/* * tclCmdMZ.c -- * * This file contains the top-level command routines for most of * the Tcl built-in commands whose names begin with the letters * M to Z. It contains only commands in the generic core (i.e. * those that don't depend much upon UNIX facilities). * * Copyright (c) 1987-1993 The Regents of the University of California. * Copyright (c) 1994-1997 Sun Microsystems, Inc. * Copyright (c) 1998-1999 by Scriptics Corporation. * * See the file "license.terms" for information on usage and redistribution * of this file, and for a DISCLAIMER OF ALL WARRANTIES. * * RCS: @(#) $Id: tclCmdMZ.c,v 1.4 1999/05/04 01:33:10 stanton Exp $ */ #include "tclInt.h" #include "tclPort.h" #include "tclCompile.h" #include "tclRegexp.h" /* * Flag values used by Tcl_ScanObjCmd. */ #define SCAN_NOSKIP 0x1 /* Don't skip blanks. */ #define SCAN_SUPPRESS 0x2 /* Suppress assignment. */ #define SCAN_UNSIGNED 0x4 /* Read an unsigned value. */ #define SCAN_WIDTH 0x8 /* A width value was supplied. */ #define SCAN_SIGNOK 0x10 /* A +/- character is allowed. */ #define SCAN_NODIGITS 0x20 /* No digits have been scanned. */ #define SCAN_NOZERO 0x40 /* No zero digits have been scanned. */ #define SCAN_XOK 0x80 /* An 'x' is allowed. */ #define SCAN_PTOK 0x100 /* Decimal point is allowed. */ #define SCAN_EXPOK 0x200 /* An exponent is allowed. */ /* * Structure used to hold information about variable traces: */ typedef struct { int flags; /* Operations for which Tcl command is * to be invoked. */ char *errMsg; /* Error message returned from Tcl command, * or NULL. Malloc'ed. */ size_t length; /* Number of non-NULL chars. in command. */ char command[4]; /* Space for Tcl command to invoke. Actual * size will be as large as necessary to * hold command. This field must be the * last in the structure, so that it can * be larger than 4 bytes. */ } TraceVarInfo; /* * Forward declarations for procedures defined in this file: */ static char * TraceVarProc _ANSI_ARGS_((ClientData clientData, Tcl_Interp *interp, char *name1, char *name2, int flags)); /* *---------------------------------------------------------------------- * * Tcl_PwdObjCmd -- * * This procedure is invoked to process the "pwd" Tcl command. * See the user documentation for details on what it does. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_PwdObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { Tcl_DString ds; if (objc != 1) { Tcl_WrongNumArgs(interp, 1, objv, NULL); return TCL_ERROR; } if (Tcl_GetCwd(interp, &ds) == NULL) { return TCL_ERROR; } Tcl_DStringResult(interp, &ds); return TCL_OK; } /* *---------------------------------------------------------------------- * * Tcl_RegexpObjCmd -- * * This procedure is invoked to process the "regexp" Tcl command. * See the user documentation for details on what it does. The * REGEXP_TEST stuff is to minimize code differences between this * and the "testregexp" command. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_RegexpObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { int i, result, indices, stringLength, wLen, match, about; int cflags, eflags; Tcl_RegExp regExpr; char *string; Tcl_DString stringBuffer, valueBuffer; Tcl_UniChar *wStart; static char *options[] = { "-indices", "-nocase", "-about", "-expanded", "-line", "-linestop", "-lineanchor", "--", (char *) NULL }; enum options { REGEXP_INDICES, REGEXP_NOCASE, REGEXP_ABOUT, REGEXP_EXPANDED, REGEXP_LINE, REGEXP_LINESTOP, REGEXP_LINEANCHOR, REGEXP_LAST }; indices = 0; about = 0; cflags = REG_ADVANCED; eflags = 0; for (i = 1; i < objc; i++) { char *name; int index; name = Tcl_GetString(objv[i]); if (name[0] != '-') { break; } if (Tcl_GetIndexFromObj(interp, objv[i], options, "switch", TCL_EXACT, &index) != TCL_OK) { return TCL_ERROR; } switch ((enum options) index) { case REGEXP_INDICES: { indices = 1; break; } case REGEXP_NOCASE: { cflags |= REG_ICASE; break; } case REGEXP_ABOUT: { about = 1; break; } case REGEXP_EXPANDED: { cflags |= REG_EXPANDED; break; } case REGEXP_LINE: { cflags |= REG_NEWLINE; break; } case REGEXP_LINESTOP: { cflags |= REG_NLSTOP; break; } case REGEXP_LINEANCHOR: { cflags |= REG_NLANCH; break; } case REGEXP_LAST: { i++; goto endOfForLoop; } } } endOfForLoop: if (objc - i < 2 - about) { Tcl_WrongNumArgs(interp, 1, objv, "?switches? exp string ?matchVar? ?subMatchVar subMatchVar ...?"); return TCL_ERROR; } objc -= i; objv += i; regExpr = Tcl_GetRegExpFromObj(interp, objv[0], cflags); if (regExpr == NULL) { return TCL_ERROR; } if (about) { if (TclRegAbout(interp, regExpr) < 0) { return TCL_ERROR; } return TCL_OK; } result = TCL_OK; string = Tcl_GetStringFromObj(objv[1], &stringLength); Tcl_DStringInit(&valueBuffer); Tcl_DStringInit(&stringBuffer); wStart = Tcl_UtfToUniCharDString(string, stringLength, &stringBuffer); wLen = Tcl_DStringLength(&stringBuffer) / sizeof(Tcl_UniChar); match = TclRegExpExecUniChar(interp, regExpr, wStart, wLen, objc-2, eflags); if (match < 0) { result = TCL_ERROR; goto done; } if (match == 0) { /* * Set the interpreter's object result to an integer object w/ value 0. */ Tcl_SetIntObj(Tcl_GetObjResult(interp), 0); goto done; } /* * If additional variable names have been specified, return * index information in those variables. */ objc -= 2; objv += 2; for (i = 0; i < objc; i++) { char *varName, *value; int start, end; varName = Tcl_GetString(objv[i]); TclRegExpRangeUniChar(regExpr, i, &start, &end); if (start < 0) { if (indices) { value = Tcl_SetVar(interp, varName, "-1 -1", 0); } else { value = Tcl_SetVar(interp, varName, "", 0); } } else { if (indices) { char info[TCL_INTEGER_SPACE * 2]; sprintf(info, "%d %d", start, end - 1); value = Tcl_SetVar(interp, varName, info, 0); } else { value = Tcl_UniCharToUtfDString(wStart + start, end - start, &valueBuffer); value = Tcl_SetVar(interp, varName, value, 0); Tcl_DStringSetLength(&valueBuffer, 0); } } if (value == NULL) { Tcl_AppendResult(interp, "couldn't set variable \"", varName, "\"", (char *) NULL); result = TCL_ERROR; goto done; } } /* * Set the interpreter's object result to an integer object w/ value 1. */ Tcl_SetIntObj(Tcl_GetObjResult(interp), 1); done: Tcl_DStringFree(&stringBuffer); Tcl_DStringFree(&valueBuffer); return result; } /* *---------------------------------------------------------------------- * * Tcl_RegsubObjCmd -- * * This procedure is invoked to process the "regsub" Tcl command. * See the user documentation for details on what it does. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_RegsubObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { int i, result, flags, all, stringLength, numMatches; Tcl_RegExp regExpr; Tcl_DString resultBuffer, stringBuffer; CONST Tcl_UniChar *w, *wStart, *wEnd; char *string, *subspec, *varname; static char *options[] = { "-all", "-nocase", "--", NULL }; enum options { REGSUB_ALL, REGSUB_NOCASE, REGSUB_LAST }; flags = 0; all = 0; for (i = 1; i < objc; i++) { char *name; int index; name = Tcl_GetString(objv[i]); if (name[0] != '-') { break; } if (Tcl_GetIndexFromObj(interp, objv[i], options, "switch", TCL_EXACT, &index) != TCL_OK) { return TCL_ERROR; } switch ((enum options) index) { case REGSUB_ALL: { all = 1; break; } case REGSUB_NOCASE: { flags |= REG_ICASE; break; } case REGSUB_LAST: { i++; goto endOfForLoop; } } } endOfForLoop: if (objc - i != 4) { Tcl_WrongNumArgs(interp, 1, objv, "?switches? exp string subSpec varName"); return TCL_ERROR; } objv += i; regExpr = Tcl_GetRegExpFromObj(interp, objv[0], flags | REG_ADVANCED); if (regExpr == NULL) { return TCL_ERROR; } result = TCL_OK; string = Tcl_GetStringFromObj(objv[1], &stringLength); subspec = Tcl_GetString(objv[2]); varname = Tcl_GetString(objv[3]); Tcl_DStringInit(&resultBuffer); /* * The following loop is to handle multiple matches within the * same source string; each iteration handles one match and its * corresponding substitution. If "-all" hasn't been specified * then the loop body only gets executed once. */ Tcl_DStringInit(&stringBuffer); wStart = Tcl_UtfToUniCharDString(string, stringLength, &stringBuffer); wEnd = wStart + Tcl_DStringLength(&stringBuffer) / sizeof(Tcl_UniChar); numMatches = 0; for (w = wStart; w < wEnd; ) { int start, end, subStart, subEnd, match; char *src, *firstChar; char c; /* * The flags argument is set if string is part of a larger string, * so that "^" won't match. */ match = TclRegExpExecUniChar(interp, regExpr, w, wEnd - w, 10, ((w > wStart) ? REG_NOTBOL : 0)); if (match < 0) { result = TCL_ERROR; goto done; } if (match == 0) { break; } numMatches++; /* * Copy the portion of the source string before the match to the * result variable. */ TclRegExpRangeUniChar(regExpr, 0, &start, &end); Tcl_UniCharToUtfDString(w, start, &resultBuffer); /* * Append the subSpec argument to the variable, making appropriate * substitutions. This code is a bit hairy because of the backslash * conventions and because the code saves up ranges of characters in * subSpec to reduce the number of calls to Tcl_SetVar. */ src = subspec; firstChar = subspec; for (c = *src; c != '\0'; src++, c = *src) { int index; if (c == '&') { index = 0; } else if (c == '\\') { c = src[1]; if ((c >= '0') && (c <= '9')) { index = c - '0'; } else if ((c == '\\') || (c == '&')) { Tcl_DStringAppend(&resultBuffer, firstChar, src - firstChar); Tcl_DStringAppend(&resultBuffer, &c, 1); firstChar = src + 2; src++; continue; } else { continue; } } else { continue; } if (firstChar != src) { Tcl_DStringAppend(&resultBuffer, firstChar, src - firstChar); } TclRegExpRangeUniChar(regExpr, index, &subStart, &subEnd); if ((subStart >= 0) && (subEnd >= 0)) { Tcl_UniCharToUtfDString(w + subStart, subEnd - subStart, &resultBuffer); } if (*src == '\\') { src++; } firstChar = src + 1; } if (firstChar != src) { Tcl_DStringAppend(&resultBuffer, firstChar, src - firstChar); } if (end == 0) { /* * Always consume at least one character of the input string * in order to prevent infinite loops. */ Tcl_UniCharToUtfDString(w, 1, &resultBuffer); w++; } w += end; if (!all) { break; } } /* * Copy the portion of the source string after the last match to the * result variable. */ if ((w < wEnd) || (numMatches == 0)) { Tcl_UniCharToUtfDString(w, wEnd - w, &resultBuffer); } if (Tcl_SetVar(interp, varname, Tcl_DStringValue(&resultBuffer), 0) == NULL) { Tcl_AppendResult(interp, "couldn't set variable \"", varname, "\"", (char *) NULL); result = TCL_ERROR; } else { /* * Set the interpreter's object result to an integer object holding the * number of matches. */ Tcl_SetIntObj(Tcl_GetObjResult(interp), numMatches); } done: Tcl_DStringFree(&stringBuffer); Tcl_DStringFree(&resultBuffer); return result; } /* *---------------------------------------------------------------------- * * Tcl_RenameObjCmd -- * * This procedure is invoked to process the "rename" Tcl command. * See the user documentation for details on what it does. * * Results: * A standard Tcl object result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_RenameObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Arbitrary value passed to the command. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { char *oldName, *newName; if (objc != 3) { Tcl_WrongNumArgs(interp, 1, objv, "oldName newName"); return TCL_ERROR; } oldName = Tcl_GetString(objv[1]); newName = Tcl_GetString(objv[2]); return TclRenameCommand(interp, oldName, newName); } /* *---------------------------------------------------------------------- * * Tcl_ReturnObjCmd -- * * This object-based procedure is invoked to process the "return" Tcl * command. See the user documentation for details on what it does. * * Results: * A standard Tcl object result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_ReturnObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { Interp *iPtr = (Interp *) interp; int optionLen, argLen, code, result; if (iPtr->errorInfo != NULL) { ckfree(iPtr->errorInfo); iPtr->errorInfo = NULL; } if (iPtr->errorCode != NULL) { ckfree(iPtr->errorCode); iPtr->errorCode = NULL; } code = TCL_OK; for (objv++, objc--; objc > 1; objv += 2, objc -= 2) { char *option = Tcl_GetStringFromObj(objv[0], &optionLen); char *arg = Tcl_GetStringFromObj(objv[1], &argLen); if (strcmp(option, "-code") == 0) { register int c = arg[0]; if ((c == 'o') && (strcmp(arg, "ok") == 0)) { code = TCL_OK; } else if ((c == 'e') && (strcmp(arg, "error") == 0)) { code = TCL_ERROR; } else if ((c == 'r') && (strcmp(arg, "return") == 0)) { code = TCL_RETURN; } else if ((c == 'b') && (strcmp(arg, "break") == 0)) { code = TCL_BREAK; } else if ((c == 'c') && (strcmp(arg, "continue") == 0)) { code = TCL_CONTINUE; } else { result = Tcl_GetIntFromObj((Tcl_Interp *) NULL, objv[1], &code); if (result != TCL_OK) { Tcl_ResetResult(interp); Tcl_AppendStringsToObj(Tcl_GetObjResult(interp), "bad completion code \"", Tcl_GetString(objv[1]), "\": must be ok, error, return, break, ", "continue, or an integer", (char *) NULL); return result; } } } else if (strcmp(option, "-errorinfo") == 0) { iPtr->errorInfo = (char *) ckalloc((unsigned) (strlen(arg) + 1)); strcpy(iPtr->errorInfo, arg); } else if (strcmp(option, "-errorcode") == 0) { iPtr->errorCode = (char *) ckalloc((unsigned) (strlen(arg) + 1)); strcpy(iPtr->errorCode, arg); } else { Tcl_AppendStringsToObj(Tcl_GetObjResult(interp), "bad option \"", option, "\": must be -code, -errorcode, or -errorinfo", (char *) NULL); return TCL_ERROR; } } if (objc == 1) { /* * Set the interpreter's object result. An inline version of * Tcl_SetObjResult. */ Tcl_SetObjResult(interp, objv[0]); } iPtr->returnCode = code; return TCL_RETURN; } /* *---------------------------------------------------------------------- * * Tcl_SourceObjCmd -- * * This procedure is invoked to process the "source" Tcl command. * See the user documentation for details on what it does. * * Results: * A standard Tcl object result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_SourceObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { char *bytes; int result; if (objc != 2) { Tcl_WrongNumArgs(interp, 1, objv, "fileName"); return TCL_ERROR; } bytes = Tcl_GetString(objv[1]); result = Tcl_EvalFile(interp, bytes); return result; } /* *---------------------------------------------------------------------- * * Tcl_SplitObjCmd -- * * This procedure is invoked to process the "split" Tcl command. * See the user documentation for details on what it does. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_SplitObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { Tcl_UniChar ch; int len; char *splitChars, *string, *end; int splitCharLen, stringLen; Tcl_Obj *listPtr, *objPtr; if (objc == 2) { splitChars = " \n\t\r"; splitCharLen = 4; } else if (objc == 3) { splitChars = Tcl_GetStringFromObj(objv[2], &splitCharLen); } else { Tcl_WrongNumArgs(interp, 1, objv, "string ?splitChars?"); return TCL_ERROR; } string = Tcl_GetStringFromObj(objv[1], &stringLen); end = string + stringLen; listPtr = Tcl_GetObjResult(interp); if (stringLen == 0) { /* * Do nothing. */ } else if (splitCharLen == 0) { /* * Handle the special case of splitting on every character. */ for ( ; string < end; string += len) { len = Tcl_UtfToUniChar(string, &ch); objPtr = Tcl_NewStringObj(string, len); Tcl_ListObjAppendElement(NULL, listPtr, objPtr); } } else { char *element, *p, *splitEnd; int splitLen; Tcl_UniChar splitChar; /* * Normal case: split on any of a given set of characters. * Discard instances of the split characters. */ splitEnd = splitChars + splitCharLen; for (element = string; string < end; string += len) { len = Tcl_UtfToUniChar(string, &ch); for (p = splitChars; p < splitEnd; p += splitLen) { splitLen = Tcl_UtfToUniChar(p, &splitChar); if (ch == splitChar) { objPtr = Tcl_NewStringObj(element, string - element); Tcl_ListObjAppendElement(NULL, listPtr, objPtr); element = string + len; break; } } } objPtr = Tcl_NewStringObj(element, string - element); Tcl_ListObjAppendElement(NULL, listPtr, objPtr); } return TCL_OK; } /* *---------------------------------------------------------------------- * * Tcl_StringObjCmd -- * * This procedure is invoked to process the "string" Tcl command. * See the user documentation for details on what it does. Note * that this command only functions correctly on properly formed * Tcl UTF strings. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_StringObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { int index, left, right; Tcl_Obj *resultPtr; char *string1, *string2; int length1, length2; static char *options[] = { "bytes", "compare", "equal", "first", "icompare", "iequal", "index", "last", "length", "map", "match", "range", "repeat", "replace", "tolower", "toupper", "totitle", "trim", "trimleft", "trimright", "wordend", "wordstart", (char *) NULL }; enum options { STR_BYTES, STR_COMPARE, STR_EQUAL, STR_FIRST, STR_ICOMPARE, STR_IEQUAL, STR_INDEX, STR_LAST, STR_LENGTH, STR_MAP, STR_MATCH, STR_RANGE, STR_REPEAT, STR_REPLACE, STR_TOLOWER, STR_TOUPPER, STR_TOTITLE, STR_TRIM, STR_TRIMLEFT, STR_TRIMRIGHT, STR_WORDEND, STR_WORDSTART }; if (objc < 2) { Tcl_WrongNumArgs(interp, 1, objv, "option arg ?arg ...?"); return TCL_ERROR; } if (Tcl_GetIndexFromObj(interp, objv[1], options, "option", 0, &index) != TCL_OK) { return TCL_ERROR; } resultPtr = Tcl_GetObjResult(interp); switch ((enum options) index) { case STR_EQUAL: case STR_COMPARE: { int match, length, reqlength = -1; if (!