b2/: add parsing of substitution rules (WIP)

b2/Makefile

@@ -13,7 +13,8 @@ SHELL = /bin/bash
 
 CFLAGS = -g -Wall $(shell pkg-config --cflags glib-2.0)
 SLOPPY = -Wno-unused -Wno-implicit-function-declaration
-OBJS = boom.o chr.o comp.o db.o dump.o eval.o param.o util.o lex.yy.o y.tab.o
+OBJS = boom.o chr.o comp.o db.o dump.o eval.o param.o subst.o util.o \
+       lex.yy.o y.tab.o
 LDLIBS = -lfl $(shell pkg-config --libs glib-2.0)
 
 YACC = bison -y
@@ -83,4 +84,5 @@ spotless:	clean
 # ----- Experiments -----------------------------------------------------------
 
 try:
-	$(VALGRIND) ./boom HIERARCHY -c CHAR -x CURR -p PROVIDER -i INV
+	$(VALGRIND) ./boom HIERARCHY -c CHAR -x CURR -p PROVIDER -i INV \
+	  -s SUBST

b2/SUBST

@@ -0,0 +1,25 @@
+REF=R[0-9]* {
+	T=R
+	VAL=.* { R=$$#R }
+	TOL = 5%
+	FN=.*% { TOL=${$#%} }
+	break REF
+//	end break again
+}
+
+/*
+pattern: () * ? .
+subst:  $1 ... $field
+
+substring:
+  $1, $2, ...
+variable:
+  $foo, ...
+with curly braces:
+  ${foo}, ...
+unit conversion:
+  $foo#V
+  ${1#R}
+input variable:
+  $$
+*/

b2/boom.c

@@ -26,6 +26,7 @@ static void usage(const char *name)
 "  -i  inventory\n"
 "  -x  currency exchange\n"
 "  -p  providers\n"
+"  -s  substitutions\n"
     , name);
 	exit(1);
 }
@@ -46,6 +47,8 @@ int main(int argc, char **argv)
 				parse = parse_currencies;
 			else if (!strcmp(argv[i], "-p"))
 				parse = parse_providers;
+			else if (!strcmp(argv[i], "-s"))
+				parse = parse_substitutions;
 			else
 				usage(*argv);
 		} else {

b2/lang.h

@@ -18,9 +18,10 @@ void parse_characteristics(const char *name);
 void parse_inventory(const char *name);
 void parse_currencies(const char *name);
 void parse_providers(const char *name);
+void parse_substitutions(const char *name);
 
 void yywarnf(const char *fmt, ...);
-void yyerrorf(const char *fmt, ...);
-void yyerror(const char *s);
+void __attribute__((noreturn)) yyerrorf(const char *fmt, ...);
+void __attribute__((noreturn)) yyerror(const char *s);
 
 #endif /* !LANG_H */

b2/lang.l

@@ -26,6 +26,7 @@ extern int yyparse(void);
 
 static int start_token;
 static int expose_nl;
+static int pattern;
 static const char *file_name;
 static int lineno;
 
@@ -46,12 +47,13 @@ static void open_stdin(const char *name)
 }
 
 
-static void do_parse(const char *name, int start, int nl)
+static void do_parse(const char *name, int start, int nl, int pat)
 {
 	open_stdin(name);
 	
 	start_token = start;
 	expose_nl = nl;
+	pattern = pat;
 	file_name = unique(name);
 	lineno = 1;
 	yyparse();
@@ -60,31 +62,37 @@ static void do_parse(const char *name, int start, int nl)
 
 void parse_hierarchy(const char *name)
 {
-	do_parse(name, START_HIERARCHY, 0);
+	do_parse(name, START_HIERARCHY, 0, 0);
 }
 
 
 void parse_characteristics(const char *name)
 {
-	do_parse(name, START_CHAR, 1);
+	do_parse(name, START_CHAR, 1, 0);
 }
 
 
 void parse_inventory(const char *name)
 {
-	do_parse(name, START_INVENTORY, 1);
+	do_parse(name, START_INVENTORY, 1, 0);
 }
 
 
 void parse_currencies(const char *name)
 {
-	do_parse(name, START_EXCHANGE, 1);
+	do_parse(name, START_EXCHANGE, 1, 0);
 }
 
 
 void parse_providers(const char *name)
 {
-	do_parse(name, START_PROVIDERS, 1);
+	do_parse(name, START_PROVIDERS, 1, 0);
+}
+
+
+void parse_substitutions(const char *name)
+{
+	do_parse(name, START_SUBST, 0, 1);
 }
 
