tinymux/parser/eval.cpp

1176 lines
46 KiB
C++
Raw Permalink Normal View History

/*
* eval.cpp - MUX expression AST evaluator study tool.
*
* Stage 3 of the parser study: walk the AST and evaluate it, producing
* output equivalent to mux_exec for the subset of expressions that
* don't require database access.
*
* Two-tier function dispatch:
* - Normal functions: arguments pre-evaluated, handler gets strings
* - FN_NOEVAL functions: arguments NOT pre-evaluated, handler gets
* AST subtrees and calls eval() selectively
*
* This demonstrates that MUX softcode CAN be evaluated from an AST
* with proper deferred evaluation for control flow and iteration.
*/
#include "mux_parse.h"
#include <cstdlib>
#include <cmath>
#include <map>
#include <functional>
#include <algorithm>
// ---------------------------------------------------------------
// Evaluation context
// ---------------------------------------------------------------
// Eval flags — mirror the defines in mux/src/externs.h.
// These control what the evaluator does with each node type.
//
enum {
EV_EVAL = 0x0004, // Evaluate %-substitutions
EV_STRIP_CURLY = 0x0008, // Strip outer {} from args
EV_FCHECK = 0x0010, // Check for function calls on (
EV_FMAND = 0x0020, // Require valid function name
EV_NOFCHECK = 0x0040, // Suppress [ evaluation
EV_NO_COMPRESS = 0x0080, // Don't compress spaces
EV_STRIP_LS = 0x1000, // Strip leading spaces
EV_STRIP_TS = 0x2000, // Strip trailing spaces
EV_TOP = 0x0800, // Top-level evaluation
};
// Default eval flags for top-level evaluation.
static constexpr int EV_DEFAULT = EV_EVAL | EV_FCHECK | EV_TOP;
struct EvalContext {
// Registers %q0-%q9, %qa-%qz, and named %q<name>
std::map<std::string, std::string> registers;
// Command arguments %0-%9
std::string args[10];
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
int nargs = 2; // number of valid args
// Iterator state for iter/list
struct IterFrame {
std::string itext; // current item (itext(n))
int inum; // current index (inum(n), 1-based)
};
std::vector<IterFrame> iterStack;
// Special substitutions
std::string enactorName; // %n
std::string enactorDbref; // %#
std::string executorDbref; // %!
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
std::string callerDbref; // %@
std::string enactorObjid; // %: (dbref:creation_time)
std::string lastCommand; // %m
std::string moniker; // %k
std::string pipedOutput; // %|
std::string switchToken; // #$
// Pronoun substitutions — subjective, possessive, objective, abs-possessive
std::string pronSub; // %s (he/she/they)
std::string pronPos; // %p (his/her/their)
std::string pronObj; // %o (him/her/them)
std::string pronAbs; // %a (his/hers/theirs)
// Variable attributes %va%vz (26 slots)
std::string varAttrs[26];
// Penn-specific
std::string rawCommand; // Penn %c — raw command line
std::string evaledCommand; // Penn %u — evaluated command line
std::string accentedName; // Penn %~ — accented name
std::string attrName; // Penn %= — current attribute name
// Current eval flags
int evalFlags = EV_DEFAULT;
// Parser/evaluator compatibility profile.
ParserProfile profile = PROFILE_MUX214_AST;
};
// ---------------------------------------------------------------
// Evaluator
// ---------------------------------------------------------------
class Evaluator {
public:
Evaluator(EvalContext &ctx) : m_ctx(ctx) {
registerBuiltins();
}
std::string eval(const ASTNode *node) {
return evalWithFlags(node, m_ctx.evalFlags);
}
// Evaluate with specific flags (used for recursive contexts).
//
std::string evalWithFlags(const ASTNode *node, int flags) {
if (!node) return "";
// Save and restore flags for this scope.
int savedFlags = m_ctx.evalFlags;
m_ctx.evalFlags = flags;
std::string result;
switch (node->type) {
case NODE_SEQUENCE:
result = evalSequence(node);
break;
case NODE_LITERAL:
case NODE_SPACE:
result = node->text;
break;
case NODE_SUBST:
// Only resolve substitutions if EV_EVAL is set.
result = (flags & EV_EVAL) ? evalSubst(node) : node->text;
break;
case NODE_ESCAPE:
result = (flags & EV_EVAL) ? evalEscape(node) : node->text;
break;
case NODE_FUNCCALL:
// Only dispatch functions if EV_FCHECK is set.
if (flags & EV_FCHECK) {
result = evalFuncCall(node);
} else {
// No function checking — treat as literal text + parens.
result = node->text + "(";
for (size_t i = 0; i < node->children.size(); i++) {
if (i > 0) result += ",";
result += evalWithFlags(node->children[i].get(), flags);
}
result += ")";
}
break;
case NODE_DYNCALL:
result = "#-1 DYNAMIC CALL NOT SUPPORTED";
break;
case NODE_EVALBRACKET:
// Only recurse into [...] if EV_NOFCHECK is NOT set.
if (!(flags & EV_NOFCHECK)) {
result = evalEvalBracket(node);
} else {
// Pass through as literal brackets.
