Increased rigor in comparisons between object labels in JSON.

FossilOrigin-Name: 2bc86d145fccc07107b7753cb1a69122676d4096fe59c454497bd81a6142d45e
This commit is contained in:
drh 2023-12-06 14:50:48 +00:00
parent 88208296c2
commit 91ec00c25a
3 changed files with 228 additions and 101 deletions

View File

@ -1,5 +1,5 @@
C Rework\sthe\sJSON\sfunctions\sso\sthat\sthey\suse\sthe\sJSONB\sformat\sinternally.\nThe\soriginal\sJsonNode\sparse\stree\sdesign\sis\sremoved.\s\sAll\sJSON\sfunctions\nthat\saccept\stext\sJSON\salso\saccept\sJSONB.\s\sNew\sfunctions\sgenerate\sJSONB.
D 2023-12-05T19:45:09.048
C Increased\srigor\sin\scomparisons\sbetween\sobject\slabels\sin\sJSON.
D 2023-12-06T14:50:48.135
F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1
F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea
F LICENSE.md df5091916dbb40e6e9686186587125e1b2ff51f022cc334e886c19a0e9982724
@ -690,7 +690,7 @@ F src/hash.h 3340ab6e1d13e725571d7cee6d3e3135f0779a7d8e76a9ce0a85971fa3953c51
F src/hwtime.h f9c2dfb84dce7acf95ce6d289e46f5f9d3d1afd328e53da8f8e9008e3b3caae6
F src/in-operator.md 10cd8f4bcd225a32518407c2fb2484089112fd71
F src/insert.c 3f0a94082d978bbdd33c38fefea15346c6c6bffb70bc645a71dc0f1f87dd3276
F src/json.c 0a6095d10a8c8251e1838a1f12abc89125b0b05f82497ad12896a7f714e397ce
F src/json.c 15efd213cc95bde5b714ec068fdb1f6817b1ed9a253644833fed5f196957445a
F src/legacy.c d7874bc885906868cd51e6c2156698f2754f02d9eee1bae2d687323c3ca8e5aa
F src/loadext.c 7432c944ff197046d67a1207790a1b13eec4548c85a9457eb0896bb3641dfb36
F src/main.c 1b89f3de98d1b59fec5bac1d66d6ece21f703821b8eaa0d53d9604c35309f6f9
@ -2147,9 +2147,11 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93
F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc
F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e
F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0
P 8abc2ccaf8106f20243568cd7fa74174386eb85d7ea381201e97e2fd527033e0 174c2b2eef5fecd96a5fc89b81032fe81f7801f12097cea10e7e7f0a02114813
R 2fc453038906bf8ee420a799d6f3c8a1
T +closed 174c2b2eef5fecd96a5fc89b81032fe81f7801f12097cea10e7e7f0a02114813
P 7f0c79b94e8f55e5013e52ba64ba8b32dad1dc4e2224d2099733cbc561de1810
R 2684d5ad233d4af8d84cd782c9755832
T *branch * json-label-compare
T *sym-json-label-compare *
T -sym-trunk *
U drh
Z fc61504cc821223c214ac8778fb2caf0
Z 93dccc3939eaec1e85cde17c2d8ca8af
# Remove this line to create a well-formed Fossil manifest.

View File

@ -1 +1 @@
7f0c79b94e8f55e5013e52ba64ba8b32dad1dc4e2224d2099733cbc561de1810
2bc86d145fccc07107b7753cb1a69122676d4096fe59c454497bd81a6142d45e

