Fossil

** performance is bounded by O(N*M) where N is the size of the regular
** expression and M is the size of the input string.  The matcher never
** exhibits exponential behavior.  Note that the X{p,q} operator expands
** to p copies of X following by q-p copies of X? and that the size of the
** regular expression in the O(N*M) performance bound is computed after
** this expansion.
**
** To help prevent DoS attacks, the size of the NFA is limit to
** SQLITE_MAX_REGEXP states, default 9999.
*/
#include <string.h>
#include <stdlib.h>
/* #include "sqlite3ext.h" */
SQLITE_EXTENSION_INIT1

#ifndef SQLITE_MAX_REGEXP
# define SQLITE_MAX_REGEXP 9999
#endif

/*
** The following #defines change the names of some functions implemented in
** this file to prevent name collisions with C-library functions of the
** same name.
*/
#define re_match   sqlite3re_match
#define re_compile sqlite3re_compile

#define RE_OP_NOTWORD    12    /* Not a perl word character */
#define RE_OP_DIGIT      13    /* digit:  [0-9] */
#define RE_OP_NOTDIGIT   14    /* Not a digit */
#define RE_OP_SPACE      15    /* space:  [ \t\n\r\v\f] */
#define RE_OP_NOTSPACE   16    /* Not a digit */
#define RE_OP_BOUNDARY   17    /* Boundary between word and non-word */
#define RE_OP_ATSTART    18    /* Currently at the start of the string */

#if defined(SQLITE_DEBUG)
/* Opcode names used for symbolic debugging */
static const char *ReOpName[] = {
  "EOF",
  "MATCH",
  "ANY",
  "ANYSTAR",
  "FORK",
  "GOTO",
  "ACCEPT",
  "CC_INC",
  "CC_EXC",
  "CC_VALUE",
  "CC_RANGE",
  "WORD",
  "NOTWORD",
  "DIGIT",
  "NOTDIGIT",
  "SPACE",
  "NOTSPACE",
  "BOUNDARY",
  "ATSTART",
};
#endif /* SQLITE_DEBUG */


/* Each opcode is a "state" in the NFA */
typedef unsigned short ReStateNumber;

/* Because this is an NFA and not a DFA, multiple states can be active at
** once.  An instance of the following object records all active states in
** the NFA.  The implementation is optimized for the common case where the

re_match_end:
  sqlite3_free(pToFree);
  return rc;
}

/* Resize the opcode and argument arrays for an RE under construction.
*/
static int re_resize(ReCompiled *p, int N){
  char *aOp;
  int *aArg;
  if( N>p->mxAlloc ){ p->zErr = "REGEXP pattern too big"; return 1; }
  aOp = sqlite3_realloc64(p->aOp, N*sizeof(p->aOp[0]));
  if( aOp==0 ){ p->zErr = "out of memory"; return 1; }
  p->aOp = aOp;
  aArg = sqlite3_realloc64(p->aArg, N*sizeof(p->aArg[0]));

static int re_append(ReCompiled *p, int op, int arg){
  return re_insert(p, p->nState, op, arg);
}

/* Make a copy of N opcodes starting at iStart onto the end of the RE
** under construction.
*/
static void re_copy(ReCompiled *p, int iStart, int N){
  if( p->nState+N>=p->nAlloc && re_resize(p, p->nAlloc*2+N) ) return;
  memcpy(&p->aOp[p->nState], &p->aOp[iStart], N*sizeof(p->aOp[0]));
  memcpy(&p->aArg[p->nState], &p->aArg[iStart], N*sizeof(p->aArg[0]));
  p->nState += N;
}

