Tcl Source Code

Check-in [11e89c25be]
Login
Bounty program for improvements to Tcl and certain Tcl packages.
Tcl 2019 Conference, Houston/TX, US, Nov 4-8
Send your abstracts to [email protected]
or submit via the online form by Sep 9.

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview
Comment:Rewrite the [string replace] compiler to take advantage of the richer set of encoded index values.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | core-8-6-branch
Files: files | file ages | folders
SHA3-256: 11e89c25be7e2227c99e48ffe7c225c01eb536b3c7fb8ddc178ec331df212705
User & Date: dgp 2018-03-13 16:03:39
Context
2018-03-13
20:46
Audit and correct Object reference counting issues. check-in: ae1f9d5726 user: pooryorick tags: core-8-6-branch
17:14
merge 8.6 check-in: 1a04240e68 user: dgp tags: core-8-branch
16:03
Rewrite the [string replace] compiler to take advantage of the richer set of encoded index values. check-in: 11e89c25be user: dgp tags: core-8-6-branch
04:01
Repair the INST_STR_REPLACE instruction. check-in: 3101f3fd20 user: dgp tags: core-8-6-branch
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to generic/tclCompCmdsSZ.c.

991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004


1005
1006
1007
1008
1009
1010

1011
1012

1013
1014
1015
1016
1017

1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037




1038
1039
1040
1041
1042
1043



1044
1045


1046
1047
1048
1049


1050
1051






1052

1053







1054









1055
1056
1057
1058



1059
1060
1061
1062
1063

1064
1065


1066
1067
1068

































1069
1070
1071
1072
1073
1074
1075
1076
1077
1078




1079










1080


1081
1082
1083
1084
1085
1086
1087



1088
1089



1090
1091
1092
1093
1094
1095







1096
1097
1098
1099







1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113

1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125

1126
1127
1128
1129
1130
1131
1132

1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
    Tcl_Interp *interp,		/* Tcl interpreter for context. */
    Tcl_Parse *parsePtr,	/* Points to a parse structure for the
				 * command. */
    Command *cmdPtr,		/* Points to defintion of command being
				 * compiled. */
    CompileEnv *envPtr)		/* Holds the resulting instructions. */
{
    Tcl_Token *tokenPtr, *valueTokenPtr, *replacementTokenPtr = NULL;
    DefineLineInformation;	/* TIP #280 */
    int idx1, idx2;

    if (parsePtr->numWords < 4 || parsePtr->numWords > 5) {
	return TCL_ERROR;
    }


    valueTokenPtr = TokenAfter(parsePtr->tokenPtr);
    if (parsePtr->numWords == 5) {
	tokenPtr = TokenAfter(valueTokenPtr);
	tokenPtr = TokenAfter(tokenPtr);
	replacementTokenPtr = TokenAfter(tokenPtr);
    }

	goto genericReplace;


    tokenPtr = TokenAfter(valueTokenPtr);
    if (TclGetIndexFromToken(tokenPtr, TCL_INDEX_START, TCL_INDEX_AFTER,
	    &idx1) != TCL_OK) {
	goto genericReplace;
    }

    /*
     * Token parsed as an index value. Indices before the string are
     * treated as index of start of string.
     */

    tokenPtr = TokenAfter(tokenPtr);
    if (TclGetIndexFromToken(tokenPtr, TCL_INDEX_BEFORE, TCL_INDEX_END,
	    &idx2) != TCL_OK) {
	goto genericReplace;
    }
    /*
     * Token parsed as an index value. Indices after the string are
     * treated as index of end of string.
     */

/* TODO...... */
    /*
     * We handle these replacements specially: first character (where
     * idx1=idx2=0) and last character (where idx1=idx2=TCL_INDEX_END). Anything
     * else and the semantics get rather screwy.




     *
     * TODO: These seem to be very narrow cases.  They are not even
     * covered by the test suite, and any programming that ends up
     * here could have been coded by the programmer using [string range]
     * and [string cat]. [*]  Not clear at all to me that the bytecode
     * generated here is worthwhile.



     *
     *  [*] Except for the empty string exceptions.  UGGGGHHHH.