result = "[";
if (!node->children.empty())
result += evalWithFlags(node->children[0].get(), flags);
result += "]";
}
break;
case NODE_BRACEGROUP:
result = evalBraceGroup(node);
break;
case NODE_SEMICOLON:
result = "";
break;
}
m_ctx.evalFlags = savedFlags;
return result;
}
private:
EvalContext &m_ctx;
// Two dispatch tables:
// - m_funcs: normal functions, receive pre-evaluated string args
// - m_noeval_funcs: FN_NOEVAL functions, receive unevaluated AST
// children and call eval() selectively (deferred evaluation)
//
using FuncHandler = std::function<std::string(const std::vector<std::string>&)>;
std::map<std::string, FuncHandler> m_funcs;
using NoevalHandler = std::function<std::string(const std::vector<std::unique_ptr<ASTNode>>&)>;
std::map<std::string, NoevalHandler> m_noeval_funcs;
static std::string toUpper(const std::string &s) {
std::string r = s;
for (auto &c : r) c = static_cast<char>(toupper(static_cast<unsigned char>(c)));
return r;
}
static long toLong(const std::string &s) {
if (s.empty()) return 0;
return strtol(s.c_str(), nullptr, 10);
}
static double toDouble(const std::string &s) {
if (s.empty()) return 0.0;
return strtod(s.c_str(), nullptr);
}
static bool toBool(const std::string &s) {
if (s.empty()) return false;
char *end;
double v = strtod(s.c_str(), &end);
if (end != s.c_str()) return v != 0.0;
return true;
}
static std::string fmtNum(double v) {
if (v == floor(v) && fabs(v) < 1e15) {
return std::to_string(static_cast<long long>(v));
}
char buf[64];
snprintf(buf, sizeof(buf), "%g", v);
return buf;
}
static std::vector<std::string> splitList(const std::string &s,
const std::string &sep = " ")
{
std::vector<std::string> result;
if (s.empty()) return result;
if (sep == " ") {
size_t i = 0;
while (i < s.size() && s[i] == ' ') i++;
while (i < s.size()) {
size_t j = i;
while (j < s.size() && s[j] != ' ') j++;
result.push_back(s.substr(i, j - i));
while (j < s.size() && s[j] == ' ') j++;
i = j;
}
} else {
size_t start = 0;
size_t pos;
while ((pos = s.find(sep, start)) != std::string::npos) {
result.push_back(s.substr(start, pos - start));
start = pos + sep.size();
}
result.push_back(s.substr(start));
}
return result;
}
std::string evalSequence(const ASTNode *node) {
std::string result;
int flags = m_ctx.evalFlags;
bool bFCheckPending = (flags & EV_FCHECK) != 0 && (flags & EV_FMAND) == 0;
for (size_t i = 0; i < node->children.size(); i++) {
const auto &child = node->children[i];
if (child->type == NODE_ESCAPE
&& child->text == "\\\\"
&& i + 1 < node->children.size()
&& node->children[i + 1]->type == NODE_SUBST) {
result += node->children[i + 1]->text;
i++;
if (bFCheckPending) {
bFCheckPending = false;
}
continue;
}
int childFlags = flags;
if (bFCheckPending) {
if (child->type != NODE_FUNCCALL) {
childFlags &= ~EV_FCHECK;
}
if (child->type != NODE_SPACE) {
bFCheckPending = false;
}
} else if ((flags & EV_FCHECK) != 0 && (flags & EV_FMAND) == 0) {
childFlags &= ~EV_FCHECK;
}
result += evalWithFlags(child.get(), childFlags);
}
return result;
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// Apply first-character uppercasing when the substitution letter
// was uppercase (%N, %S, %P, %O, %A, %K, %M, %Q, %V).
//
static std::string maybeCapitalize(const std::string &s, char origLetter) {
if (s.empty()) return s;
if (isupper(static_cast<unsigned char>(origLetter))) {
std::string r = s;
r[0] = static_cast<char>(toupper(static_cast<unsigned char>(r[0])));
return r;
}
return s;
}
std::string evalSubst(const ASTNode *node) {
const std::string &sub = node->text;
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// Hash forms: ##, #@, #$
if (sub.size() == 2 && sub[0] == '#') {
switch (sub[1]) {
case '#': {
// ## — same as %i0 (current loop item)
int idx = static_cast<int>(m_ctx.iterStack.size()) - 1;
return (idx >= 0) ? m_ctx.iterStack[idx].itext : "";
}
case '@': {
// #@ — same as inum(0) (current loop position)
int idx = static_cast<int>(m_ctx.iterStack.size()) - 1;
return (idx >= 0) ? std::to_string(m_ctx.iterStack[idx].inum) : "";
}
case '$':
// #$ — switch matched value
return m_ctx.switchToken;
}
}
if (sub.size() < 2) return "%";
char ch = sub[1];
char upper = static_cast<char>(toupper(static_cast<unsigned char>(ch)));
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// %0%9: command arguments
if (ch >= '0' && ch <= '9') {
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
int idx = ch - '0';
return (idx < m_ctx.nargs) ? m_ctx.args[idx] : "";
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// %q/%Q: registers
if (upper == 'Q') {
std::string regname;
if (sub.size() >= 4 && sub[2] == '<') {
regname = sub.substr(3, sub.size() - 4);
} else if (sub.size() >= 3) {
regname = std::string(1, sub[2]);
}
auto it = m_ctx.registers.find(regname);
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
std::string val = (it != m_ctx.registers.end()) ? it->second : "";
return maybeCapitalize(val, ch);
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// %v/%V: variable attributes AZ
if (upper == 'V' && sub.size() >= 3) {
char attrCh = static_cast<char>(toupper(static_cast<unsigned char>(sub[2])));
if (attrCh >= 'A' && attrCh <= 'Z') {
std::string val = m_ctx.varAttrs[attrCh - 'A'];
return maybeCapitalize(val, ch);
}
return "";
}
// %w/%W: Penn W-attributes (profile-dependent)
if (upper == 'W' && sub.size() >= 3) {
if (m_ctx.profile == PROFILE_PENN) {
return "";
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
}
2026-03-21 04:28:44 -06:00
// MUX: unknown → literal. In MUX profiles the tokenizer
// should not have consumed the trailing attribute letter.
