selection->rhs);i++)
ceallnodesr((DCEnode*)nclistget(selection->rhs,i),allnodes,which);
} break;
case CES_PROJECT: {
DCEprojection* projection = (DCEprojection*)node;
if(projection->discrim == CES_VAR)
ceallnodesr((DCEnode*)projection->var,allnodes,which);
else
ceallnodesr((DCEnode*)projection->fcn,allnodes,which);
} break;
case CES_CONSTRAINT: {
DCEconstraint* constraint = (DCEconstraint*)node;
for(i=0;i<nclistlength(constraint->projections);i++)
ceallnodesr((DCEnode*)nclistget(constraint->projections,i),allnodes,which);
for(i=0;i<nclistlength(constraint->selections);i++)
ceallnodesr((DCEnode*)nclistget(constraint->selections,i),allnodes,which);
} break;
/* All others have no subnodes */
default:
break;
}
}
DCEnode*
dcecreate(CEsort sort)
{
DCEnode* node = NULL;
switch (sort) {
case CES_SLICE: {
DCEslice* target = (DCEslice*)calloc(1,sizeof(DCEslice));
if(target == NULL) return NULL;
node = (DCEnode*)target;
} break;
case CES_SEGMENT: {
int i;
DCEsegment* target = (DCEsegment*)calloc(1,sizeof(DCEsegment));
if(target == NULL) return NULL;
/* Initialize the sort of the slices */
for(i=0;i<NC_MAX_VAR_DIMS;i++)
target->slices[i].node.sort = CES_SLICE;
node = (DCEnode*)target;
} break;
case CES_CONST: {
DCEconstant* target = (DCEconstant*)calloc(1,sizeof(DCEconstant));
if(target == NULL) return NULL;
node = (DCEnode*)target;
target->discrim = CES_NIL;
2011-04-18 02:56:10 +08:00
} break;
case CES_VALUE: {
DCEvalue* target = (DCEvalue*)calloc(1,sizeof(DCEvalue));
if(target == NULL) return NULL;
node = (DCEnode*)target;
target->discrim = CES_NIL;
2011-04-18 02:56:10 +08:00
} break;
case CES_VAR: {
DCEvar* target = (DCEvar*)calloc(1,sizeof(DCEvar));
if(target == NULL) return NULL;
node = (DCEnode*)target;
} break;
case CES_FCN: {
DCEfcn* target = (DCEfcn*)calloc(1,sizeof(DCEfcn));
if(target == NULL) return NULL;
node = (DCEnode*)target;
} break;
case CES_PROJECT: {
DCEprojection* target = (DCEprojection*)calloc(1,sizeof(DCEprojection));
if(target == NULL) return NULL;
node = (DCEnode*)target;
} break;
case CES_SELECT: {
DCEselection* target = (DCEselection*)calloc(1,sizeof(DCEselection));
if(target == NULL) return NULL;
node = (DCEnode*)target;
target->operator = CEO_NIL;
} break;
case CES_CONSTRAINT: {
DCEconstraint* target = (DCEconstraint*)calloc(1,sizeof(DCEconstraint));
if(target == NULL) return NULL;
node = (DCEnode*)target;
} break;
default:
assert(0);
}
/* final action */
node->sort = sort;
return node;
}
int
dceiswholeslice(DCEslice* slice)
{
if(slice->first != 0
|| slice->stride != 1
|| slice->length != slice->declsize) return 0;
2011-04-18 02:56:10 +08:00
return 1;
}
int
dceiswholesegment(DCEsegment* seg)
{
int i,whole;
2011-04-29 01:11:21 +08:00
2011-04-18 02:56:10 +08:00
if(!seg->slicesdefined) return 0; /* actually, we don't know */
whole = 1; /* assume so */
for(i=0;i<seg->rank;i++) {
if(!dceiswholeslice(&seg->slices[i])) {whole = 0; break;}
}
return whole;
}
2012-02-04 05:31:50 +08:00
void
dcemakewholeslice(DCEslice* slice, size_t declsize)
{
slice->first = 0;
slice->stride = 1;
slice->length = declsize;
slice->declsize = declsize;
slice->count = declsize;
slice->last = slice->length - 1;
2012-02-04 05:31:50 +08:00
}