     */

    if (idx1 == 0 && idx2 == 0) {
	int notEq, end;



	/*






	 * Just working with the first character.

	 */

















	CompileWord(envPtr, valueTokenPtr, interp, 1);
	if (replacementTokenPtr == NULL) {
	    /* Drop first */
	    OP44(	STR_RANGE_IMM, 1, TCL_INDEX_END);



	    return TCL_OK;
	}
	/* Replace first */
	CompileWord(envPtr, replacementTokenPtr, interp, 4);


	/*
	 * NOTE: The following tower of bullshit is present because


	 * [string replace] was boneheadedly defined not to replace
	 * empty strings, so we actually have to detect the empty
	 * string case and treat it differently.

































	 */

	OP4(		OVER, 1);
	PUSH(		"");
	OP(		STR_EQ);
	JUMP1(		JUMP_FALSE, notEq);
	OP(		POP);
	JUMP1(		JUMP, end);
	FIXJUMP1(notEq);
	TclAdjustStackDepth(1, envPtr);




	OP4(		REVERSE, 2);










	OP44(		STR_RANGE_IMM, 1, TCL_INDEX_END);


	OP1(		STR_CONCAT1, 2);
	FIXJUMP1(end);
	return TCL_OK;

    } else if (idx1 == TCL_INDEX_END && idx2 == TCL_INDEX_END) {
	int notEq, end;




	/*
	 * Just working with the last character.



	 */

	CompileWord(envPtr, valueTokenPtr, interp, 1);
	if (replacementTokenPtr == NULL) {
	    /* Drop last */
	    OP44(	STR_RANGE_IMM, 0, TCL_INDEX_END-1);







	    return TCL_OK;
	}
	/* Replace last */
	CompileWord(envPtr, replacementTokenPtr, interp, 4);








	/* More bullshit; see NOTE above. */

	OP4(		OVER, 1);
	PUSH(		"");
	OP(		STR_EQ);
	JUMP1(		JUMP_FALSE, notEq);
	OP(		POP);
	JUMP1(		JUMP, end);
	FIXJUMP1(notEq);
	TclAdjustStackDepth(1, envPtr);
	OP4(		REVERSE, 2);
	OP44(		STR_RANGE_IMM, 0, TCL_INDEX_END-1);
	OP4(		REVERSE, 2);

	OP1(		STR_CONCAT1, 2);
	FIXJUMP1(end);
	return TCL_OK;

    } else {
	/*
	 * Need to process indices at runtime. This could be because the
	 * indices are not constants, or because we need to resolve them to
	 * absolute indices to work out if a replacement is going to happen.
	 * In any case, to runtime it is.
	 */


    genericReplace:
	CompileWord(envPtr, valueTokenPtr, interp, 1);
	tokenPtr = TokenAfter(valueTokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 2);
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 3);
	if (replacementTokenPtr != NULL) {

	    CompileWord(envPtr, replacementTokenPtr, interp, 4);
	} else {
	    PUSH(	"");
	}
	OP(		STR_REPLACE);
	return TCL_OK;
    }
}

int
TclCompileStringTrimLCmd(
    Tcl_Interp *interp,		/* Used for error reporting. */
    Tcl_Parse *parsePtr,	/* Points to a parse structure for the command
				 * created by Tcl_ParseCommand. */






|

|




>
>

<
<
|
<
|
>
|
<
>

|
|


>

<
|

<

|
|


<
<
<
<

<
|
<
<
<
>
>
>
>

<
<
<
<
<
>
>
>

<
>
>


<
<
>
>


>
>
>
>
>
>
|
>

>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
|
<
<
<
>
>
>
|
|
<
<

>
|
<
>
>
|
<
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|

<
<
<
<
<
<
<
<
>
>
>
>

>
>
>
>
>
>
>
>
>
>
|
>
>

<

|
<
<

>
>
>
|
<
>
>
>


<
<
<
<
>
>
>
>
>
>
>
|
|
<
<
>
>
>
>
>
>
>
|
<
<
<
<
<
<
|
<
<
<
<
|
|
>
|
<
|
|
<
<
<
<
<
<
<
|
>