 %}
@@ -96,6 +104,9 @@ void parse_providers(const char *name)
  */
 
 ID	[-_A-Za-z0-9()+./]
+PAT	"\\"[^\t\n]|[^ \t\n\\]
+
+%s ID PAT
 
 %%
 
@@ -108,9 +119,17 @@ ID	[-_A-Za-z0-9()+./]
 	}
 %}
 
-{ID}({ID}|"%")*			{ yylval.s = unique(yytext);
+<INITIAL,ID>{ID}({ID}|"%")*		{ yylval.s = unique(yytext);
 				  return WORD; }
 
+<PAT>{PAT}*			{ BEGIN(ID);
+				  yylval.s = stralloc(yytext);
+				  return PATTERN; }
+
+"="				{ if (pattern)
+					BEGIN(PAT);
+				  return '='; }
+
 "<="				return TOK_LE;
 ">="				return TOK_GE;
 
@@ -142,7 +161,7 @@ void yywarnf(const char *fmt, ...)
 }
 
 
-void yyerrorf(const char *fmt, ...)
+void __attribute__((noreturn)) yyerrorf(const char *fmt, ...)
 {
 	va_list ap;
 
@@ -155,7 +174,7 @@ void yyerrorf(const char *fmt, ...)
 }
 
 
-void yyerror(const char *s)
+void __attribute__((noreturn)) yyerror(const char *s)
 {
 	fprintf(stderr, "%s:%d: %s\n", file_name, lineno, s);
 	exit(1);

b2/lang.y

@@ -18,6 +18,7 @@
 #include "param.h"
 #include "chr.h"
 #include "db.h"
+#include "subst.h"
 #include "lang.h"
 
 #include "y.tab.h"
@@ -81,13 +82,14 @@ static const struct field *top_field(void)
 	struct price *price;
 	struct stock *stock;
 	struct provider *prov;
+	struct subst *subst;
 };
 
 
 %token			START_HIERARCHY START_CHAR START_INVENTORY
-%token			START_EXCHANGE START_PROVIDERS
+%token			START_EXCHANGE START_PROVIDERS START_SUBST
 %token			TOK_LE TOK_GE TOK_NL
-%token	<s>		WORD
+%token	<s>		WORD PATTERN
 
 %type	<num>		int
 %type	<fnum>		float
@@ -104,6 +106,7 @@ static const struct field *top_field(void)
 %type	<price>		prices price
 %type	<stock>		stock
 %type	<prov>		providers provider
+%type	<subst>		block opt_block
 
 %%
 
@@ -113,6 +116,7 @@ all:
 	| START_INVENTORY inventory
 	| START_EXCHANGE exchange
 	| START_PROVIDERS providers
+	| START_SUBST substitutions
 	;
 
 
@@ -554,3 +558,57 @@ provider:
 			$$->minimum = $4;
 		}
 	;
+
+
+/* ----- Substitutions ----------------------------------------------------- */
+
+
+substitutions:
+	block
+		{
+			subst_finalize($1);
+			subst_dump(stderr, $1);
+		}
+	;
+
+block:
+		{
+			$$ = NULL;
+		}
+	| WORD '=' PATTERN opt_block block
+		{
+			if ($4) {
+				$$ = subst_match($1, $3);
+				$$->u.match.block = $4;
+			} else {
+				$$ = subst_assign($1, $3);
+			}
+			free((void *) $3);
+			$$->next = $5;
+		}
+	| WORD WORD
+		{
+			if (!strcmp($1, "break"))
+				$$ = subst_break($2);
+			else if (!strcmp($1, "again"))
+				$$ = subst_again($2);
+			else
+				yyerrorf("unknown keyword \"%s\"", $1);
+		}
+	| WORD
+		{
+			if (strcmp($1, "end"))
+				yyerrorf("unknown keyword \"%s\"", $1);
+			$$ = subst_end();
+		}
+	;
+
+opt_block:
+		{
+			$$ = NULL;
+		}
+	| '{' block '}'
+		{
+			$$ = $2;
+		}
+	;