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
return std::string(1, ch);
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// %c/%C, %x/%X: color codes (MUX) or command/X-attr (Penn)
if (upper == 'C' || upper == 'X') {
if (m_ctx.profile == PROFILE_PENN) {
if (upper == 'C') {
// Penn: %c = raw command line
return m_ctx.rawCommand;
}
// Penn: %x = X-attribute (if sub.size() >= 3)
if (sub.size() >= 3) {
return "thing";
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
}
return "";
}
// MUX: color codes — simplified for study tool
if (sub.size() >= 3) {
return "[color:" + sub.substr(1) + "]";
}
return "";
}
// %i/%I: loop itext at depth
if (upper == 'I' && sub.size() == 2 && m_ctx.profile == PROFILE_MUX214_AST) {
return "";
}
if (upper == 'I' && sub.size() >= 3) {
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
char depthCh = sub[2];
if (depthCh >= '0' && depthCh <= '9') {
int depth = depthCh - '0';
int idx = static_cast<int>(m_ctx.iterStack.size()) - 1 - depth;
if (idx >= 0 && idx < static_cast<int>(m_ctx.iterStack.size())) {
return m_ctx.iterStack[idx].itext;
}
return "";
}
if (toupper(static_cast<unsigned char>(depthCh)) == 'L'
&& m_ctx.profile == PROFILE_PENN) {
if (m_ctx.iterStack.empty()) {
return "#-1 ARGUMENT OUT OF RANGE";
}
return m_ctx.iterStack.back().itext;
}
return "";
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// %=: attribute name (Penn) or extended arg/attr lookup (MUX)
if (ch == '=') {
if (sub.size() == 2) {
// Bare %=
if (m_ctx.profile == PROFILE_PENN) {
return m_ctx.attrName;
}
return "=";
}
// %=<...> — extended form (MUX only)
if (sub.size() >= 4 && sub[2] == '<') {
std::string inner = sub.substr(3, sub.size() - 4);
// Check if numeric → extended arg
bool isNum = !inner.empty();
for (char c : inner) {
if (!isdigit(static_cast<unsigned char>(c))) {
isNum = false;
break;
}
}
if (isNum) {
int idx = atoi(inner.c_str());
return (idx < m_ctx.nargs) ? m_ctx.args[idx] : "";
}
return "[attr:" + inner + "]";
}
return "=";
}
// %$: Penn stack variables
if (ch == '$' && m_ctx.profile == PROFILE_PENN) {
return "[stack]";
}
// Simple single-character forms
switch (upper) {
case 'R':
// MUX: \r\n. Penn: \n only.
return (m_ctx.profile == PROFILE_PENN) ? "\n" : "\r\n";
case 'B': return " ";
case 'T': return "\t";
case 'N': return maybeCapitalize(m_ctx.enactorName, ch);
case 'S': return maybeCapitalize(m_ctx.pronSub, ch);
case 'P': return maybeCapitalize(m_ctx.pronPos, ch);
case 'O': return maybeCapitalize(m_ctx.pronObj, ch);
case 'A': return maybeCapitalize(m_ctx.pronAbs, ch);
case 'K': return maybeCapitalize(m_ctx.moniker, ch);
case 'M': return maybeCapitalize(m_ctx.lastCommand, ch);
case 'L': return "[location]";
}
switch (ch) {
case '%': return "%";
case '#': return m_ctx.enactorDbref;
case '!': return m_ctx.executorDbref;
case '@': return m_ctx.callerDbref;
case ':': return m_ctx.enactorObjid;
case '+': return std::to_string(m_ctx.nargs);
case '|': return m_ctx.pipedOutput;
}
// Penn-specific single-char forms
if (m_ctx.profile == PROFILE_PENN) {
if (ch == ' ') return "% ";
if (ch == '~') return m_ctx.accentedName;
if (ch == '?') return "[limits]";
if (upper == 'U') return m_ctx.evaledCommand;
}
// Unknown substitution — output the character literally
// (matches iCode == 0 fallback in all engines).
return std::string(1, ch);
}
std::string evalEscape(const ASTNode *node) {
if (node->text.size() >= 2) {
if (node->text == "\\%" && m_ctx.profile != PROFILE_PENN) {
return "";
}
return node->text.substr(1);
}
return "\\";
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// Replicate 2.13's noeval pass on an AST subtree.
//
// In 2.13, mux_exec with EV_EVAL off still processes backslash
// escapes (the handler at line 2439 has no EV_EVAL guard).