/* Return true if c is a hexadecimal digit character:  [0-9a-fA-F]

  return 0;
}

/* Free and reclaim all the memory used by a previously compiled
** regular expression.  Applications should invoke this routine once
** for every call to re_compile() to avoid memory leaks.
*/
static void re_free(void *p){
  ReCompiled *pRe = (ReCompiled*)p;
  if( pRe ){
    sqlite3_free(pRe->aOp);
    sqlite3_free(pRe->aArg);
    sqlite3_free(pRe);
  }
}

  const char *zErr;
  ReCompiled *pRe;
  sqlite3_str *pStr;
  int i;
  int n;
  char *z;
  (void)argc;






















  zPattern = (const char*)sqlite3_value_text(argv[0]);
  if( zPattern==0 ) return;
  zErr = re_compile(&pRe, zPattern, re_maxlen(context),
                    sqlite3_user_data(context)!=0);
  if( zErr ){
    re_free(pRe);

** the text of this file.  Search for "Begin file sqlite3.h" to find the start
** of the embedded sqlite3.h header file.) Additional code files may be needed
** if you want a wrapper to interface SQLite with your choice of programming
** language. The code for the "sqlite3" command-line shell is also in a
** separate file. This file contains only code for the core SQLite library.
**
** The content in this amalgamation comes from Fossil check-in
** 869c968569b09d05a5b7d587d8fddb3b4611 with changes in files:
**
**    
*/
#ifndef SQLITE_AMALGAMATION
#define SQLITE_CORE 1
#define SQLITE_AMALGAMATION 1
#ifndef SQLITE_PRIVATE

**
** See also: [sqlite3_libversion()],
** [sqlite3_libversion_number()], [sqlite3_sourceid()],
** [sqlite_version()] and [sqlite_source_id()].
*/
#define SQLITE_VERSION        "3.51.0"
#define SQLITE_VERSION_NUMBER 3051000
#define SQLITE_SOURCE_ID      "2025-09-26 15:38:52 869c968569b09d05a5b7d587d8fddb3b4611daf7467dc157701e5dc6c960alt1"
#define SQLITE_SCM_BRANCH     "trunk"
#define SQLITE_SCM_TAGS       ""
#define SQLITE_SCM_DATETIME   "2025-09-26T15:38:52.279Z"

/*
** CAPI3REF: Run-Time Library Version Numbers
** KEYWORDS: sqlite3_version sqlite3_sourceid
**
** These interfaces provide the same information as the [SQLITE_VERSION],
** [SQLITE_VERSION_NUMBER], and [SQLITE_SOURCE_ID] C preprocessor macros

static void fts5SourceIdFunc(
  sqlite3_context *pCtx,          /* Function call context */
  int nArg,                       /* Number of args */
  sqlite3_value **apUnused        /* Function arguments */
){
  assert( nArg==0 );
  UNUSED_PARAM2(nArg, apUnused);
  sqlite3_result_text(pCtx, "fts5: 2025-09-26 11:47:13 d022ee167b90a7c32049a93d476e869270018017f60551185024409730d77640", -1, SQLITE_TRANSIENT);
}

/*
** Implementation of fts5_locale(LOCALE, TEXT) function.
**
** If parameter LOCALE is NULL, or a zero-length string, then a copy of
** TEXT is returned. Otherwise, both LOCALE and TEXT are interpreted as

**
** See also: [sqlite3_libversion()],
** [sqlite3_libversion_number()], [sqlite3_sourceid()],
** [sqlite_version()] and [sqlite_source_id()].
*/
#define SQLITE_VERSION        "3.51.0"
#define SQLITE_VERSION_NUMBER 3051000
#define SQLITE_SOURCE_ID      "2025-09-26 15:38:52 869c968569b09d05a5b7d587d8fddb3b4611daf7467dc157701e5dc6c960alt1"
#define SQLITE_SCM_BRANCH     "trunk"
#define SQLITE_SCM_TAGS       ""
#define SQLITE_SCM_DATETIME   "2025-09-26T15:38:52.279Z"

/*
** CAPI3REF: Run-Time Library Version Numbers
** KEYWORDS: sqlite3_version sqlite3_sourceid
**
** These interfaces provide the same information as the [SQLITE_VERSION],
** [SQLITE_VERSION_NUMBER], and [SQLITE_SOURCE_ID] C preprocessor macros

** A nondeterministic finite automaton (NFA) is used for matching, so the
** performance is bounded by O(N*M) where N is the size of the regular
** expression and M is the size of the input string.  The matcher never
** exhibits exponential behavior.  Note that the X{p,q} operator expands
** to p copies of X following by q-p copies of X? and that the size of the
** regular expression in the O(N*M) performance bound is computed after
** this expansion.


