X-Git-Url: https://ocean-lang.org/code/?a=blobdiff_plain;f=csrc%2Fparsergen.mdc;h=6d291312547f0aad1ddadc57ee2affeb7e8b3802;hb=afa1be3c3a7f14a8c476d937a33376490272ba45;hp=a7330351d767b52f52ba7b14a802e1b2556dd842;hpb=fa35c2e89220fb586700f934d9beba2f3e806f0f;p=ocean

diff --git a/csrc/parsergen.mdc b/csrc/parsergen.mdc
index a733035..6d29131 100644
--- a/csrc/parsergen.mdc
+++ b/csrc/parsergen.mdc
@@ -277,6 +277,9 @@ declares how it associates.  This level is stored in each symbol
 listed and may be inherited by any production which uses the symbol.  A
 production inherits from the last symbol which has a precedence.
 
+The symbols on the first precedence line have the lowest precedence.
+Subsequent lines introduce symbols with higher precedence.
+
 ###### grammar fields
 	struct text current_type;
 	int type_isref;
@@ -1287,12 +1290,11 @@ When itemsets are created we assign a precedence to the itemset from
 the complete item, if there is one.  We ignore the possibility of
 there being two and don't (currently) handle precedence in such
 grammars.  When completing a grammar we ignore any item where DOT is
-followed by a terminal with a precedence lower (numerically higher)
-than that for the itemset.  Unless the terminal has right
-associativity, we also ignore items where the terminal has the same
-precedence.  The result is that unwanted items are still in the
-itemset, but the terminal doesn't get into the go to set, so the item
-is ineffective.
+followed by a terminal with a precedence lower than that for the
+itemset.  Unless the terminal has right associativity, we also ignore
+items where the terminal has the same precedence.  The result is that
+unwanted items are still in the itemset, but the terminal doesn't get
+into the go to set, so the item is ineffective.
 
 ###### complete itemset
 	for (i = 0; i < is->items.cnt; i++) {
@@ -1311,7 +1313,7 @@ is ineffective.
 			continue;
 		s = pr->body[bs];
 		if (s->precedence && is->precedence &&
-		    is->precedence < s->precedence)
+		    is->precedence > s->precedence)
 			/* This terminal has a low precedence and
 			 * shouldn't be shifted
 			 */
@@ -1404,8 +1406,7 @@ with a pre-existing itemset).
 			pos = symset_find(&newitemset, pr->head->num);
 			if (bp + 1 == pr->body_size &&
 			    pr->precedence > 0 &&
-			    (precedence == 0 ||
-			     pr->precedence < precedence)) {
+			    pr->precedence > precedence) {
 				// new itemset is reducible and has a precedence.
 				precedence = pr->precedence;
 				assoc = pr->assoc;
@@ -1776,10 +1777,10 @@ As a special case, if we find a SHIFT/REDUCE conflict, where a
 terminal that could be shifted is in the lookahead set of some
 reducable item, then set check if the reducable item also have
 `TK_newline` in its lookahead set.  If it does, then a newline will
-force and reduction, but anything else can reasonably be shifts, so
+force the reduction, but anything else can reasonably be shifted, so
 that isn't really a conflict.  Such apparent conflicts do not get
-reported.  This will not affect a "tradtional" grammar that does not
-include newlines as token.
+counted, and are reported as non-critical.  This will not affect a
+"traditional" grammar that does not include newlines as token.
 
 	static int conflicts_slr(struct grammar *g, enum grammar_type type)
 	{
@@ -1799,14 +1800,20 @@ include newlines as token.
 				int p = item_prod(itm);
 				int bp = item_index(itm);
 				struct production *pr = g->productions[p];
+				struct symbol *s;
 
-				if (bp < pr->body_size &&
-				    pr->body[bp]->type == Terminal) {
-					/* shiftable */
-					int sym = pr->body[bp]->num;
-					if (symset_find(&shifts, sym) < 0)
-						symset_add(&shifts, sym, itm);
-				}
+				if (bp >= pr->body_size ||
+				    pr->body[bp]->type != Terminal)
+					/* not shiftable */
+					continue;
+
+				s = pr->body[bp];
+				if (s->precedence && is->precedence)
+					/* Precedence resolves this, so no conflict */
+					continue;
+
+				if (symset_find(&shifts, s->num) < 0)
+					symset_add(&shifts, s->num, itm);
 			}
 			/* Now look for reductions and conflicts */
 			for (j = 0; j < is->items.cnt; j++) {
@@ -1826,13 +1833,16 @@ include newlines as token.
 				int k;
 				for (k = 0; k < la.cnt; k++) {
 					int pos = symset_find(&shifts, la.syms[k]);
-					if (pos >= 0 && symset_find(&la, TK_newline) < 0) {
-						printf("  State %d has SHIFT/REDUCE conflict on ", i);
+					if (pos >= 0) {
+						if (symset_find(&la, TK_newline) < 0) {
+							printf("  State %d has SHIFT/REDUCE conflict on ", i);
+							cnt++;
+						} else
+							printf("  State %d has non-critical SHIFT/REDUCE conflict on ", i);
 						prtxt(g->symtab[la.syms[k]]->name);
 						printf(":\n");
 						report_item(g, shifts.data[pos]);
 						report_item(g, itm);
-						cnt++;
 					}
 					pos = symset_find(&reduce, la.syms[k]);
 					if (pos < 0) {
@@ -1869,13 +1879,14 @@ pieces of code provided in the grammar file, so they are generated first.
 