// Percent substitutions are copied literally (guarded by EV_EVAL).
//
// This produces a string with one layer of backslash stripping,
// which can then be re-tokenized and evaluated.
//
std::string noevalPass(const ASTNode *node) {
if (!node) return "";
switch (node->type) {
case NODE_LITERAL:
case NODE_SPACE:
case NODE_SEMICOLON:
return node->text;
case NODE_SUBST:
// Percent handler IS guarded by EV_EVAL in 2.13.
// With EV_EVAL off, it copies % and following char literally.
return node->text;
case NODE_ESCAPE:
// Backslash handler is NOT guarded by EV_EVAL in 2.13.
// It unconditionally consumes the \ and emits the next char.
if (node->text.size() >= 2) {
return node->text.substr(1);
}
return "\\";
case NODE_FUNCCALL:
// In noeval, function calls aren't dispatched.
// Copy the name and parens literally.
{
std::string r = node->text + "(";
for (size_t i = 0; i < node->children.size(); i++) {
if (i > 0) r += ",";
r += noevalPass(node->children[i].get());
}
return r + ")";
}
case NODE_EVALBRACKET:
// In noeval with EV_NOFCHECK, [...] is not evaluated.
// Copy brackets and contents literally.
{
std::string r = "[";
for (const auto &c : node->children)
r += noevalPass(c.get());
return r + "]";
}
case NODE_BRACEGROUP:
// Nested brace groups: copy braces and recurse.
{
std::string r = "{";
for (const auto &c : node->children)
r += noevalPass(c.get());
return r + "}";
}
case NODE_SEQUENCE:
{
std::string r;
2026-03-21 12:01:17 -06:00
for (size_t i = 0; i < node->children.size(); i++) {
const auto &c = node->children[i];
2026-04-04 17:48:22 -06:00
if (c->type == NODE_ESCAPE
&& c->text == "\\\\"
&& i + 1 < node->children.size()
&& node->children[i + 1]->type == NODE_SUBST) {
r += node->children[i + 1]->text;
i++;
continue;
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
r += noevalPass(c.get());
2026-03-21 12:01:17 -06:00
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
return r;
}
case NODE_DYNCALL:
return "#-1 DYNAMIC";
}
return "";
}
// Evaluate a selected argument from a FN_NOEVAL function using the
// 2.13-style noeval pass followed by reparse/re-eval.
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
//
// This replicates the two-pass behavior observed in 2.13:
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
// Pass 1: noeval — strip one layer of backslash (noevalPass)
// Pass 2: eval — re-tokenize the result and evaluate it
//
std::string evalNoevalLegacyArg(const ASTNode *node) {
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
if (!node) return "";
const ASTNode *inner = node;
if (node->type == NODE_BRACEGROUP && !node->children.empty()) {
inner = node->children[0].get();
}
// Pass 1: produce text with one backslash layer stripped.
std::string text = noevalPass(inner);
// Pass 2: re-tokenize and evaluate.
auto tokens = tokenize(text.c_str(), m_ctx.profile);
Parser parser(tokens);
auto ast = parser.parse();
int flags = (m_ctx.evalFlags & ~(EV_TOP | EV_FMAND))
| EV_EVAL | EV_FCHECK;
return evalWithFlags(ast.get(), flags);
}
// Evaluate an argument to a FN_NOEVAL function.
//
std::string evalNoevalArg(const ASTNode *node) {
if (!node) return "";
if (m_ctx.profile == PROFILE_MUX213_COMPAT) {
return evalNoevalLegacyArg(node);
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
if (node->type == NODE_BRACEGROUP) {
2026-03-21 04:28:44 -06:00
// Baseline AST/Penn study behavior: brace groups selected by
// FN_NOEVAL functions are evaluated with brace stripping.
2026-03-21 04:28:44 -06:00
int flags = (m_ctx.evalFlags & ~(EV_TOP | EV_FMAND))
| EV_EVAL | EV_FCHECK | EV_STRIP_CURLY;
return evalWithFlags(node, flags);
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
}
return eval(node);
}
std::string evalFuncCall(const ASTNode *node) {
std::string fname = toUpper(node->text);
// Check FN_NOEVAL first — deferred evaluation.
auto nit = m_noeval_funcs.find(fname);
if (nit != m_noeval_funcs.end()) {
return nit->second(node->children);
}
// Normal — evaluate all arguments first.
auto it = m_funcs.find(fname);
if (it == m_funcs.end()) {
return "#-1 FUNCTION (" + fname + ") NOT FOUND";
}
std::vector<std::string> args;
int flags = (m_ctx.evalFlags & ~(EV_TOP | EV_FMAND)) | EV_EVAL | EV_FCHECK;
for (const auto &child : node->children) {
args.push_back(evalWithFlags(child.get(), flags));
}
return it->second(args);
}
std::string evalEvalBracket(const ASTNode *node) {
if (node->children.empty()) return "";
// Inside [...], functions are checked and mandatory.
// This matches mux_exec: eval | EV_FCHECK | EV_FMAND | EV_EVAL
int flags = (m_ctx.evalFlags & ~EV_TOP) | EV_EVAL | EV_FCHECK | EV_FMAND;
return evalWithFlags(node->children[0].get(), flags);
}
std::string evalBraceGroup(const ASTNode *node) {
if (node->children.empty()) return "";
int flags = m_ctx.evalFlags;
if (flags & EV_STRIP_CURLY) {
// Strip braces, evaluate contents without function checking.