/* Remove slicing from terminal segment of p */
void
dcemakewholeprojection(DCEprojection* p)
{
/* Remove the slicing (if any) from the last segment */
if(p->discrim == CES_VAR && p->var != NULL && p->var->segments != NULL) {
int lastindex = nclistlength(p->var->segments) - 1;
DCEsegment* lastseg = (DCEsegment*)nclistget(p->var->segments,lastindex);
lastseg->rank = 0;
}
}
2011-04-18 02:56:10 +08:00
int
dcesamepath(NClist* list1, NClist* list2)
{
size_t i;
size_t len = nclistlength(list1);
2011-04-18 02:56:10 +08:00
if(len != nclistlength(list2)) return 0;
for(i=0;i<len;i++) {
DCEsegment* s1 = (DCEsegment*)nclistget(list1,i);
DCEsegment* s2 = (DCEsegment*)nclistget(list2,i);
if(strcmp(s1->name,s2->name) != 0) return 0;
}
return 1;
}
void
dcesegment_transpose(DCEsegment* segment,
size_t* start,
size_t* count,
size_t* stride,
size_t* sizes
)
{
int i;
if(segment != NULL && sizes != NULL) {
for(i=0;i<segment->rank;i++) {
if(start != NULL) start[i] = segment->slices[i].first;
if(count != NULL) count[i] = segment->slices[i].count;
if(stride != NULL) stride[i] = (size_t)segment->slices[i].stride;
if(sizes != NULL) sizes[i] = segment->slices[i].declsize;
}
}
}
/* Compute segment size for subset of slices */
size_t
dcesegmentsize(DCEsegment* seg, size_t start, size_t stop)
{
int i, count;
if(!seg->slicesdefined) return 0; /* actually, we don't know */
for(count=1,i=start;i<stop;i++) {
count *= seg->slices[i].count;
}
return count;
}
/* Return the index of the leftmost slice
starting at start and up to, but not including
stop, such that it and all slices to the right
are "safe". Safe means dceiswholeslice() is true.
In effect, we can read the safe index set as a
single chunk. Return stop if there is no safe index.
*/
size_t
dcesafeindex(DCEsegment* seg, size_t start, size_t stop)
{
size_t safe;
if(!seg->slicesdefined) return stop; /* actually, we don't know */
if(stop == 0) return stop;
/* watch out because safe is unsigned */
for(safe=stop-1;safe>start;safe--) {
if(!dceiswholeslice(&seg->slices[safe])) return safe+1;
}
return dceiswholeslice(&seg->slices[start]) ? start /*every slice is safe*/
: start+1 ;
}
static const char*
dcesortname(CEsort sort)
{
switch (sort) {
case CES_SLICE: return "SLICE";
case CES_SEGMENT: return "SEGMENT";
case CES_VAR: return "VAR";
case CES_FCN: return "FCN";
case CES_CONST: return "CONST";
case CES_VALUE: return "VALUE";
case CES_PROJECT: return "PROJECT";
case CES_SELECT: return "SELECT";
case CES_CONSTRAINT: return "CONSTRAINT";
case CES_STR: return "STR";
case CES_INT: return "INT";
case CES_FLOAT: return "FLOAT";
default: break;
}
return "UNKNOWN";
}
static void
dcedumpraw(DCEnode* node, NCbytes* buf)
{
int i;
char tmp[1024];
if(buf == NULL) return;
if(node == NULL) {ncbytescat(buf,"<null>"); return;}
ncbytescat(buf,LBRACE);
ncbytescat(buf,(char*)dcesortname(node->sort));
switch (node->sort) {
case CES_SLICE: {
DCEslice* slice = (DCEslice*)node;
snprintf(tmp,sizeof(tmp),
" [first=%lu stride=%lu last=%lu len=%lu count=%lu size=%lu]",
(unsigned long)slice->first,
(unsigned long)slice->stride,
(unsigned long)slice->last,
(unsigned long)slice->length,
(unsigned long)slice->count,
(unsigned long)slice->declsize);
ncbytescat(buf,tmp);