b2/subst.c

@@ -0,0 +1,349 @@
+/*
+ * subst.c - Substitution rules
+ *
+ * Copyright 2012 by Werner Almesberger
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+#include <sys/types.h>
+#include <regex.h>
+#include <assert.h>
+
+#include "util.h"
+#include "lang.h"
+#include "subst.h"
+
+
+/* ----- Rule set construction --------------------------------------------- */
+
+
+static struct subst *alloc_subst(enum subst_type type)
+{
+	struct subst *sub;
+
+	sub = alloc_type(struct subst);
+	sub->type = type;
+	sub->next = NULL;
+	return sub;
+}
+
+
+struct subst *subst_match(const char *src, const char *re)
+{
+	char error[1000];
+	struct subst *sub;
+	char *tmp;
+	int err, len;
+
+	sub = alloc_subst(st_match);
+	sub->u.match.src = src;
+	len = strlen(re);
+	tmp = alloc_size(len+3);
+	tmp[0] = '^';
+	memcpy(tmp+1, re, len);
+	tmp[len+1] = '$';
+	tmp[len+2] = 0;
+	err = regcomp(&sub->u.match.re, tmp, REG_EXTENDED);
+	free(tmp);
+	if (err) {
+		regerror(err, &sub->u.match.re, error, sizeof(error));
+		yyerrorf("%s", error);
+	}
+	return sub;
+}
+
+
+static void end_chunk(struct chunk ***last, const char *start, const char *s)
+{
+	struct chunk *c;
+
+	if (s == start)
+		return;
+
+	c = alloc_type(struct chunk);
+	c->type = ct_string;
+	c->u.s = stralloc_n(start, s-start);;
+	c->unit = NULL;
+	c->next = NULL;
+	**last = c;
+	*last = &c->next;
+}
+
+
+static const char *parse_var(struct chunk *c, const char *s)
+{
+	const char *t;
+	int braced;
+
+	if (!*s)
+		yyerror("trailing dollar sign");
+
+	braced = *s == '{';
+	if (braced)
+		s++;
+
+	t = s;
+	while (*t) {
+		if (braced && *t == '}')
+			break;
+		if (s == t && *t == '$') {
+			t++;
+			break;
+		}
+		if (!isalnum(*t))
+			break;
+		t++;
+	}
+	if (s == t)
+		yyerror("invalid variable name");
+	if (braced && !*t)
+		yyerror("unterminated variable name");
+	if (isdigit(*s)) {
+		if (t != s+1 || *s == '0')
+			yyerror("invalid variable name");
+		c->type = ct_sub;
+		c->u.sub = *s-'0';
+	} else if (isalnum(*s)) {
+		char *tmp;
+
+		c->type = ct_var;
+		tmp = stralloc_n(s, t-s);
+		c->u.var = unique(tmp);
+		free(tmp);
+	} else {
+		c->type = ct_sub;
+		c->u.sub = 0;
+	}
+
+	if (*t != '#') {
+		if (braced) {
+			assert(*t == '}');
+			t++;
+		}
+		return t;
+	}
+
+	s = ++t;
+	while (*t) {
+		if (braced && *t == '}')
+			break;
+		if (!braced && t != s)
+			break;
+		t++;
+	}
+	if (s == t)
+		yyerror("invalid unit");
+	c->unit = stralloc_n(s, t-s);
+	if (braced) {
+		if (!*t)
+			yyerror("unterminated unit");
+		t++;
+	}
+	return t;
+}
+
+
+static struct chunk *parse_pattern(const char *s)
+{
+	struct chunk *res = NULL, **last = &res;
+	struct chunk *c;
+	const char *start = s;
+
+	while (*s) {
+		if (*s == '\\') {
+			if (!s[1])
+				yyerror("trailing backslash");
+			end_chunk(&last, start, s);
+			start = s+1;
+			s += 2;
+			continue;
+		}
+		if (*s != '$') {
+			s++;
+			continue;
+		}
+
+		end_chunk(&last, start, s);
+		c = alloc_type(struct chunk);
+		c->unit = NULL;
+		c->next = NULL;
+		*last = c;
+		last = &c->next;
+		start = s = parse_var(c, s+1);
+	}
+	end_chunk(&last, start, s);
+	return res;
+}
+
+
+struct subst *subst_assign(const char *dst, const char *pat)
+{
+	struct subst *sub;
+
+	sub = alloc_subst(st_assign);
+	sub->u.assign.dst = dst;
+	sub->u.assign.pat = parse_pattern(pat);
+	return sub;
+}
+
+
+struct subst *subst_end(void)
+{
+	return alloc_subst(st_end);
+}
+
+
+struct subst *subst_break(const char *block)
+{
+	struct subst *sub;
+
+	sub = alloc_subst(st_break);
+	sub->u.tmp = block;
+	return sub;
+}
+
+
+struct subst *subst_again(const char *block)
+{
+	struct subst *sub;
+
+	sub = alloc_subst(st_again);
+	sub->u.tmp = block;
+	return sub;
+}
+
+
+/* ----- Jump resolution --------------------------------------------------- */
+
+
+struct parent {
+	const struct subst *sub;
+	const struct parent *parent;
+};
+
+
+static const struct subst *resolve_jump(const char *name,
+    const struct parent *parent)
+{
+	while (parent) {
+		assert(parent->sub->type == st_match);
+		if (!strcmp(name, parent->sub->u.match.src))
+			return parent->sub;
+		parent = parent->parent;
+	}
+	yyerrorf("cannot find \"%s\"", name);
+}
+
+
+static void recurse_fin(struct subst *sub, const struct parent *parent)
+{
+	struct parent next = {
+		.sub = sub,
+		.parent = parent,
+	};
+
+	while (sub) {
+		switch (sub->type) {
+		case st_match:
+			recurse_fin(sub->u.match.block, &next);
+		case st_assign:
+		case st_end:
+			break;
+		case st_break:
+		case st_again:
+			sub->u.jump = resolve_jump(sub->u.tmp, parent);
+			break;
+		default:
+			abort();
+		}
+		sub = sub->next;
+	}
+}
+
+
+void subst_finalize(struct subst *sub)
+{
+	recurse_fin(sub, NULL);
+}
+
+
+/* ----- Dumping ----------------------------------------------------------- */
+
+
+#define	INDENT	4
+
+
+static void dump_chunks(FILE *file, const struct chunk *c)
+{
+	while (c) {
+		switch (c->type) {
+		case ct_string:
+			fprintf(file, "%s", c->u.s);
+			break;
+		case ct_var:
+			fprintf(file, "${%s", c->u.var);
+			break;
+		case ct_sub:
+			if (c->u.sub)
+				fprintf(file, "${%d", c->u.sub);
+			else
+				fprintf(file, "${$");
+			break;
+		default:
+			abort();
+		}
+		if (c->type != ct_string) {
+			if (c->unit)
+				fprintf(file, "#%s", c->unit);
+			fprintf(file, "}");
+		}
+		c = c->next;
+	}
+}
+
+
+static void recurse_dump(FILE *file, const struct subst *sub, int level)
+{
+	while (sub) {
+		fprintf(file, "%*s", INDENT*level, "");
+		switch (sub->type) {
+		case st_match:
+			fprintf(file, "%s=RE {\n", sub->u.match.src);
+			recurse_dump(file, sub->u.match.block, level+1);
+			fprintf(file, "%*s}\n", INDENT*level, "");
+			break;
+		case st_assign:
+			fprintf(file, "%s=", sub->u.assign.dst);
+			dump_chunks(file, sub->u.assign.pat);
+			fprintf(file, "\n");
+			break;
+		case st_end:
+			fprintf(file, "end\n");
+			break;
+		case st_break:
+			fprintf(file, "break %s\n", sub->u.jump->u.match.src);
+			break;
+		case st_again:
+			fprintf(file, "again %s\n", sub->u.jump->u.match.src);
+			break;
+		default:
+			abort();
+		}
+		sub = sub->next;
+	}
+}
+
+
+void subst_dump(FILE *file, const struct subst *sub)
+{
+	recurse_dump(file, sub, 0);
+}