// This matches mux_exec behavior: inside {} with EV_EVAL,
// the flags become eval & ~(EV_STRIP_CURLY|EV_FCHECK|EV_FMAND).
int innerFlags = (flags & ~(EV_STRIP_CURLY | EV_FCHECK | EV_FMAND));
return evalWithFlags(node->children[0].get(), innerFlags);
} else {
// No strip — return the raw text including braces.
return "{" + ast_raw_text(node->children[0].get()) + "}";
}
}
// ---------------------------------------------------------------
// Builtin function registration
// ---------------------------------------------------------------
void registerBuiltins() {
// -- Arithmetic --
m_funcs["ADD"] = [](const std::vector<std::string> &args) -> std::string {
double sum = 0;
for (const auto &a : args) sum += toDouble(a);
return fmtNum(sum);
};
m_funcs["SUB"] = [](const std::vector<std::string> &args) -> std::string {
if (args.size() < 2) return "0";
return fmtNum(toDouble(args[0]) - toDouble(args[1]));
};
m_funcs["MUL"] = [](const std::vector<std::string> &args) -> std::string {
double prod = 1;
for (const auto &a : args) prod *= toDouble(a);
return fmtNum(prod);
};
m_funcs["DIV"] = [](const std::vector<std::string> &args) -> std::string {
if (args.size() < 2) return "0";
long b = toLong(args[1]);
if (b == 0) return "#-1 DIVIDE BY ZERO";
return std::to_string(toLong(args[0]) / b);
};
m_funcs["MOD"] = [](const std::vector<std::string> &args) -> std::string {
if (args.size() < 2) return "0";
long b = toLong(args[1]);
if (b == 0) return "#-1 DIVIDE BY ZERO";
return std::to_string(toLong(args[0]) % b);
};
m_funcs["ABS"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
return fmtNum(fabs(toDouble(args[0])));
};
m_funcs["INC"] = [](const std::vector<std::string> &args) -> std::string {
return std::to_string((args.empty() ? 0 : toLong(args[0])) + 1);
};
m_funcs["DEC"] = [](const std::vector<std::string> &args) -> std::string {
return std::to_string((args.empty() ? 0 : toLong(args[0])) - 1);
};
m_funcs["FLOOR"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
return std::to_string(static_cast<long long>(floor(toDouble(args[0]))));
};
m_funcs["CEIL"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
return std::to_string(static_cast<long long>(ceil(toDouble(args[0]))));
};
m_funcs["ROUND"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
double v = toDouble(args[0]);
int places = args.size() > 1 ? static_cast<int>(toLong(args[1])) : 0;
double factor = pow(10.0, places);
v = round(v * factor) / factor;
if (places <= 0) return std::to_string(static_cast<long long>(v));
char buf[64];
snprintf(buf, sizeof(buf), "%.*f", places, v);
return buf;
};
m_funcs["MAX"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
double m = toDouble(args[0]);
for (size_t i = 1; i < args.size(); i++) {
double v = toDouble(args[i]);
if (v > m) m = v;
}
return fmtNum(m);
};
m_funcs["MIN"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
double m = toDouble(args[0]);
for (size_t i = 1; i < args.size(); i++) {
double v = toDouble(args[i]);
if (v < m) m = v;
}
return fmtNum(m);
};
m_funcs["POWER"] = [](const std::vector<std::string> &args) -> std::string {
if (args.size() < 2) return "0";
return fmtNum(pow(toDouble(args[0]), toDouble(args[1])));
};
m_funcs["SQRT"] = [](const std::vector<std::string> &args) -> std::string {
if (args.empty()) return "0";
double v = toDouble(args[0]);
if (v < 0) return "#-1 SQUARE ROOT OF NEGATIVE";
return fmtNum(sqrt(v));
};
// -- Comparison --
m_funcs["EQ"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : (toLong(a[0]) == toLong(a[1]) ? "1" : "0"); };
m_funcs["NEQ"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : (toLong(a[0]) != toLong(a[1]) ? "1" : "0"); };
m_funcs["GT"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : (toLong(a[0]) > toLong(a[1]) ? "1" : "0"); };
m_funcs["GTE"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : (toLong(a[0]) >= toLong(a[1]) ? "1" : "0"); };
m_funcs["LT"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : (toLong(a[0]) < toLong(a[1]) ? "1" : "0"); };
m_funcs["LTE"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : (toLong(a[0]) <= toLong(a[1]) ? "1" : "0"); };
m_funcs["COMP"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "0";
int r = a[0].compare(a[1]);
return std::to_string(r < 0 ? -1 : (r > 0 ? 1 : 0));
};
// -- Boolean --
m_funcs["AND"] = [](const std::vector<std::string> &a) -> std::string {
for (const auto &x : a) if (!toBool(x)) return "0";
return "1";
};
m_funcs["OR"] = [](const std::vector<std::string> &a) -> std::string {
for (const auto &x : a) if (toBool(x)) return "1";
return "0";
};
m_funcs["NOT"] = [](const std::vector<std::string> &a) { return (a.empty() || !toBool(a[0])) ? "1" : "0"; };
m_funcs["XOR"] = [](const std::vector<std::string> &a) { return a.size() < 2 ? "0" : ((toBool(a[0]) != toBool(a[1])) ? "1" : "0"); };
m_funcs["T"] = [](const std::vector<std::string> &a) { return (a.empty() || !toBool(a[0])) ? "0" : "1"; };