} break;
case CES_SEGMENT: {
DCEsegment* segment = (DCEsegment*)node;
int rank = segment->rank;
char* name = (segment->name?segment->name:"<unknown>");
ncbytescat(buf," name=");
ncbytescat(buf,name);
snprintf(tmp,sizeof(tmp)," rank=%lu",(unsigned long)rank);
ncbytescat(buf,tmp);
ncbytescat(buf," defined=");
ncbytescat(buf,(segment->slicesdefined?"1":"0"));
ncbytescat(buf," declized=");
ncbytescat(buf,(segment->slicesdeclized?"1":"0"));
if(rank > 0) {
ncbytescat(buf," slices=");
for(i=0;i<rank;i++) {
DCEslice* slice = segment->slices+i;
dcedumpraw((DCEnode*)slice,buf);
}
}
} break;
case CES_VAR: {
DCEvar* var = (DCEvar*)node;
ncbytescat(buf," segments=");
dcedumprawlist(var->segments,buf);
} break;
case CES_FCN: {
DCEfcn* fcn = (DCEfcn*)node;
ncbytescat(buf," name=");
ncbytescat(buf,fcn->name);
ncbytescat(buf,"args=");
dcedumprawlist(fcn->args,buf);
} break;
case CES_CONST: {
DCEconstant* value = (DCEconstant*)node;
ncbytescat(buf," discrim=");
ncbytescat(buf,dcesortname(value->discrim));
ncbytescat(buf," value=");
switch (value->discrim) {
case CES_STR:
ncbytescat(buf,"|");
ncbytescat(buf,value->text);
ncbytescat(buf,"|");
break;
case CES_INT:
snprintf(tmp,sizeof(tmp),"%lld",value->intvalue);
ncbytescat(buf,tmp);
break;
case CES_FLOAT:
snprintf(tmp,sizeof(tmp),"%g",value->floatvalue);
ncbytescat(buf,tmp);
break;
default: assert(0);
}
} break;
case CES_VALUE: {
DCEvalue* value = (DCEvalue*)node;
ncbytescat(buf," discrim=");
ncbytescat(buf,dcesortname(value->discrim));
switch (value->discrim) {
case CES_CONST:
dcedumpraw((DCEnode*)value->constant,buf);
break;
case CES_VAR:
dcedumpraw((DCEnode*)value->var,buf);
break;
case CES_FCN:
dcedumpraw((DCEnode*)value->fcn,buf);
break;
default: assert(0);
}
} break;
case CES_PROJECT: {
DCEprojection* target = (DCEprojection*)node;
ncbytescat(buf," discrim=");
ncbytescat(buf,dcesortname(target->discrim));
switch (target->discrim) {
case CES_VAR:
dcedumpraw((DCEnode*)target->var,buf);
break;
case CES_FCN:
dcedumpraw((DCEnode*)target->fcn,buf);
break;
default: assert(0);
}
} break;
case CES_SELECT: {
DCEselection* sel = (DCEselection*)node;
ncbytescat(buf," ");
dcedumpraw((DCEnode*)sel->lhs,buf);
if(sel->operator == CES_NIL) break;
ncbytescat(buf,opstrings[(int)sel->operator]);
if(nclistlength(sel->rhs) > 1)
ncbytescat(buf,"{");
dcedumprawlist(sel->rhs,buf);
if(nclistlength(sel->rhs) > 1)
ncbytescat(buf,"}");
} break;
case CES_CONSTRAINT: {
DCEconstraint* con = (DCEconstraint*)node;
if(con->projections != NULL && nclistlength(con->projections) > 0) {
ncbytescat(buf,"projections=");
dcedumprawlist(con->projections,buf);
}
if(con->selections != NULL && nclistlength(con->selections) > 0) {
ncbytescat(buf,"selections=");
dcedumprawlist(con->selections,buf);
}
} break;
case CES_NIL: {
ncbytescat(buf,"<nil>");
} break;
default:
assert(0);
}
ncbytescat(buf,RBRACE);
}
static void
dcedumprawlist(NClist* list, NCbytes* buf)
{
size_t i;
if(list == NULL || buf == NULL) return;
ncbytescat(buf,"(");
for(i=0;i<nclistlength(list);i++) {
DCEnode* node = (DCEnode*)nclistget(list,i);
if(node == NULL) continue;
if(i>0) ncbytescat(buf,",");
dcedumpraw((DCEnode*)node,buf);
}
ncbytescat(buf,")");
}
2013-04-18 02:58:37 +08:00