View File

@ -2119,6 +2119,188 @@ static void jsonBlobEdit(
if( nIns && aIns ) memcpy(&pParse->aBlob[iDel], aIns, nIns);
}
/*
** Return the number of escaped newlines to be ignored.
** An escaped newline is a one of the following byte sequences:
**
** 0x5c 0x0a
** 0x5c 0x0d
** 0x5c 0x0d 0x0a
** 0x5c 0xe2 0x80 0xa8
** 0x5c 0xe2 0x80 0xa9
*/
static u32 jsonBytesToBypass(const char *z, u32 n){
u32 i = 0;
while( i+1<n ){
if( z[i]!='\\' ) return i;
if( z[i+1]=='\n' ){
i += 2;
continue;
}
if( z[i+1]=='\r' ){
if( i+2<n && z[i+2]=='\n' ){
i += 3;
}else{
i += 2;
}
continue;
}
if( 0xe2==(u8)z[i+1]
&& i+3<n
&& 0x80==(u8)z[i+2]
&& (0xa8==(u8)z[i+3] || 0xa9==(u8)z[i+3])
){
i += 4;
continue;
}
break;
}
return i;
}
/*
** Input z[0..n] defines JSON escape sequence including the leading '\\'.
** Decode that escape sequence into a single character. Write that
** character into *piOut. Return the number of bytes in the escape sequence.
*/
static u32 jsonUnescapeOneChar(const char *z, u32 n, u32 *piOut){
assert( n>0 );
assert( z[0]=='\\' );
if( n<2 ){
*piOut = 0xFFFD;
return n;
}
switch( (u8)z[1] ){
case 'u': {
u32 v, vlo;
if( n<6 ){
*piOut = 0xFFFD;
return n;
}
v = jsonHexToInt4(&z[2]);
if( (v & 0xfc00)==0xd800
&& n>=12
&& z[6]=='\\'
&& z[7]=='u'
&& ((vlo = jsonHexToInt4(&z[8]))&0xfc00)==0xdc00
){
*piOut = ((v&0x3ff)<<10) + (vlo&0x3ff) + 0x10000;
return 12;
}else{
*piOut = v;
return 6;
}
}
case 'b': { *piOut = '\b'; return 2; }
case 'f': { *piOut = '\f'; return 2; }
case 'n': { *piOut = '\n'; return 2; }
case 'r': { *piOut = '\r'; return 2; }
case 't': { *piOut = '\t'; return 2; }
case 'v': { *piOut = '\v'; return 2; }
case '0': { *piOut = 0; return 2; }
case '\'':
case '"':
case '/':
case '\\':{ *piOut = z[1]; return 2; }
case 'x': {
if( n<4 ){
*piOut = 0xFFFD;
return n;
}
*piOut = (jsonHexToInt(z[2])<<4) | jsonHexToInt(z[3]);
return 4;
}
case 0xe2:
case '\r':
case '\n': {
u32 nSkip = jsonBytesToBypass(z, n);
if( nSkip==0 ){
*piOut = 0xFFFD;
return n;
}else if( nSkip==n ){
*piOut = 0;
return n;
}else if( z[nSkip]=='\\' ){
return nSkip + jsonUnescapeOneChar(&z[nSkip], n-nSkip, piOut);
}else{
*piOut = z[nSkip];
return nSkip+1;
}
}
default: {
*piOut = 0xFFFD;
return 2;
}
}
}
/*
** Compare two object labels. Return 1 if they are equal and
** 0 if they differ.
**
** In this version, we know that one or the other or both of the
** two comparands contains an escape sequence.
*/
static SQLITE_NOINLINE int jsonLabelCompareEscaped(
const char *zLeft, /* The left label */
u32 nLeft, /* Size of the left label in bytes */
int rawLeft, /* True if zLeft contains no escapes */
const char *zRight, /* The right label */
u32 nRight, /* Size of the right label in bytes */
int rawRight /* True if zRight is escape-free */
){
u32 cLeft, cRight;
assert( rawLeft==0 || rawRight==0 );
while( nLeft>0 && nRight>0 ){
if( rawLeft || zLeft[0]!='\\' ){
cLeft = ((u8*)zLeft)[0];
zLeft++;
nLeft--;
}else{
u32 n = jsonUnescapeOneChar(zLeft, nLeft, &cLeft);
zLeft += n;
assert( n<=nLeft );
nLeft -= n;
}
if( rawRight || zRight[0]!='\\' ){