// -- String --
m_funcs["STRLEN"] = [](const std::vector<std::string> &a) { return a.empty() ? "0" : std::to_string(a[0].size()); };
m_funcs["MID"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 3) return "";
long pos = toLong(a[1]), len = toLong(a[2]);
if (pos < 0 || len < 0 || pos >= static_cast<long>(a[0].size())) return "";
return a[0].substr(pos, len);
};
m_funcs["LEFT"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
long len = toLong(a[1]);
return len <= 0 ? "" : a[0].substr(0, len);
};
m_funcs["RIGHT"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
long len = toLong(a[1]);
if (len <= 0) return "";
if (len >= static_cast<long>(a[0].size())) return a[0];
return a[0].substr(a[0].size() - len);
};
m_funcs["CAPSTR"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty() || a[0].empty()) return "";
std::string s = a[0];
s[0] = static_cast<char>(toupper(static_cast<unsigned char>(s[0])));
return s;
};
m_funcs["LCSTR"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
std::string s = a[0];
for (auto &c : s) c = static_cast<char>(tolower(static_cast<unsigned char>(c)));
return s;
};
m_funcs["UCSTR"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
std::string s = a[0];
for (auto &c : s) c = static_cast<char>(toupper(static_cast<unsigned char>(c)));
return s;
};
m_funcs["CAT"] = [](const std::vector<std::string> &a) -> std::string {
std::string r;
for (size_t i = 0; i < a.size(); i++) { if (i > 0) r += " "; r += a[i]; }
return r;
};
m_funcs["STRCAT"] = [](const std::vector<std::string> &a) -> std::string {
std::string r;
for (const auto &x : a) r += x;
return r;
};
m_funcs["REPEAT"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
long n = toLong(a[1]);
if (n <= 0) return "";
std::string r;
for (long i = 0; i < n; i++) r += a[0];
return r;
};
m_funcs["SPACE"] = [](const std::vector<std::string> &a) -> std::string {
long n = a.empty() ? 1 : toLong(a[0]);
return n <= 0 ? "" : std::string(n, ' ');
};
m_funcs["TRIM"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
size_t start = a[0].find_first_not_of(' ');
if (start == std::string::npos) return "";
size_t end = a[0].find_last_not_of(' ');
return a[0].substr(start, end - start + 1);
};
// -- List --
m_funcs["WORDS"] = [](const std::vector<std::string> &a) {
return a.empty() ? "0" : std::to_string(splitList(a[0]).size());
};
m_funcs["FIRST"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
auto w = splitList(a[0]);
return w.empty() ? "" : w[0];
};
m_funcs["REST"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
auto w = splitList(a[0]);
if (w.size() <= 1) return "";
std::string r;
for (size_t i = 1; i < w.size(); i++) { if (i > 1) r += " "; r += w[i]; }
return r;
};
m_funcs["LAST"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
auto w = splitList(a[0]);
return w.empty() ? "" : w.back();
};
m_funcs["EXTRACT"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 3) return "";
auto w = splitList(a[0]);
long first = toLong(a[1]) - 1, count = toLong(a[2]);
if (first < 0) first = 0;
std::string r;
for (long i = first; i < first + count && i < static_cast<long>(w.size()); i++) {
if (i > first) r += " ";
r += w[i];
}
return r;
};
m_funcs["LNUM"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
long n = toLong(a[0]);
std::string sep = a.size() > 1 ? a[1] : " ";
std::string r;
for (long i = 0; i < n; i++) { if (i > 0) r += sep; r += std::to_string(i); }
return r;
};
m_funcs["SORT"] = [](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
auto w = splitList(a[0]);
bool allNum = true;
for (const auto &x : w) {
char *end;
strtod(x.c_str(), &end);
if (end == x.c_str() || *end != '\0') { allNum = false; break; }
}
if (allNum) {
std::sort(w.begin(), w.end(), [](const std::string &x, const std::string &y) {
return strtod(x.c_str(), nullptr) < strtod(y.c_str(), nullptr);
});
} else {
std::sort(w.begin(), w.end());
}
std::string r;
for (size_t i = 0; i < w.size(); i++) { if (i > 0) r += " "; r += w[i]; }
return r;
};
m_funcs["MEMBER"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "0";
auto w = splitList(a[0]);
for (size_t i = 0; i < w.size(); i++) if (w[i] == a[1]) return std::to_string(i + 1);
return "0";
};
m_funcs["INDEX"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 4) return "";
auto items = splitList(a[0], a[1]);
long first = toLong(a[2]) - 1, count = toLong(a[3]);
if (first < 0) first = 0;
std::string r;
for (long i = first; i < first + count && i < static_cast<long>(items.size()); i++) {
if (i > first) r += a[1];
r += items[i];
}
return r;
};
// -- Set operations --
m_funcs["SETUNION"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
auto x = splitList(a[0]), y = splitList(a[1]);
std::vector<std::string> result;
std::sort(x.begin(), x.end()); std::sort(y.begin(), y.end());
std::set_union(x.begin(), x.end(), y.begin(), y.end(), std::back_inserter(result));
std::string r;
