Various improvements to parsing in battlestar, mostly from OpenBSD.

Define a constant WORDLEN.  Always use this constant and NWORD where
appropriate.

Use NWORD - 1 in battlestar.c to avoid off-by-one error.

Increment wordnumber after the INVEN verb to allow it to be followed
by a comma and other actions.

Avoid overflowing elements of the words array if input words are too
long.

Parse "," as AND except when followed by a verb, to allow such
constructions as "take foo, bar, and baz".  Trim AND AND which may
occur from the ", and" in such a list.

Avoid crashes from EVERYTHING in the wrong place by moving it to the
start of OBJECT AND EVERYTHING and NOUNS AND EVERYTHING sequences, and
trimming EVERYTHING AND EVERYTHING.

1 files changed, 60 insertions, 2 deletions

diff --git a/battlestar/parse.c b/battlestar/parse.c
index 8d85e6c5..e79722ba 100644
--- a/battlestar/parse.c
+++ b/battlestar/parse.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: parse.c,v 1.7 1999/02/10 01:36:50 hubertf Exp $	*/
+/*	$NetBSD: parse.c,v 1.8 2000/09/21 17:44:34 jsm Exp $	*/
 
 /*
  * Copyright (c) 1983, 1993
@@ -38,7 +38,7 @@
 #if 0
 static char sccsid[] = "@(#)parse.c	8.2 (Berkeley) 4/28/95";
 #else
-__RCSID("$NetBSD: parse.c,v 1.7 1999/02/10 01:36:50 hubertf Exp $");
+__RCSID("$NetBSD: parse.c,v 1.8 2000/09/21 17:44:34 jsm Exp $");
 #endif
 #endif				/* not lint */
 
@@ -98,6 +98,7 @@ parse()
 {
 	struct wlist *wp;
 	int     n;
+	int     flag;
 
 	wordnumber = 0;		/* for cypher */
 	for (n = 0; n <= wordcount; n++) {
@@ -109,4 +110,61 @@ parse()
 			wordtype[n] = wp->article;
 		}
 	}
+	/* Don't let a comma mean AND if followed by a verb. */
+	for (n = 0; n < wordcount; n++)
+		if (wordvalue[n] == AND && words[n][0] == ','
+		    && wordtype[n + 1] == VERB) {
+			wordvalue[n] = -1;
+			wordtype[n] = -1;
+		}
+	/* Trim "AND AND" which can happen naturally at the end of a
+	 * comma-delimited list.
+	 */
+	for (n = 1; n < wordcount; n++)
+		if (wordvalue[n - 1] == AND && wordvalue[n] == AND) {
+			int i;
+			for (i = n + 1; i < wordcount; i++) {
+				wordtype[i - 1] = wordtype[i];
+				wordvalue[i - 1] = wordvalue[i];
+				strcpy(words[i - 1], words[i]);
+			}
+			wordcount--;
+		}
+
+	/* If there is a sequence (NOUN | OBJECT) AND EVERYTHING
+	 * then move all the EVERYTHINGs to the beginning, since that's where
+	 * they're expected.  We can't get rid of the NOUNs and OBJECTs in
+	 * case they aren't in EVERYTHING (i.e. not here or nonexistant).
+	 */
+	flag = 1;
+	while (flag) {
+		flag = 0;
+		for (n = 1; n < wordcount; n++)
+			if ((wordtype[n - 1] == NOUNS || wordtype[n - 1] == OBJECT) &&
+			    wordvalue[n] == AND && wordvalue[n + 1] == EVERYTHING) {
+				char tmpword[WORDLEN];
+				wordvalue[n + 1] = wordvalue[n - 1];
+				wordvalue[n - 1] = EVERYTHING;
+				wordtype[n + 1] = wordtype[n - 1];
+				wordtype[n - 1] = OBJECT;
+				strcpy(tmpword, words[n - 1]);
+				strcpy(words[n - 1], words[n + 1]);
+				strcpy(words[n + 1], tmpword);
+				flag = 1;
+		}
+		/* And trim EVERYTHING AND EVERYTHING. */
+		for (n = 1; n < wordcount; n++)
+			if (wordvalue[n - 1] == EVERYTHING &&
+			    wordvalue[n] == AND && wordvalue[n + 1] == EVERYTHING) {
+				int i;
+				for (i = n + 1; i < wordcount; i++) {
+					wordtype[i - 1] = wordtype[i + 1];
+					wordvalue[i - 1] = wordvalue[i + 1];
+					strcpy(words[i - 1], words[i + 1]);
+				}
+				wordcount--;
+				wordcount--;
+				flag = 1;
+			}
+	}
 }