Index: ext/misc/json1.c ================================================================== --- ext/misc/json1.c +++ ext/misc/json1.c @@ -519,10 +519,41 @@ jsonInit(&s, pCtx); jsonRenderNode(pNode, &s, aReplace); jsonResult(&s); sqlite3_result_subtype(pCtx, JSON_SUBTYPE); } + +/* +** Translate a single byte of Hex into an integer. +** This routine only works if h really is a valid hexadecimal +** character: 0..9a..fA..F +*/ +static u8 jsonHexToInt(int h){ + assert( (h>='0' && h<='9') || (h>='a' && h<='f') || (h>='A' && h<='F') ); +#ifdef SQLITE_EBCDIC + h += 9*(1&~(h>>4)); +#else + h += 9*(1&(h>>6)); +#endif + return (u8)(h & 0xf); +} + +/* +** Convert a 4-byte hex string into an integer +*/ +static u32 jsonHexToInt4(const char *z){ + u32 v; + assert( safe_isxdigit(z[0]) ); + assert( safe_isxdigit(z[1]) ); + assert( safe_isxdigit(z[2]) ); + assert( safe_isxdigit(z[3]) ); + v = (jsonHexToInt(z[0])<<12) + + (jsonHexToInt(z[1])<<8) + + (jsonHexToInt(z[2])<<4) + + jsonHexToInt(z[3]); + return v; +} /* ** Make the JsonNode the return value of the function. */ static void jsonReturn( @@ -613,29 +644,38 @@ if( c!='\\' ){ zOut[j++] = c; }else{ c = z[++i]; if( c=='u' ){ - u32 v = 0, k; - for(k=0; k<4; i++, k++){ - assert( i>6)); zOut[j++] = 0x80 | (v&0x3f); }else{ - zOut[j++] = (char)(0xe0 | (v>>12)); - zOut[j++] = 0x80 | ((v>>6)&0x3f); - zOut[j++] = 0x80 | (v&0x3f); + u32 vlo; + if( (v&0xfc00)==0xd800 + && i>18); + zOut[j++] = 0x80 | ((v>>12)&0x3f); + zOut[j++] = 0x80 | ((v>>6)&0x3f); + zOut[j++] = 0x80 | (v&0x3f); + }else{ + zOut[j++] = 0xe0 | (v>>12); + zOut[j++] = 0x80 | ((v>>6)&0x3f); + zOut[j++] = 0x80 | (v&0x3f); + } } }else{ if( c=='b' ){ c = '\b'; }else if( c=='f' ){ Index: test/json101.test ================================================================== --- test/json101.test +++ test/json101.test @@ -829,7 +829,22 @@ SELECT * FROM (JSON_EACH('{"a":1, "b":2}')); } {a 1 integer 1 2 {} {$.a} {$} b 2 integer 2 4 {} {$.b} {$}} do_execsql_test json-15.130 { SELECT xyz.* FROM (JSON_EACH('{"a":1, "b":2}')) AS xyz; } {a 1 integer 1 2 {} {$.a} {$} b 2 integer 2 4 {} {$.b} {$}} + +# 2019-11-10 +# Mailing list bug report on the handling of surrogate pairs +# in JSON. +# +do_execsql_test json-16.10 { + SELECT length(json_extract('"abc\uD834\uDD1Exyz"','$')); +} {7} +do_execsql_test json-16.20 { + SELECT length(json_extract('"\uD834\uDD1E"','$')); +} {1} +do_execsql_test json-16.30 { + SELECT unicode(json_extract('"\uD834\uDD1E"','$')); +} {119070} + finish_test