for (size_t i = 0; i < result.size(); i++) { if (i > 0) r += " "; r += result[i]; }
return r;
};
m_funcs["SETDIFF"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
auto x = splitList(a[0]), y = splitList(a[1]);
std::vector<std::string> result;
std::sort(x.begin(), x.end()); std::sort(y.begin(), y.end());
std::set_difference(x.begin(), x.end(), y.begin(), y.end(), std::back_inserter(result));
std::string r;
for (size_t i = 0; i < result.size(); i++) { if (i > 0) r += " "; r += result[i]; }
return r;
};
m_funcs["SETINTER"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
auto x = splitList(a[0]), y = splitList(a[1]);
std::vector<std::string> result;
std::sort(x.begin(), x.end()); std::sort(y.begin(), y.end());
std::set_intersection(x.begin(), x.end(), y.begin(), y.end(), std::back_inserter(result));
std::string r;
for (size_t i = 0; i < result.size(); i++) { if (i > 0) r += " "; r += result[i]; }
return r;
};
// -- Registers --
m_funcs["SETQ"] = [this](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
m_ctx.registers[a[0]] = a[1];
return "";
};
m_funcs["SETR"] = [this](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "";
m_ctx.registers[a[0]] = a[1];
return a[1];
};
m_funcs["R"] = [this](const std::vector<std::string> &a) -> std::string {
if (a.empty()) return "";
auto it = m_ctx.registers.find(a[0]);
return (it != m_ctx.registers.end()) ? it->second : "";
};
// -- Misc (normal eval) --
m_funcs["NULL"] = [](const std::vector<std::string> &) -> std::string { return ""; };
m_funcs["MATCH"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "0";
auto w = splitList(a[0]);
for (size_t i = 0; i < w.size(); i++) if (w[i] == a[1]) return std::to_string(i + 1);
return "0";
};
m_funcs["STRMATCH"] = [](const std::vector<std::string> &a) -> std::string {
if (a.size() < 2) return "0";
if (a[1] == "*") return "1";
return (a[0] == a[1]) ? "1" : "0";
};
m_funcs["ITEXT"] = [this](const std::vector<std::string> &a) -> std::string {
int depth = a.empty() ? 0 : static_cast<int>(toLong(a[0]));
int idx = static_cast<int>(m_ctx.iterStack.size()) - 1 - depth;
if (idx >= 0 && idx < static_cast<int>(m_ctx.iterStack.size()))
return m_ctx.iterStack[idx].itext;
return "";
};
m_funcs["INUM"] = [this](const std::vector<std::string> &a) -> std::string {
int depth = a.empty() ? 0 : static_cast<int>(toLong(a[0]));
int idx = static_cast<int>(m_ctx.iterStack.size()) - 1 - depth;
if (idx >= 0 && idx < static_cast<int>(m_ctx.iterStack.size()))
return std::to_string(m_ctx.iterStack[idx].inum);
return "";
};
// ---------------------------------------------------------------
// FN_NOEVAL functions — deferred evaluation
// ---------------------------------------------------------------
m_noeval_funcs["IFELSE"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
if (c.size() < 3) return "";
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
return toBool(eval(c[0].get())) ? evalNoevalArg(c[1].get()) : evalNoevalArg(c[2].get());
};
// switch(val, pat1, result1, ..., default)
m_noeval_funcs["SWITCH"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
if (c.size() < 2) return "";
std::string val = eval(c[0].get());
for (size_t i = 1; i + 1 < c.size(); i += 2) {
std::string pat = eval(c[i].get());
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
if (pat == val || pat == "*") return evalNoevalArg(c[i + 1].get());
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
if (c.size() % 2 == 0) return evalNoevalArg(c.back().get());
return "";
};
m_noeval_funcs["CASE"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
if (c.size() < 2) return "";
std::string val = eval(c[0].get());
for (size_t i = 1; i + 1 < c.size(); i += 2) {
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
if (eval(c[i].get()) == val) return evalNoevalArg(c[i + 1].get());
}
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
if (c.size() % 2 == 0) return evalNoevalArg(c.back().get());
return "";
};
// Short-circuit boolean
m_noeval_funcs["CAND"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
for (const auto &x : c) if (!toBool(eval(x.get()))) return "0";
return "1";
};
m_noeval_funcs["COR"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
for (const auto &x : c) if (toBool(eval(x.get()))) return "1";
return "0";
};
// @@(comment) — discard without evaluating
m_noeval_funcs["@@"] = [](const std::vector<std::unique_ptr<ASTNode>> &) -> std::string {
return "";
};
// lit(text) — return unevaluated source text
m_noeval_funcs["LIT"] = [](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
return c.empty() ? "" : ast_raw_text(c[0].get());
};
// iter(list, body, osep, isep) — evaluate body per item
m_noeval_funcs["ITER"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
if (c.size() < 2) return "";
std::string listVal = eval(c[0].get());
std::string sep = c.size() > 3 ? eval(c[3].get()) : " ";