<




|
>
|





<







991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007


1008

1009
1010
1011

1012
1013
1014
1015
1016
1017
1018
1019

1020
1021

1022
1023
1024
1025
1026




1027

1028



1029
1030
1031
1032
1033





1034
1035
1036
1037

1038
1039
1040
1041


1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072



1073
1074
1075
1076
1077


1078
1079
1080

1081
1082
1083


1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118








1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137

1138
1139


1140
1141
1142
1143
1144

1145
1146
1147
1148
1149




1150
1151
1152
1153
1154
1155
1156
1157
1158


1159
1160
1161
1162
1163
1164
1165
1166






1167




1168
1169
1170
1171

1172
1173







1174
1175
1176

1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188

1189
1190
1191
1192
1193
1194
1195
    Tcl_Interp *interp,		/* Tcl interpreter for context. */
    Tcl_Parse *parsePtr,	/* Points to a parse structure for the
				 * command. */
    Command *cmdPtr,		/* Points to defintion of command being
				 * compiled. */
    CompileEnv *envPtr)		/* Holds the resulting instructions. */
{
    Tcl_Token *tokenPtr, *valueTokenPtr;
    DefineLineInformation;	/* TIP #280 */
    int first, last;

    if (parsePtr->numWords < 4 || parsePtr->numWords > 5) {
	return TCL_ERROR;
    }
 
    /* Bytecode to compute/push string argument being replaced */
    valueTokenPtr = TokenAfter(parsePtr->tokenPtr);


    CompileWord(envPtr, valueTokenPtr, interp, 1);


    /*
     * Check for first index known and useful at compile time. 

     */
    tokenPtr = TokenAfter(valueTokenPtr);
    if (TclGetIndexFromToken(tokenPtr, TCL_INDEX_BEFORE, TCL_INDEX_AFTER,
	    &first) != TCL_OK) {
	goto genericReplace;
    }

    /*

     * Check for last index known and useful at compile time. 
     */

    tokenPtr = TokenAfter(tokenPtr);
    if (TclGetIndexFromToken(tokenPtr, TCL_INDEX_BEFORE, TCL_INDEX_AFTER,
	    &last) != TCL_OK) {
	goto genericReplace;
    }






    /* 



     * [string replace] is an odd bird.  For many arguments it is
     * a conventional substring replacer.  However it also goes out
     * of its way to become a no-op for many cases where it would be
     * replacing an empty substring.  Precisely, it is a no-op when
     *





     *		(last < first)		OR
     *		(last < 0)		OR
     *		(end < first)
     *

     * For some compile-time values we can detect these cases, and
     * compile direct to bytecode implementing the no-op.
     */



    if ((last == TCL_INDEX_BEFORE)		/* Know (last < 0) */
	    || (first == TCL_INDEX_AFTER)	/* Know (first > end) */

	/*
	 * Tricky to determine when runtime (last < first) can be
	 * certainly known based on the encoded values. Consider the
	 * cases...
	 *
	 * (first <= TCL_INDEX_END) &&
	 *	(last == TCL_INDEX_AFTER) => cannot tell REJECT
	 *	(last <= TCL_INDEX END) && (last < first) => ACCEPT
	 *	else => cannot tell REJECT
	 */
	    || ((first <= TCL_INDEX_END) && (last <= TCL_INDEX_END)
		&& (last < first))		/* Know (last < first) */
	/*
	 * (first == TCL_INDEX_BEFORE) &&
	 *	(last == TCL_INDEX_AFTER) => (first < last) REJECT
	 *	(last <= TCL_INDEX_END) => cannot tell REJECT
	 *	else		=> (first < last) REJECT
	 *
	 * else [[first >= TCL_INDEX_START]] &&
	 *	(last == TCL_INDEX_AFTER) => cannot tell REJECT
	 *	(last <= TCL_INDEX_END) => cannot tell REJECT
	 *	else [[last >= TCL_INDEX START]] && (last < first) => ACCEPT
	 */
	    || ((first >= TCL_INDEX_START) && (last >= TCL_INDEX_START)
		&& (last < first))) {		/* Know (last < first) */
	if (parsePtr->numWords == 5) {
	    tokenPtr = TokenAfter(tokenPtr);
	    CompileWord(envPtr, tokenPtr, interp, 4);