(objc == 4 || objc == 5)) { Tcl_WrongNumArgs(interp, 2, objv, "string1 string2 ?length?"); return TCL_ERROR; } if ((objc == 5) && Tcl_GetIntFromObj(interp, objv[4], &reqlength) != TCL_OK) { return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); string2 = Tcl_GetStringFromObj(objv[3], &length2); length = (length1 < length2) ? length1 : length2; if ((reqlength >= 0) && (reqlength < length)) { Tcl_UniChar ch1, ch2; /* * reqlength must be interpreted as chars, not bytes * we will only enter here when both strings are of * at least reqlength chars long (no need for \0 check) */ match = 0; while (reqlength-- > 0) { string1 += Tcl_UtfToUniChar(string1, &ch1); string2 += Tcl_UtfToUniChar(string2, &ch2); if (ch1 != ch2) { match = ch1 - ch2; break; } } } else { match = memcmp(string1, string2, (unsigned) length); if (match == 0) { match = length1 - length2; } } if ((enum options) index == STR_EQUAL) { Tcl_SetIntObj(resultPtr, (match) ? 0 : 1); } else { Tcl_SetIntObj(resultPtr, ((match > 0) ? 1 : (match < 0) ? -1 : 0)); } break; } case STR_IEQUAL: case STR_ICOMPARE: { int match, length, reqlength = -1; Tcl_UniChar ch1, ch2; if (objc < 4 || objc > 5) { Tcl_WrongNumArgs(interp, 2, objv, "string1 string2 ?length?"); return TCL_ERROR; } if ((objc == 5) && Tcl_GetIntFromObj(interp, objv[4], &reqlength) != TCL_OK) { return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); string2 = Tcl_GetStringFromObj(objv[3], &length2); length = (length1 < length2) ? length1 : length2; if ((reqlength >= 0) && (reqlength < length)) { length = reqlength; } /* * length must be interpreted as chars, not bytes * we will only enter here when both strings are of * at least length chars long (no need for \0 check) */ match = 0; while (length-- > 0) { string1 += Tcl_UtfToUniChar(string1, &ch1); string2 += Tcl_UtfToUniChar(string2, &ch2); if ((ch1 != ch2) && (Tcl_UniCharToLower(ch1) != Tcl_UniCharToLower(ch2))) { match = ch1 - ch2; break; } } if ((match == 0) && (reqlength >= length)) { match = length1 - length2; } if ((enum options) index == STR_IEQUAL) { Tcl_SetIntObj(resultPtr, (match) ? 0 : 1); } else { Tcl_SetIntObj(resultPtr, ((match > 0) ? 1 : (match < 0) ? -1 : 0)); } break; } case STR_FIRST: { register char *p, *end; int match; if (objc != 4) { badFirstLastArgs: Tcl_WrongNumArgs(interp, 2, objv, "string1 string2"); return TCL_ERROR; } /* * This algorithm fails on improperly formed UTF strings. */ match = -1; string1 = Tcl_GetStringFromObj(objv[2], &length1); string2 = Tcl_GetStringFromObj(objv[3], &length2); if (length1 > 0) { end = string2 + length2 - length1 + 1; for (p = string2; p < end; p++) { /* * Scan forward to find the first character. */ p = memchr(p, *string1, (unsigned) (end - p)); if (p == NULL) { break; } if (memcmp(string1, p, (unsigned) length1) == 0) { match = p - string2; break; } } } /* * Compute the character index of the matching string by counting * the number of characters before the match. */ if (match != -1) { match = Tcl_NumUtfChars(string2, match); } Tcl_SetIntObj(resultPtr, match); break; } case STR_INDEX: { int index; char buf[TCL_UTF_MAX]; if (objc != 4) { Tcl_WrongNumArgs(interp, 2, objv, "string charIndex"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); /* * establish what 'end' really means */ length2 = Tcl_NumUtfChars(string1, length1) - 1; if (TclGetIntForIndex(interp, objv[3], length2, &index) != TCL_OK) { return TCL_ERROR; } if ((index >= 0) && (index < length1)) { length2 = Tcl_UniCharToUtf(Tcl_UniCharAtIndex(string1, index), buf); Tcl_SetStringObj(resultPtr, buf, length2); } break; } case STR_LAST: { register char *p; int match; if (objc != 4) { goto badFirstLastArgs; } /* * This algorithm fails on improperly formed UTF strings. */ match = -1; string1 = Tcl_GetStringFromObj(objv[2], &length1); string2 = Tcl_GetStringFromObj(objv[3], &length2); if (length1 > 0) { for (p = string2 + length2 - length1; p >= string2; p--) { /* * Scan backwards to find the first character. */ while ((p != string2) && (*p != *string1)) { p--; } if (memcmp(string1, p, (unsigned) length1) == 0) { match = p - string2; break; } } } /* * Compute the character index of the matching string by counting * the number of characters before the match. */ if (match != -1) { match = Tcl_NumUtfChars(string2, match); } Tcl_SetIntObj(resultPtr, match); break; } case STR_BYTES: case STR_LENGTH: { if (objc != 3) { Tcl_WrongNumArgs(interp, 2, objv, "string"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); if ((enum options) index == STR_BYTES) { Tcl_SetIntObj(resultPtr, length1); } else { Tcl_SetIntObj(resultPtr, Tcl_NumUtfChars(string1, length1)); } break; } case STR_MAP: { int mapElemc, len; Tcl_Obj **mapElemv; char *end; Tcl_UniChar ch; if (objc != 4) { Tcl_WrongNumArgs(interp, 2, objv, "charMap string"); return TCL_ERROR; } if (Tcl_ListObjGetElements(interp, objv[objc-2], &mapElemc, &mapElemv) != TCL_OK) { return TCL_ERROR; } if (mapElemc & 1) { /* * The charMap must be an even number of key/value items */ Tcl_SetStringObj(resultPtr, "char map list