b2/subst.h

@@ -0,0 +1,75 @@
+/*
+ * subst.h - Substitution rules
+ *
+ * Copyright 2012 by Werner Almesberger
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+
+#ifndef SUBST_H
+#define	SUBST_H
+
+#include <stdio.h>
+#include <sys/types.h>
+#include <regex.h>
+
+
+enum chunk_type {
+	ct_string,
+	ct_var,
+	ct_sub,
+};
+
+struct chunk {
+	enum chunk_type type;
+	union {
+		const char *s;
+		const char *var;
+		int sub; /* 0 if $$ */
+	} u;
+	const char *unit; /* NULL if no conversion specified */
+	struct chunk *next;
+};
+
+enum subst_type {
+	st_match,
+	st_assign,
+	st_end,
+	st_break,
+	st_again,
+};
+
+struct subst {
+	enum subst_type type;
+	union {
+		struct {
+			const char *src;
+			regex_t re;
+			struct subst *block;
+		} match;
+		struct {
+			const char *dst;
+			struct chunk *pat;
+		} assign;
+		const struct subst *jump;
+		const char *tmp; /* jump target name; for subst_finalize */
+	} u;
+	struct subst *next;
+};
+
+
+struct subst *subst_match(const char *src, const char *re);
+struct subst *subst_assign(const char *dst, const char *pat);
+struct subst *subst_end(void);
+struct subst *subst_break(const char *block);
+struct subst *subst_again(const char *block);
+
+void subst_finalize(struct subst *sub);
+
+void subst_dump(FILE *file, const struct subst *sub);
+
+#endif /* !SUBST_H */

b2/util.h

@@ -13,6 +13,7 @@
 #define	UTIL_H
 
 #include <stdlib.h>
+#include <string.h>
 
 
 #define alloc_size(s)				\
@@ -24,6 +25,30 @@
 #define alloc_type(t) ((t *) alloc_size(sizeof(t)))
 
 
+static inline char *stralloc(const char *s)
+{
+	char *t;
+
+	t = strdup(s);
+	if (!t)
+		abort();
+	return t;
+}
+
+
+static inline char *stralloc_n(const char *s, int n)
+{
+	char *t;
+
+	t = alloc_size(n+1);
+	if (!t)
+		abort();
+	memcpy(t, s, n);
+	t[n] = 0;
+	return t;
+}
+
+
 const char *unique(const char *s);
 
 #endif /* !UTIL_H */