*/
#include "config.h"
#include "regexp.h"

#ifndef SQLITE_MAX_REGEXP_REPEAT
# define SQLITE_MAX_REGEXP_REPEAT 999
#endif

/* The end-of-input character */
#define RE_EOF            0    /* End of input */
#define RE_START  0xfffffff    /* Start of input - larger than an UTF-8 */

/* The NFA is implemented as sequence of opcodes taken from the following
** set.  Each opcode has a single integer argument.
*/

struct ReCompiled {
  ReInput sIn;                /* Regular expression text */
  const char *zErr;           /* Error message to return */
  char *aOp;                  /* Operators for the virtual machine */
  int *aArg;                  /* Arguments to each operator */
  unsigned (*xNextChar)(ReInput*);  /* Next character function */
  unsigned char zInit[12];    /* Initial text to match */
  int nInit;                  /* Number of characters in zInit */
  unsigned nState;            /* Number of entries in aOp[] and aArg[] */
  unsigned nAlloc;            /* Slots allocated for aOp[] and aArg[] */
  unsigned mxAlloc;           /* Complexity limit */
};
#endif

/* Add a state to the given state set if it is not already there */

      c = (c&0x1f)<<6 | (p->z[p->i++]&0x3f);
      if( c<0x80 ) c = 0xfffd;
    }else if( (c&0xf0)==0xe0 && p->i+1<p->mx && (p->z[p->i]&0xc0)==0x80
           && (p->z[p->i+1]&0xc0)==0x80 ){
      c = (c&0x0f)<<12 | ((p->z[p->i]&0x3f)<<6) | (p->z[p->i+1]&0x3f);
      p->i += 2;
      if( c<=0x7ff || (c>=0xd800 && c<=0xdfff) ) c = 0xfffd;
    }else if( (c&0xf8)==0xf0 && p->i+3<p->mx && (p->z[p->i]&0xc0)==0x80
           && (p->z[p->i+1]&0xc0)==0x80 && (p->z[p->i+2]&0xc0)==0x80 ){
      c = (c&0x07)<<18 | ((p->z[p->i]&0x3f)<<12) | ((p->z[p->i+1]&0x3f)<<6)
                       | (p->z[p->i+2]&0x3f);
      p->i += 3;
      if( c<=0xffff || c>0x10ffff ) c = 0xfffd;
    }else{
      c = 0xfffd;

        }
        case RE_OP_ACCEPT: {
          rc = 1;
          goto re_match_end;
        }
        case RE_OP_CC_EXC: {
          if( c==0 ) break;
          /* fall-through */
        }
        case RE_OP_CC_INC: {
          int j = 1;
          int n = pRe->aArg[x];
          int hit = 0;
          for(j=1; j>0 && j<n; j++){
            if( pRe->aOp[x+j]==RE_OP_CC_VALUE ){
              if( pRe->aArg[x+j]==c ){
                hit = 1;

          while( (c=rePeek(p))>='0' && c<='9' ){
            n = n*10 + c-'0';
            if( n*2>p->mxAlloc ) return "REGEXP pattern too big";
            p->sIn.i++;
          }
        }
        if( c!='}' ) return "unmatched '{'";
        if( n>0 && n<m ) return "n less than m in '{m,n}'";
        p->sIn.i++;
        sz = p->nState - iPrev;
        if( m==0 ){
          if( n==0 ) return "both m and n are zero in '{m,n}'";
          re_insert(p, iPrev, RE_OP_FORK, sz+1);
          iPrev++;
          n--;