unbalanced", -1); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[objc-1], &length1); if (length1 == 0) { break; } end = string1 + length1; for ( ; string1 < end; string1 += len) { len = Tcl_UtfToUniChar(string1, &ch); for (index = 0; index < mapElemc; index +=2) { /* * Get the key string to match on */ string2 = Tcl_GetStringFromObj(mapElemv[index], &length2); if ((*string2 == *string1) && (memcmp(string2, string1, length2) == 0)) { /* * Adjust len to be full length of matched string */ len = length2; /* * Change string2 and length2 to the replacement value */ string2 = Tcl_GetStringFromObj(mapElemv[index+1], &length2); Tcl_AppendToObj(resultPtr, string2, length2); break; } } if (index == mapElemc) { /* * No match was found, put the char onto result */ Tcl_AppendToObj(resultPtr, string1, len); } } break; } case STR_MATCH: { if (objc != 4) { Tcl_WrongNumArgs(interp, 2, objv, "pattern string"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); string2 = Tcl_GetStringFromObj(objv[3], &length2); Tcl_SetBooleanObj(resultPtr, Tcl_StringMatch(string2, string1)); break; } case STR_RANGE: { int first, last; if (objc != 5) { Tcl_WrongNumArgs(interp, 2, objv, "string first last"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); length1 = Tcl_NumUtfChars(string1, length1); if (TclGetIntForIndex(interp, objv[3], length1 - 1, &first) != TCL_OK) { return TCL_ERROR; } if (TclGetIntForIndex(interp, objv[4], length1 - 1, &last) != TCL_OK) { return TCL_ERROR; } if (first < 0) { first = 0; } if (last >= length1 - 1) { last = length1 - 1; } if (last >= first) { char *start, *end; start = Tcl_UtfAtIndex(string1, first); end = Tcl_UtfAtIndex(start, last - first + 1); Tcl_SetStringObj(resultPtr, start, end - start); } break; } case STR_REPEAT: { int count; if (objc != 4) { Tcl_WrongNumArgs(interp, 2, objv, "string count"); return TCL_ERROR; } if (Tcl_GetIntFromObj(interp, objv[3], &count) != TCL_OK) { return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); if (length1 > 0) { for (index = 0; index < count; index++) { Tcl_AppendToObj(resultPtr, string1, length1); } } break; } case STR_REPLACE: { int first, last; if (!(objc == 5 || objc == 6)) { Tcl_WrongNumArgs(interp, 2, objv, "string first last ?string?"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); length1 = Tcl_NumUtfChars(string1, length1) - 1; if (TclGetIntForIndex(interp, objv[3], length1, &first) != TCL_OK) { return TCL_ERROR; } if (TclGetIntForIndex(interp, objv[4], length1, &last) != TCL_OK) { return TCL_ERROR; } if ((last < first) || (first > length1) || (last < 0)) { Tcl_SetObjResult(interp, objv[2]); } else { char *start, *end; if (first < 0) { first = 0; } start = Tcl_UtfAtIndex(string1, first); end = Tcl_UtfAtIndex(start, ((last > length1) ? length1 : last) - first + 1); Tcl_SetStringObj(resultPtr, string1, start - string1); if (objc == 6) { Tcl_AppendObjToObj(resultPtr, objv[5]); } if (last < length1) { Tcl_AppendToObj(resultPtr, end, -1); } } break; } case STR_TOLOWER: case STR_TOUPPER: case STR_TOTITLE: if (objc < 3 || objc > 5) { Tcl_WrongNumArgs(interp, 2, objv, "string ?first? ?last?"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); if (objc == 3) { /* * Since the result object is not a shared object, it is * safe to copy the string into the result and do the * conversion in place. The conversion may change the length * of the string, so reset the length after conversion. */ Tcl_SetStringObj(resultPtr, string1, length1); if ((enum options) index == STR_TOLOWER) { length1 = Tcl_UtfToLower(Tcl_GetStringFromObj(resultPtr, NULL)); } else if ((enum options) index == STR_TOUPPER) { length1 = Tcl_UtfToUpper(Tcl_GetStringFromObj(resultPtr, NULL)); } else { length1 = Tcl_UtfToTitle(Tcl_GetStringFromObj(resultPtr, NULL)); } Tcl_SetObjLength(resultPtr, length1); } else { int first, last; char *start, *end; length1 = Tcl_NumUtfChars(string1, length1) - 1; if (TclGetIntForIndex(interp, objv[3], length1, &first) != TCL_OK) { return TCL_ERROR; } if (first < 0) { first = 0; } last = first; if ((objc == 5) && (TclGetIntForIndex(interp, objv[4], length1, &last) != TCL_OK)) { return TCL_ERROR; } if (last >= length1) { last = length1; } if (last < first) { Tcl_SetObjResult(interp, objv[2]); break; } start = Tcl_UtfAtIndex(string1, first); end = Tcl_UtfAtIndex(start, last - first + 1); length2 = end-start; string2 = ckalloc(length2+1); memcpy(string2, start, length2); string2[length2] = '\0'; if ((enum options) index == STR_TOLOWER) { length2 = Tcl_UtfToLower(string2); } else if ((enum options) index == STR_TOUPPER) { length2 = Tcl_UtfToUpper(string2); } else { length2 = Tcl_UtfToTitle(string2); } Tcl_SetStringObj(resultPtr, string1, start - string1); Tcl_AppendToObj(resultPtr, string2, length2); Tcl_AppendToObj(resultPtr, end, -1); } break; case STR_TRIM: { Tcl_UniChar ch, trim; register char *p, *end; char *check, *checkEnd; int offset; left = 1; right = 1; dotrim: if (objc == 4) { string2 = Tcl_GetStringFromObj(objv[3], &length2); } else if (objc == 3) { string2 = " \t\n\r"; length2 = strlen(string2); } else { Tcl_WrongNumArgs(interp, 2, objv, "string ?chars?"