 ###### parser_generate
 
-	static void gen_parser(FILE *f, struct grammar *g, char *file, char *name)
+	static void gen_parser(FILE *f, struct grammar *g, char *file, char *name,
+		               struct code_node *pre_reduce)
 	{
 		gen_known(f, g);
 		gen_non_term(f, g);
 		gen_goto(f, g);
 		gen_states(f, g);
-		gen_reduce(f, g, file);
+		gen_reduce(f, g, file, pre_reduce);
 		gen_free(f, g);
 
 		fprintf(f, "#line 0 \"gen_parser\"\n");
@@ -1896,7 +1907,9 @@ pieces of code provided in the grammar file, so they are generated first.
 ### Known words table
 
 The known words table is simply an array of terminal symbols.
-The table of nonterminals used for tracing is a similar array.
+The table of nonterminals used for tracing is a similar array.  We
+include virtual symbols in the table of non_terminals to keep the
+numbers right.
 
 ###### functions
 
@@ -1921,7 +1934,7 @@ The table of nonterminals used for tracing is a similar array.
 		for (i = TK_reserved;
 		     i < g->num_syms;
 		     i++)
-			if (g->symtab[i]->type == Nonterminal)
+			if (g->symtab[i]->type != Terminal)
 				fprintf(f, "\t\"%.*s\",\n", g->symtab[i]->name.len,
 					g->symtab[i]->name.txt);
 		fprintf(f, "};\n\n");
@@ -2109,14 +2122,18 @@ automatically freed.  This is equivalent to assigning `NULL` to the pointer.
 
 ###### functions
 
-	static void gen_reduce(FILE *f, struct grammar *g, char *file)
+	static void gen_reduce(FILE *f, struct grammar *g, char *file,
+	                       struct code_node *code)
 	{
 		int i;
-		fprintf(f, "#line 0 \"gen_reduce\"\n");
+		fprintf(f, "#line 1 \"gen_reduce\"\n");
 		fprintf(f, "static int do_reduce(int prod, void **body, struct token_config *config, void *ret)\n");
 		fprintf(f, "{\n");
 		fprintf(f, "\tint ret_size = 0;\n");
+		if (code)
+			code_node_print(f, code, file);
 
+		fprintf(f, "#line 4 \"gen_reduce\"\n");
 		fprintf(f, "\tswitch(prod) {\n");
 		for (i = 0; i < g->production_count; i++) {
 			struct production *p = g->productions[i];
@@ -2313,6 +2330,7 @@ parser with neither. "grammar" must be provided.
 	struct code_node *hdr = NULL;
 	struct code_node *code = NULL;
 	struct code_node *gram = NULL;
+	struct code_node *pre_reduce = NULL;
 	for (s = table; s; s = s->next) {
 		struct text sec = s->section;
 		if (tag && !strip_tag(&sec, tag))
@@ -2323,6 +2341,8 @@ parser with neither. "grammar" must be provided.
 			code = s->code;
 		else if (text_is(sec, "grammar"))
 			gram = s->code;
+		else if (text_is(sec, "reduce"))
+			pre_reduce = s->code;
 		else {
 			fprintf(stderr, "Unknown content section: %.*s\n",
 			        s->section.len, s->section.txt);
@@ -2394,7 +2414,7 @@ file with the code section (if any) and the parser tables and function.
 		if (f) {
 			if (code)
 				code_node_print(f, code, infile);
-			gen_parser(f, g, infile, name);
+			gen_parser(f, g, infile, name, pre_reduce);
 			fclose(f);
 		} else {
 			fprintf(stderr, "Cannot create %s.c\n",
@@ -3076,8 +3096,8 @@ an error.
 
 # calc: grammar
 
-	$LEFT * /
 	$LEFT + -
+	$LEFT * /
 
 	Session -> Session Line
 		| Line