std::string osep = c.size() > 2 ? eval(c[2].get()) : " ";
auto items = splitList(listVal, sep);
std::string result;
for (size_t i = 0; i < items.size(); i++) {
if (i > 0) result += osep;
m_ctx.iterStack.push_back({items[i], static_cast<int>(i + 1)});
result += evalNoevalArg(c[1].get());
m_ctx.iterStack.pop_back();
}
return result;
};
m_noeval_funcs["IF"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
if (c.empty()) return "";
std::string cond_str = evalWithFlags(c[0].get(), (m_ctx.evalFlags & ~EV_TOP) | EV_EVAL | EV_FCHECK);
bool cond = !cond_str.empty() && cond_str != "0";
if (cond) {
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
return (c.size() > 1) ? evalNoevalArg(c[1].get()) : "";
} else {
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
return (c.size() > 2) ? evalNoevalArg(c[2].get()) : "";
}
};
m_noeval_funcs["EVAL"] = [this](const std::vector<std::unique_ptr<ASTNode>> &c) -> std::string {
if (c.empty()) return "";
std::string text = evalWithFlags(c[0].get(), (m_ctx.evalFlags & ~EV_TOP) | EV_EVAL | EV_FCHECK);
2026-03-21 04:28:44 -06:00
auto tokens = tokenize(text.c_str(), m_ctx.profile);
Parser parser(tokens);
auto ast = parser.parse();
return evalWithFlags(ast.get(), (m_ctx.evalFlags & ~EV_TOP) | EV_EVAL | EV_FCHECK);
};
}
};
// ---------------------------------------------------------------
// Main
// ---------------------------------------------------------------
int main(int argc, char *argv[])
{
bool showAST = false;
ParserProfile profile = PROFILE_MUX214_AST;
for (int i = 1; i < argc; i++) {
if (strcmp(argv[i], "-a") == 0 || strcmp(argv[i], "--ast") == 0) {
showAST = true;
} else if (0 == strcmp(argv[i], "--profile")) {
if (i + 1 >= argc || !parse_profile_string(argv[i + 1], profile)) {
fprintf(stderr, "usage: %s [--ast] [--profile mux214|mux213|penn]\n", argv[0]);
return 2;
}
i++;
} else if (0 == strncmp(argv[i], "--profile=", 10)) {
if (!parse_profile_string(argv[i] + 10, profile)) {
fprintf(stderr, "usage: %s [--ast] [--profile mux214|mux213|penn]\n", argv[0]);
return 2;
}
} else {
fprintf(stderr, "usage: %s [--ast] [--profile mux214|mux213|penn]\n", argv[0]);
return 2;
}
}
EvalContext ctx;
ctx.profile = profile;
ctx.enactorName = "testplayer";
ctx.enactorDbref = "#1234";
ctx.executorDbref = "#1234";
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
ctx.callerDbref = "#1234";
ctx.enactorObjid = "#1234:1700000000";
ctx.lastCommand = "test";
ctx.moniker = "TestPlayer";
ctx.pronSub = "they";
ctx.pronPos = "their";
ctx.pronObj = "them";
ctx.pronAbs = "theirs";
ctx.rawCommand = "@pemit me=%cg";
ctx.evaledCommand = "@pemit me=%cg";
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
ctx.accentedName = "testplayer";
ctx.attrName = "";
ctx.varAttrs['G' - 'A'] = "thing";
ctx.args[0] = "hello";
ctx.args[1] = "world";
Add parser percent-substitution matrix and fix noeval brace-group backslash handling Root cause confirmed via debugger: 2.13's mux_exec() backslash handler (eval.cpp line 2439) has no EV_EVAL guard — it runs unconditionally on every pass. When brace-group arguments go through FN_NOEVAL functions like switch(), the text passes through mux_exec() twice: once for noeval arg collection (stripping one backslash layer) and once for evaluation. This two-pass behavior turns \\% into \% then into literal %. The 2.14 AST scanner tokenizes once greedily — \\ becomes one ESC token and % becomes one SUBST token, evaluated independently. This produces \<space> instead of % for the bboard case. Fix: add noevalPass() and evalNoevalArg() to the parser prototype. For brace-group args to FN_NOEVAL functions, strip one backslash layer via noevalPass(), re-tokenize, then evaluate. This replicates 2.13's unconditional backslash consumption. Changes: - docs/parser-percent-matrix.md: complete semantic matrix of all % forms across 2.13, 2.14, and PennMUSH with confirmed root cause - parser/eval.cpp: full % substitution coverage (pronouns, color stubs, caller, objid, moniker, hash forms, etc.), noevalPass/evalNoevalArg for two-pass brace-group handling in if/switch/case - parser/mux_parse.h: tokenizer support for hash forms (##/#@/#$), Penn %$ stack vars, %w attrs, %iL, angle-bracket helper - parser/test_eval.sh: 123 tests (was 86), cross-profile escape+percent tests including the confirmed bboard divergence case Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-21 04:17:43 -06:00
ctx.nargs = 2;
Evaluator evaluator(ctx);
char line[8192];
while (fgets(line, sizeof(line), stdin)) {
size_t len = strlen(line);
if (len > 0 && line[len - 1] == '\n') {
line[len - 1] = '\0';
}
auto tokens = tokenize(line, profile);
Parser parser(tokens);
auto ast = parser.parse();
if (showAST) {
printf("INPUT: %s\n", line);
printf("PROFILE: %s\n", profile_name(profile));
printf("AST:\n");
ast_print(ast.get(), 2);
}
std::string result = evaluator.eval(ast.get());
printf("%s\n", result.c_str());
}
return 0;
}