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); checkEnd = string2 + length2; if (left) { end = string1 + length1; /* * The outer loop iterates over the string. The inner * loop iterates over the trim characters. The loops * terminate as soon as a non-trim character is discovered * and string1 is left pointing at the first non-trim * character. */ for (p = string1; p < end; p += offset) { offset = Tcl_UtfToUniChar(p, &ch); for (check = string2; ; ) { if (check >= checkEnd) { p = end; break; } check += Tcl_UtfToUniChar(check, &trim); if (ch == trim) { length1 -= offset; string1 += offset; break; } } } } if (right) { end = string1; /* * The outer loop iterates over the string. The inner * loop iterates over the trim characters. The loops * terminate as soon as a non-trim character is discovered * and length1 marks the last non-trim character. */ for (p = string1 + length1; p > end; ) { p = Tcl_UtfPrev(p, string1); offset = Tcl_UtfToUniChar(p, &ch); for (check = string2; ; ) { if (check >= checkEnd) { p = end; break; } check += Tcl_UtfToUniChar(check, &trim); if (ch == trim) { length1 -= offset; break; } } } } Tcl_SetStringObj(resultPtr, string1, length1); break; } case STR_TRIMLEFT: { left = 1; right = 0; goto dotrim; } case STR_TRIMRIGHT: { left = 0; right = 1; goto dotrim; } case STR_WORDEND: { int cur; Tcl_UniChar ch; char *p, *end; int numChars; if (objc != 4) { Tcl_WrongNumArgs(interp, 2, objv, "string index"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); numChars = Tcl_NumUtfChars(string1, length1); if (TclGetIntForIndex(interp, objv[3], numChars-1, &index) != TCL_OK) { return TCL_ERROR; } if (index < 0) { index = 0; } if (index < numChars) { p = Tcl_UtfAtIndex(string1, index); end = string1+length1; for (cur = index; p < end; cur++) { p += Tcl_UtfToUniChar(p, &ch); if (!Tcl_UniCharIsWordChar(ch)) { break; } } if (cur == index) { cur++; } } else { cur = numChars; } Tcl_SetIntObj(resultPtr, cur); break; } case STR_WORDSTART: { int cur; Tcl_UniChar ch; char *p; int numChars; if (objc != 4) { Tcl_WrongNumArgs(interp, 2, objv, "string index"); return TCL_ERROR; } string1 = Tcl_GetStringFromObj(objv[2], &length1); numChars = Tcl_NumUtfChars(string1, length1); if (TclGetIntForIndex(interp, objv[3], numChars-1, &index) != TCL_OK) { return TCL_ERROR; } if (index >= numChars) { index = numChars - 1; } cur = 0; if (index > 0) { p = Tcl_UtfAtIndex(string1, index); for (cur = index; cur >= 0; cur--) { Tcl_UtfToUniChar(p, &ch); if (!Tcl_UniCharIsWordChar(ch)) { break; } p = Tcl_UtfPrev(p, string1); } if (cur != index) { cur += 1; } } Tcl_SetIntObj(resultPtr, cur); break; } } return TCL_OK; } /* *---------------------------------------------------------------------- * * Tcl_SubstObjCmd -- * * This procedure is invoked to process the "subst" Tcl command. * See the user documentation for details on what it does. This * command is an almost direct copy of an implementation by * Andrew Payne. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_SubstObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { static char *substOptions[] = { "-nobackslashes", "-nocommands", "-novariables", (char *) NULL }; enum substOptions { SUBST_NOBACKSLASHES, SUBST_NOCOMMANDS, SUBST_NOVARS }; Interp *iPtr = (Interp *) interp; Tcl_DString result; char *p, *old, *value; int optionIndex, code, count, doVars, doCmds, doBackslashes, i; /* * Parse command-line options. */ doVars = doCmds = doBackslashes = 1; for (i = 1; i < (objc-1); i++) { p = Tcl_GetString(objv[i]); if (*p != '-') { break; } if (Tcl_GetIndexFromObj(interp, objv[i], substOptions, "switch", 0, &optionIndex) != TCL_OK) { return TCL_ERROR; } switch (optionIndex) { case SUBST_NOBACKSLASHES: { doBackslashes = 0; break; } case SUBST_NOCOMMANDS: { doCmds = 0; break; } case SUBST_NOVARS: { doVars = 0; break; } default: { panic("Tcl_SubstObjCmd: bad option index to SubstOptions"); } } } if (i != (objc-1)) { Tcl_WrongNumArgs(interp, 1, objv, "?-nobackslashes? ?-nocommands? ?-novariables? string"); return TCL_ERROR; } /* * Scan through the string one character at a time, performing * command, variable, and backslash substitutions. */ Tcl_DStringInit(&result); old = p = Tcl_GetString(objv[i]); while (*p != 0) { switch (*p) { case '\\': if (doBackslashes) { char buf[TCL_UTF_MAX]; if (p != old) { Tcl_DStringAppend(&result, old, p-old); } Tcl_DStringAppend(&result, buf, Tcl_UtfBackslash(p, &count, buf)); p += count; old = p; } else { p++; } break; case '$': if (doVars) { if (p != old) { Tcl_DStringAppend(&result, old, p-old); } value = Tcl_ParseVar(interp, p, &p); if (value == NULL) { Tcl_DStringFree(&result); return TCL_ERROR; } Tcl_DStringAppend(&result, value, -1); old = p; } else { p++; } break; case '[': if (doCmds) { if (p != old) { Tcl_DStringAppend(&result, old, p-old); } iPtr->evalFlags = TCL_BRACKET_TERM; code = Tcl_Eval(interp, p+1); if (code == TCL_ERROR) { Tcl_DStringFree(&result); return code; } old = p = (p+1 + iPtr->termOffset+1); Tcl_DStringAppend(&result, iPtr->result, -1); Tcl_ResetResult(interp); } else { p++; } break; default: p++; break; } } if (p != old) { Tcl_DStringAppend(&result, old, p-old); } Tcl_DStringResult(interp, &result); return TCL_OK; } /* *---------------------------------------------------------------------- * * Tcl_SwitchObjCmd -- * * This object-based procedure is invoked to process the "switch" Tcl * command. See the user documentation for details on what it does. * * Results: * A standard Tcl object result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_SwitchObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { int i, j, index, mode, matched, result; char *string, *pattern; static char *options[] = { "-exact", "-glob", "-regexp", "--", NULL }; enum options { OPT_EXACT, OPT_GLOB, OPT_REGEXP, OPT_LAST }; mode = OPT_EXACT; for (i = 1; i < objc; i++) { string = Tcl_GetString(objv[i]); if (string[0] != '-') { break; } if (Tcl_GetIndexFromObj(interp, objv[i], options, "option", 0, &index) != TCL_OK) { return TCL_ERROR; } if (index == OPT_LAST) { i++; break; } mode = index; } if (objc - i < 2) { Tcl_WrongNumArgs(interp, 1, objv, "?switches? string pattern body ... ?default body?"); return TCL_ERROR; } string = Tcl_GetString(objv[i]); objc -= i + 1; objv += i + 1; /* * If all of the pattern/command pairs are lumped into a single * argument, split them out again. */ if (objc == 1) { Tcl_Obj **listv; if (Tcl_ListObjGetElements(interp, objv[0], &objc, &listv) != TCL_OK) { return TCL_ERROR; } objv = listv; } for (i = 0; i < objc; i += 2) { if (i == objc - 1) { Tcl_ResetResult(interp); Tcl_AppendToObj(Tcl_GetObjResult(interp), "extra switch pattern with no body", -1); return TCL_ERROR; } /* * See if the pattern matches the string. */ pattern = Tcl_GetString(objv[i]); matched = 0; if ((i == objc - 2) && (*pattern == 'd') && (strcmp(pattern, "default") == 0)) { matched = 1; } else { switch (mode) { case OPT_EXACT: matched = (strcmp(string, pattern) == 0); break; case OPT_GLOB: matched = Tcl_StringMatch(string, pattern); break; case OPT_REGEXP: matched = TclRegExpMatchObj(interp, string, objv[i]); if (matched < 0) { return TCL_ERROR; } break; } } if (matched == 0) { continue; } /* * We've got a match. Find a body to execute, skipping bodies * that are "-". */ for (j = i + 1; ; j += 2) { if (j >= objc) { Tcl_AppendStringsToObj(Tcl_GetObjResult(interp), "no body specified for pattern \"", pattern, "\"", (char *) NULL); return TCL_ERROR; } if (strcmp(Tcl_GetString(objv[j]), "-") != 0) { break; } } result = Tcl_EvalObjEx(interp, objv[j], 0); if (result == TCL_ERROR) { char msg[100 + TCL_INTEGER_SPACE]; sprintf(msg, "\n (\"%.50s\" arm line %d)", pattern, interp->errorLine); Tcl_AddObjErrorInfo(interp, msg, -1); } return result; } return TCL_OK; } /* *---------------------------------------------------------------------- * * Tcl_TimeObjCmd -- * * This object-based procedure is invoked to process the "time" Tcl * command. See the user documentation for details on what it does. * * Results: * A standard Tcl object result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_TimeObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { register Tcl_Obj *objPtr; register int i, result; int count; double totalMicroSec; Tcl_Time start, stop; char buf[100]; if (objc == 2) { count = 1; } else if (objc == 3) { result = Tcl_GetIntFromObj(interp, objv[2], &count); if (result != TCL_OK) { return result; } } else { Tcl_WrongNumArgs(interp, 1, objv, "command ?count?"); return TCL_ERROR; } objPtr = objv[1]; i = count; TclpGetTime(&start); while (i-- > 0) { result = Tcl_EvalObjEx(interp, objPtr, 0); if (result != TCL_OK) { return result; } } TclpGetTime(&stop); totalMicroSec = (stop.sec - start.sec)*1000000 + (stop.usec - start.usec); sprintf(buf, "%.0f microseconds per iteration", ((count <= 0) ? 0 : totalMicroSec/count)); Tcl_ResetResult(interp); Tcl_AppendToObj(Tcl_GetObjResult(interp), buf, -1); return TCL_OK; } /* *---------------------------------------------------------------------- * * Tcl_TraceObjCmd -- * * This procedure is invoked to process the "trace" Tcl command. * See the user documentation for details on what it does. * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_TraceObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { int optionIndex, commandLength; char *name, *rwuOps, *command, *p; size_t length; static char *traceOptions[] = { "variable", "vdelete", "vinfo", (char *) NULL }; enum traceOptions { TRACE_VARIABLE, TRACE_VDELETE, TRACE_VINFO }; if (objc < 2) { Tcl_WrongNumArgs(interp, 1, objv, "option [arg arg ...]"); return TCL_ERROR; } if (Tcl_GetIndexFromObj(interp, objv[1], traceOptions, "option", 0, &optionIndex) != TCL_OK) { return TCL_ERROR; } switch ((enum traceOptions) optionIndex) { case TRACE_VARIABLE: { int flags; TraceVarInfo *tvarPtr; if (objc != 5) { Tcl_WrongNumArgs(interp, 2, objv, "name ops command"); return TCL_ERROR; } flags = 0; rwuOps = Tcl_GetString(objv[3]); for (p = rwuOps; *p != 0; p++) { if (*p == 'r') { flags |= TCL_TRACE_READS; } else if (*p == 'w') { flags |= TCL_TRACE_WRITES; } else if (*p == 'u') { flags |= TCL_TRACE_UNSETS; } else { goto badOps; } } if (flags == 0) { goto badOps; } command = Tcl_GetStringFromObj(objv[4], &commandLength); length = (size_t) commandLength; tvarPtr = (TraceVarInfo *) ckalloc((unsigned) (sizeof(TraceVarInfo) - sizeof(tvarPtr->command) + length + 1)); tvarPtr->flags = flags; tvarPtr->errMsg = NULL; tvarPtr->length = length; flags |= TCL_TRACE_UNSETS; strcpy(tvarPtr->command, command); name = Tcl_GetString(objv[2]); if (Tcl_TraceVar(interp, name, flags, TraceVarProc, (ClientData) tvarPtr) != TCL_OK) { ckfree((char *) tvarPtr); return TCL_ERROR; } break; } case TRACE_VDELETE: { int flags; TraceVarInfo *tvarPtr; ClientData clientData; if (objc != 5) { Tcl_WrongNumArgs(interp, 2, objv, "name ops command"); return TCL_ERROR; } flags = 0; rwuOps = Tcl_GetString(objv[3]); for (p = rwuOps; *p != 0; p++) { if (*p == 'r') { flags |= TCL_TRACE_READS; } else if (*p == 'w') { flags |= TCL_TRACE_WRITES; } else if (*p == 'u') { flags |= TCL_TRACE_UNSETS; } else { goto badOps; } } if (flags == 0) { goto badOps; } /* * Search through all of our traces on this variable to * see if there's one with the given command. If so, then * delete the first one that matches. */ command = Tcl_GetStringFromObj(objv[4], &commandLength); length = (size_t) commandLength; clientData = 0; name = Tcl_GetString(objv[2]); while ((clientData = Tcl_VarTraceInfo(interp, name, 0, TraceVarProc, clientData)) != 0) { tvarPtr = (TraceVarInfo *) clientData; if ((tvarPtr->length == length) && (tvarPtr->flags == flags) && (strncmp(command, tvarPtr->command, (size_t) length) == 