cRight = ((u8*)zRight)[0];
zRight++;
nRight--;
}else{
u32 n = jsonUnescapeOneChar(zRight, nRight, &cRight);
zRight += n;
assert( n<=nRight );
nRight -= n;
}
if( cLeft!=cRight ) return 0;
}
return nLeft==0 && nRight==0;
}
/*
** Compare two object labels. Return 1 if they are equal and
** 0 if they differ. Return -1 if an OOM occurs.
*/
static int jsonLabelCompare(
const char *zLeft, /* The left label */
u32 nLeft, /* Size of the left label in bytes */
int rawLeft, /* True if zLeft contains no escapes */
const char *zRight, /* The right label */
u32 nRight, /* Size of the right label in bytes */
int rawRight /* True if zRight is escape-free */
){
if( rawLeft && rawRight ){
/* Simpliest case: Neither label contains escapes. A simple
** memcmp() is sufficient. */
if( nLeft!=nRight ) return 0;
return memcmp(zLeft, zRight, nLeft)==0;
}else{
return jsonLabelCompareEscaped(zLeft, nLeft, rawLeft,
zRight, nRight, rawRight);
}
}
/*
** Error returns from jsonLookupStep()
*/
@ -2224,6 +2406,7 @@ static u32 jsonLookupStep(
return iRoot;
}
if( zPath[0]=='.' ){
int rawKey = 1;
x = pParse->aBlob[iRoot];
zPath++;
if( zPath[0]=='"' ){
@ -2236,6 +2419,7 @@ static u32 jsonLookupStep(
return JSON_LOOKUP_PATHERROR;
}
testcase( nKey==0 );
rawKey = memchr(zKey, '\\', nKey)==0;
}else{
zKey = zPath;
for(i=0; zPath[i] && zPath[i]!='.' && zPath[i]!='['; i++){}
@ -2249,13 +2433,17 @@ static u32 jsonLookupStep(
j = iRoot + n; /* j is the index of a label */
iEnd = j+sz;
while( j<iEnd ){
int rawLabel;
const char *zLabel;
x = pParse->aBlob[j] & 0x0f;
if( x<JSONB_TEXT || x>JSONB_TEXTRAW ) return JSON_LOOKUP_ERROR;
n = jsonbPayloadSize(pParse, j, &sz);
if( n==0 ) return JSON_LOOKUP_ERROR;
k = j+n; /* k is the index of the label text */
if( k+sz>=iEnd ) return JSON_LOOKUP_ERROR;
if( sz==nKey && memcmp(&pParse->aBlob[k], zKey, nKey)==0 ){
zLabel = (const char*)&pParse->aBlob[k];
rawLabel = x==JSONB_TEXT || x==JSONB_TEXTRAW;
if( jsonLabelCompare(zKey, nKey, rawKey, zLabel, sz, rawLabel) ){
u32 v = k+sz; /* v is the index of the value */
if( ((pParse->aBlob[v])&0x0f)>JSONB_OBJECT ) return JSON_LOOKUP_ERROR;
n = jsonbPayloadSize(pParse, v, &sz);
@ -2279,7 +2467,7 @@ static u32 jsonLookupStep(
testcase( pParse->eEdit==JEDIT_INS );
testcase( pParse->eEdit==JEDIT_SET );
memset(&ix, 0, sizeof(ix));
jsonBlobAppendNode(&ix,JSONB_TEXTRAW, nKey, 0);
jsonBlobAppendNode(&ix, rawKey?JSONB_TEXTRAW:JSONB_TEXT5, nKey, 0);
pParse->oom |= ix.oom;
rc = jsonCreateEditSubstructure(pParse, &v, &zPath[i]);
if( !JSON_LOOKUP_ISERROR(rc)
@ -2483,72 +2671,27 @@ static void jsonReturnFromBlob(
for(iIn=iOut=0; iIn<sz; iIn++){
char c = z[iIn];
if( c=='\\' ){
c = z[++iIn];
if( c=='u' ){
u32 v = jsonHexToInt4(z+iIn+1);
iIn += 4;
if( v==0 ) break;
if( v<=0x7f ){
zOut[iOut++] = (char)v;
}else if( v<=0x7ff ){
zOut[iOut++] = (char)(0xc0 | (v>>6));
zOut[iOut++] = 0x80 | (v&0x3f);
}else{
u32 vlo;
if( (v&0xfc00)==0xd800
&& iIn<sz-6
&& z[iIn+1]=='\\'
&& z[iIn+2]=='u'
&& ((vlo = jsonHexToInt4(z+iIn+3))&0xfc00)==0xdc00
){
/* We have a surrogate pair */