/*
** Compile a textual regular expression in zIn[] into a compiled regular
** expression suitable for us by re_match() and return a pointer to the
** compiled regular expression in *ppRe.  Return NULL on success or an
** error message if something goes wrong.
*/
const char *re_compile(
  ReCompiled **ppRe,      /* OUT: write compiled NFA here */
  const char *zIn,        /* Input regular expression */
  int mxRe,               /* Complexity limit */
  int noCase              /* True for caseless comparisons */
){
  ReCompiled *pRe;
  const char *zErr;

      }
    }
    if( j>0 && pRe->zInit[j-1]==0 ) j--;
    pRe->nInit = j;
  }
  return pRe->zErr;
}






































































/*
** The input zIn is a string that we want to match exactly as part of
** a regular expression.  Return a new string (in space obtained from
** fossil_malloc() or the equivalent) that escapes all regexp syntax
** characters in zIn.
*/

** SETTING:  regexp-limit                  width=8 default=1000
**
** Limit the size of the bytecode used to implement a regular expression
** to this many steps.  It is important to limit this to avoid possible
** DoS attacks.
*/

/*
** Compute a reasonable limit on the length of the REGEXP NFA.
*/
int re_maxlen(void){
  return g.db ? db_get_int("regexp-limit", 1000) : 1000;
}

/*
** Compile an RE using re_maxlen().
*/
const char *fossil_re_compile(
  ReCompiled **ppRe,      /* OUT: write compiled NFA here */
  const char *zIn,        /* Input regular expression */
  int noCase              /* True for caseless comparisons */
){

  return re_compile(ppRe, zIn, re_maxlen(), noCase);
}

/*
** Implementation of the regexp() SQL function.  This function implements
** the build-in REGEXP operator.  The first argument to the function is the
** pattern and the second argument is the string.  So, the SQL statements:
**
**       A REGEXP B
**
** is implemented as regexp(B,A).
*/
static void re_sql_func(
  sqlite3_context *context,
  int argc,
  sqlite3_value **argv
){
  ReCompiled *pRe;          /* Compiled regular expression */
  const char *zPattern;     /* The regular expression */
  const unsigned char *zStr;/* String being searched */
  const char *zErr;         /* Compile error message */
  int setAux = 0;           /* True to invoke sqlite3_set_auxdata() */

  (void)argc;  /* Unused */
  pRe = sqlite3_get_auxdata(context, 0);
  if( pRe==0 ){
    zPattern = (const char*)sqlite3_value_text(argv[0]);
    if( zPattern==0 ) return;
    zErr = fossil_re_compile(&pRe, zPattern, sqlite3_user_data(context)!=0);
    if( zErr ){
      re_free(pRe);
      sqlite3_result_int(context, 0);
      /* sqlite3_result_error(context, zErr, -1); */
      return;
    }
    if( pRe==0 ){
      sqlite3_result_error_nomem(context);
      return;
    }
    setAux = 1;
  }
  zStr = (const unsigned char*)sqlite3_value_text(argv[1]);
  if( zStr!=0 ){
    sqlite3_result_int(context, re_match(pRe, zStr, -1));
  }
  if( setAux ){
    sqlite3_set_auxdata(context, 0, pRe, (void(*)(void*))re_free);
  }
}

/*
** Invoke this routine to register the regexp() function with the
** SQLite database connection.
*/
int re_add_sql_func(sqlite3 *db){
  int rc;
  rc = sqlite3_create_function(db, "regexp", 2,
                           SQLITE_UTF8|SQLITE_INNOCUOUS|SQLITE_DETERMINISTIC,
                               0, re_sql_func, 0, 0);
  if( rc==SQLITE_OK ){
    /* The regexpi(PATTERN,STRING) function is a case-insensitive version
    ** of regexp(PATTERN,STRING). */
    rc = sqlite3_create_function(db, "regexpi", 2,
                           SQLITE_UTF8|SQLITE_INNOCUOUS|SQLITE_DETERMINISTIC,
                                 (void*)db, re_sql_func, 0, 0);
  }
  return rc;
}

/*
** Run a "grep" over a single file read from disk.
*/
static void grep_file(ReCompiled *pRe, const char *zFile, FILE *in){
  int ln = 0;