0)) { Tcl_UntraceVar(interp, name, flags | TCL_TRACE_UNSETS, TraceVarProc, clientData); if (tvarPtr->errMsg != NULL) { ckfree(tvarPtr->errMsg); } ckfree((char *) tvarPtr); break; } } break; } case TRACE_VINFO: { ClientData clientData; char ops[4]; Tcl_Obj *resultListPtr, *pairObjPtr, *elemObjPtr; if (objc != 3) { Tcl_WrongNumArgs(interp, 2, objv, "name"); return TCL_ERROR; } resultListPtr = Tcl_GetObjResult(interp); clientData = 0; name = Tcl_GetString(objv[2]); while ((clientData = Tcl_VarTraceInfo(interp, name, 0, TraceVarProc, clientData)) != 0) { TraceVarInfo *tvarPtr = (TraceVarInfo *) clientData; pairObjPtr = Tcl_NewListObj(0, (Tcl_Obj **) NULL); p = ops; if (tvarPtr->flags & TCL_TRACE_READS) { *p = 'r'; p++; } if (tvarPtr->flags & TCL_TRACE_WRITES) { *p = 'w'; p++; } if (tvarPtr->flags & TCL_TRACE_UNSETS) { *p = 'u'; p++; } *p = '\0'; /* * Build a pair (2-item list) with the ops string as * the first obj element and the tvarPtr->command string * as the second obj element. Append the pair (as an * element) to the end of the result object list. */ elemObjPtr = Tcl_NewStringObj(ops, -1); Tcl_ListObjAppendElement(NULL, pairObjPtr, elemObjPtr); elemObjPtr = Tcl_NewStringObj(tvarPtr->command, -1); Tcl_ListObjAppendElement(NULL, pairObjPtr, elemObjPtr); Tcl_ListObjAppendElement(interp, resultListPtr, pairObjPtr); } Tcl_SetObjResult(interp, resultListPtr); break; } default: { panic("Tcl_TraceObjCmd: bad option index to TraceOptions"); } } return TCL_OK; badOps: Tcl_AppendResult(interp, "bad operations \"", rwuOps, "\": should be one or more of rwu", (char *) NULL); return TCL_ERROR; } /* *---------------------------------------------------------------------- * * TraceVarProc -- * * This procedure is called to handle variable accesses that have * been traced using the "trace" command. * * Results: * Normally returns NULL. If the trace command returns an error, * then this procedure returns an error string. * * Side effects: * Depends on the command associated with the trace. * *---------------------------------------------------------------------- */ /* ARGSUSED */ static char * TraceVarProc(clientData, interp, name1, name2, flags) ClientData clientData; /* Information about the variable trace. */ Tcl_Interp *interp; /* Interpreter containing variable. */ char *name1; /* Name of variable or array. */ char *name2; /* Name of element within array; NULL means * scalar variable is being referenced. */ int flags; /* OR-ed bits giving operation and other * information. */ { Tcl_SavedResult state; TraceVarInfo *tvarPtr = (TraceVarInfo *) clientData; char *result; int code; Tcl_DString cmd; result = NULL; if (tvarPtr->errMsg != NULL) { ckfree(tvarPtr->errMsg); tvarPtr->errMsg = NULL; } if ((tvarPtr->flags & flags) && !(flags & TCL_INTERP_DESTROYED)) { /* * Generate a command to execute by appending list elements * for the two variable names and the operation. The five * extra characters are for three space, the opcode character, * and the terminating null. */ if (name2 == NULL) { name2 = ""; } Tcl_DStringInit(&cmd); Tcl_DStringAppend(&cmd, tvarPtr->command, (int) tvarPtr->length); Tcl_DStringAppendElement(&cmd, name1); Tcl_DStringAppendElement(&cmd, name2); if (flags & TCL_TRACE_READS) { Tcl_DStringAppend(&cmd, " r", 2); } else if (flags & TCL_TRACE_WRITES) { Tcl_DStringAppend(&cmd, " w", 2); } else if (flags & TCL_TRACE_UNSETS) { Tcl_DStringAppend(&cmd, " u", 2); } /* * Execute the command. Save the interp's result used for * the command. We discard any object result the command returns. */ Tcl_SaveResult(interp, &state); code = Tcl_Eval(interp, Tcl_DStringValue(&cmd)); if (code != TCL_OK) { /* copy error msg to result */ char *string; int length; string = Tcl_GetStringFromObj(Tcl_GetObjResult(interp), &length); tvarPtr->errMsg = (char *) ckalloc((unsigned) (length + 1)); memcpy(tvarPtr->errMsg, string, (size_t) (length + 1)); result = tvarPtr->errMsg; } Tcl_RestoreResult(interp, &state); Tcl_DStringFree(&cmd); } if (flags & TCL_TRACE_DESTROYED) { result = NULL; if (tvarPtr->errMsg != NULL) { ckfree(tvarPtr->errMsg); } ckfree((char *) tvarPtr); } return result; } /* *---------------------------------------------------------------------- * * Tcl_WhileObjCmd -- * * This procedure is invoked to process the "while" Tcl command. * See the user documentation for details on what it does. * * With the bytecode compiler, this procedure is only called when * a command name is computed at runtime, and is "while" or the name * to which "while" was renamed: e.g., "set z while; $z {$i<100} {}" * * Results: * A standard Tcl result. * * Side effects: * See the user documentation. * *---------------------------------------------------------------------- */ /* ARGSUSED */ int Tcl_WhileObjCmd(dummy, interp, objc, objv) ClientData dummy; /* Not used. */ Tcl_Interp *interp; /* Current interpreter. */ int objc; /* Number of arguments. */ Tcl_Obj *CONST objv[]; /* Argument objects. */ { int result, value; if (objc != 3) { Tcl_WrongNumArgs(interp, 1, objv, "test command"); return TCL_ERROR; } while (1) { result = Tcl_ExprBooleanObj(interp, objv[1], &value); if (result != TCL_OK) { return result; } if (!value) { break; } result = Tcl_EvalObjEx(interp, objv[2], 0); if ((result != TCL_OK) && (result != TCL_CONTINUE)) { if (result == TCL_ERROR) { char msg[32 + TCL_INTEGER_SPACE]; sprintf(msg, "\n (\"while\" body line %d)", interp->errorLine); Tcl_AddErrorInfo(interp, msg); } break; } } if (result == TCL_BREAK) { result = TCL_OK; } if (result == TCL_OK) { Tcl_ResetResult(interp); } return result; }