	    OP(		POP);		/* Pop newString */
	}
	/* Original string argument now on TOS as result */
	return TCL_OK;
    }



    if (parsePtr->numWords == 5) {
    /*

     * When we have a string replacement, we have to take care about
     * not replacing empty substrings that [string replace] promises
     * not to replace


     *
     * The remaining index values might be suitable for conventional
     * string replacement, but only if they cannot possibly meet the
     * conditions described above at runtime. If there's a chance they
     * might, we would have to emit bytecode to check and at that point
     * we're paying more in bytecode execution time than would make
     * things worthwhile. Trouble is we are very limited in
     * how much we can detect that at compile time. After decoding,
     * we need, first:
     *
     *		(first <= end)
     *
     * The encoded indices (first <= TCL_INDEX END) and
     * (first == TCL_INDEX_BEFORE) always meets this condition, but
     * any other encoded first index has some list for which it fails.
     *
     * We also need, second:
     *
     *		(last >= 0)
     *
     * The encoded indices (last >= TCL_INDEX_START) and
     * (last == TCL_INDEX_AFTER) always meet this condition but any
     * other encoded last index has some list for which it fails.
     *
     * Finally we need, third:
     *
     *		(first <= last)
     * 
     * Considered in combination with the constraints we already have,
     * we see that we can proceed when (first == TCL_INDEX_BEFORE)
     * or (last == TCL_INDEX_AFTER). These also permit simplification
     * of the prefix|replace|suffix construction. The other constraints,
     * though, interfere with getting a guarantee that first <= last. 
     */









    if ((first == TCL_INDEX_BEFORE) && (last >= TCL_INDEX_START)) {
	/* empty prefix */
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 4);
	OP4(		REVERSE, 2);
	if (last == TCL_INDEX_AFTER) {
	    OP(		POP);		/* Pop  original */
	} else {
	    OP44(	STR_RANGE_IMM, last + 1, TCL_INDEX_END);
	    OP1(	STR_CONCAT1, 2);
	}
	return TCL_OK;
    }

    if ((last == TCL_INDEX_AFTER) && (first <= TCL_INDEX_END)) {
	OP44(		STR_RANGE_IMM, 0, first-1);
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 4);
	OP1(		STR_CONCAT1, 2);

	return TCL_OK;
    }



	/* FLOW THROUGH TO genericReplace */

    } else {
	/* 

	 * When we have no replacement string to worry about, we may
	 * have more luck, because the forbidden empty string replacements
	 * are harmless when they are replaced by another empty string.
	 */





	if ((first == TCL_INDEX_BEFORE) || (first == TCL_INDEX_START)) {
	    /* empty prefix - build suffix only */

	    if ((last == TCL_INDEX_END) || (last == TCL_INDEX_AFTER)) {
		/* empty suffix too => empty result */
		OP(	POP);		/* Pop  original */
		PUSH	(	"");
		return TCL_OK;
	    }


	    OP44(	STR_RANGE_IMM, last + 1, TCL_INDEX_END);
	    return TCL_OK;
	} else {
	    if ((last == TCL_INDEX_END) || (last == TCL_INDEX_AFTER)) {
		/* empty suffix - build prefix only */
		OP44(	STR_RANGE_IMM, 0, first-1);
		return TCL_OK;
	    }






	    OP(		DUP);




	    OP44(	STR_RANGE_IMM, 0, first-1);
	    OP4(	REVERSE, 2);
	    OP44(	STR_RANGE_IMM, last + 1, TCL_INDEX_END);
	    OP1(	STR_CONCAT1, 2);

	    return TCL_OK;
	}







    }

    genericReplace:

	tokenPtr = TokenAfter(valueTokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 2);
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 3);
	if (parsePtr->numWords == 5) {
	    tokenPtr = TokenAfter(tokenPtr);
	    CompileWord(envPtr, tokenPtr, interp, 4);
	} else {
	    PUSH(	"");
	}
	OP(		STR_REPLACE);
	return TCL_OK;

}

int
TclCompileStringTrimLCmd(
    Tcl_Interp *interp,		/* Used for error reporting. */
    Tcl_Parse *parsePtr,	/* Points to a parse structure for the command
				 * created by Tcl_ParseCommand. */