v = ((v&0x3ff)<<10) + (vlo&0x3ff) + 0x10000;
iIn += 6;
zOut[iOut++] = 0xf0 | (v>>18);
zOut[iOut++] = 0x80 | ((v>>12)&0x3f);
zOut[iOut++] = 0x80 | ((v>>6)&0x3f);
zOut[iOut++] = 0x80 | (v&0x3f);
}else{
zOut[iOut++] = 0xe0 | (v>>12);
zOut[iOut++] = 0x80 | ((v>>6)&0x3f);
zOut[iOut++] = 0x80 | (v&0x3f);
}
}
continue;
}else if( c=='b' ){
c = '\b';
}else if( c=='f' ){
c = '\f';
}else if( c=='n' ){
c = '\n';
}else if( c=='r' ){
c = '\r';
}else if( c=='t' ){
c = '\t';
}else if( c=='v' ){
c = '\v';
}else if( c=='\'' || c=='"' || c=='/' || c=='\\' ){
/* pass through unchanged */
}else if( c=='0' ){
c = 0;
}else if( c=='x' ){
c = (jsonHexToInt(z[iIn+1])<<4) | jsonHexToInt(z[iIn+2]);
iIn += 2;
}else if( c=='\r' && z[i+1]=='\n' ){
iIn++;
continue;
}else if( 0xe2==(u8)c
&& iIn<sz-2
&& 0x80==(u8)z[iIn+1]
&& (0xa8==(u8)z[iIn+2] || 0xa9==(u8)z[iIn+2])
){
iIn += 2;
continue;
u32 v;
u32 szEscape = jsonUnescapeOneChar(&z[iIn], sz-iIn, &v);
if( v<=0x7f ){
zOut[iOut++] = (char)v;
}else if( v<=0x7ff ){
zOut[iOut++] = (char)(0xc0 | (v>>6));
zOut[iOut++] = 0x80 | (v&0x3f);
}else if( v<0x10000 ){
zOut[iOut++] = 0xe0 | (v>>12);
zOut[iOut++] = 0x80 | ((v>>6)&0x3f);
zOut[iOut++] = 0x80 | (v&0x3f);
}else{
continue;
zOut[iOut++] = 0xf0 | (v>>18);
zOut[iOut++] = 0x80 | ((v>>12)&0x3f);
zOut[iOut++] = 0x80 | ((v>>6)&0x3f);
zOut[iOut++] = 0x80 | (v&0x3f);
}
} /* end if( c=='\\' ) */
zOut[iOut++] = c;
iIn += szEscape - 1;
}else{
zOut[iOut++] = c;
}
} /* end for() */
zOut[iOut] = 0;
sqlite3_result_text(pCtx, zOut, iOut, sqlite3_free);
@ -3384,6 +3527,7 @@ static int jsonMergePatch(
iTCursor = iTStart;
iTEnd = iTEndBE + pTarget->delta;
while( iTCursor<iTEnd ){
int isEqual; /* true if the patch and target labels match */
iTLabel = iTCursor;
eTLabel = pTarget->aBlob[iTCursor] & 0x0f;
if( eTLabel<JSONB_TEXT || eTLabel>JSONB_TEXTRAW ){
@ -3396,33 +3540,14 @@ static int jsonMergePatch(
nTValue = jsonbPayloadSize(pTarget, iTValue, &szTValue);
if( nTValue==0 ) return JSON_MERGE_BADTARGET;
if( iTValue + nTValue + szTValue > iTEnd ) return JSON_MERGE_BADTARGET;
if( eTLabel==ePLabel ){
/* Common case */
if( szTLabel==szPLabel
&& memcmp(&pTarget->aBlob[iTLabel+nTLabel],
&pPatch->aBlob[iPLabel+nPLabel], szTLabel)==0
){
break; /* Labels match. */
}
}else{
/* Should rarely happen */
JsonString s1, s2;
int isEqual, isOom;
jsonStringInit(&s1, 0);
jsonXlateBlobToText(pTarget, iTLabel, &s1);
jsonStringInit(&s2, 0);
jsonXlateBlobToText(pPatch, iPLabel, &s2);
isOom = s1.eErr || s2.eErr;
if( s1.nUsed==s2.nUsed && memcmp(s1.zBuf, s2.zBuf, s1.nUsed)==0 ){
isEqual = 1;
}else{
isEqual = 0;
}
jsonStringReset(&s1);
jsonStringReset(&s2);
if( isOom ) return JSON_MERGE_OOM;
if( isEqual ) break;
}
isEqual = jsonLabelCompare(
(const char*)&pPatch->aBlob[iPLabel+nPLabel],
szPLabel,
(ePLabel==JSONB_TEXT || ePLabel==JSONB_TEXTRAW),
(const char*)&pTarget->aBlob[iTLabel+nTLabel],
szTLabel,
(eTLabel==JSONB_TEXT || eTLabel==JSONB_TEXTRAW));
if( isEqual ) break;
iTCursor = iTValue + nTValue + szTValue;
}
x = pPatch->aBlob[iPValue] & 0x0f;