Parse equals sign as keyword

The parser treats unexpected keyword tokens (eg. if, while, ~) as literals and
not as syntax errors. It does not do this for the equals sign, but it could, as
it only has a special meaning at the beginning of a command, much like the
keywords. Doing this allows one to write commands like this:

    dd if'='/dev/zero of'='data bs'='1024 count'='1

without the quotes around the equals signs. All shells from the Bourne family
do this.

For this to work, the lexer must not automatically terminate words at equals
signs, instead it has to generate a free caret around it like for the twiddle
and other such special characters.

In the parser's grammar we need to get rid of the left recursion in the rules
for 'simple' to turn the reduce/reduce conflict at 'first' on '=' (empty ->
'optcaret' or 'first' -> 'simple') into a shift/reduce conflict, which then
can be solved with precedence/associativity so that the reduction of the
empty 'optcaret' wins over shifting '='.

3 files changed, 25 insertions, 12 deletions

diff --git a/lex.c b/lex.c
index 92732f8..195e8f6 100644
--- a/lex.c
+++ b/lex.c
@@ -178,9 +178,9 @@ top:	while ((c = gchar()) == ' ' || c == '\t')
 		y->word.q = FALSE;
 		return WORD;
 	}
-	if (c == '`' || c == '!' || c == '@' || c == '~' || c == '$' || c == '\'') {
+	if (c == '`' || c == '!' || c == '@' || c == '~' || c == '$' || c == '\'' || c == '=') {
 		checkfreecaret;
-		if (c == '!' || c == '@' || c == '~')
+		if (c == '!' || c == '@' || c == '~' || c == '=')
 			w = KW;
 	}
 	switch (c) {
@@ -255,9 +255,9 @@ top:	while ((c = gchar()) == ' ' || c == '\t')
 	case ';':
 	case '^':
 	case ')':
-	case '=':
 	case '{': case '}':
 		w = NW;
+	case '=':
 		return c;
 	case '&':
 		w = NW;
diff --git a/parse.y b/parse.y
index 8be65bb..bde299e 100644
--- a/parse.y
+++ b/parse.y
@@ -16,6 +16,7 @@ Node *parsetree;	/* not using yylval because bison declares it as an auto */
 %token ANDAND BACKBACK BANG CASE COUNT DUP ELSE END FLAT FN FOR IF IN
 %token OROR PIPE REDIR SREDIR SUB SUBSHELL SWITCH TWIDDLE WHILE WORD HUH
 
+%left '^' '='
 %left WHILE ')' ELSE
 %left ANDAND OROR '\n'
 %left BANG SUBSHELL
@@ -41,6 +42,7 @@ Node *parsetree;	/* not using yylval because bison declares it as an auto */
 %type <keyword> keyword
 %type <node> assign body brace case cbody cmd cmdsa cmdsan comword epilog
 	     first line nlwords paren redir sword simple iftail word words
+	     arg args
 
 %start rc
 
@@ -72,7 +74,7 @@ brace	: '{' body '}'		{ $$ = $2; }
 
 paren	: '(' body ')'		{ $$ = $2; }
 
-assign	: first '=' word	{ $$ = mk(nAssign,$1,$3); }
+assign	: first optcaret '=' optcaret word	{ $$ = mk(nAssign,$1,$5); }
 
 epilog	:			{ $$ = NULL; }
 	| redir epilog		{ $$ = mk(nEpilog,$1,$2); }
@@ -115,12 +117,17 @@ cmd	: /* empty */	%prec WHILE		{ $$ = NULL; }
 	| FN words brace			{ $$ = mk(nNewfn,$2,$3); }
 	| FN words				{ $$ = mk(nRmfn,$2); }
 
-optcaret : /* empty */
+optcaret : /* empty */	%prec '^'
 	| '^'
 
 simple	: first
-	| simple word			{ $$ = ($2 != NULL ? mk(nArgs,$1,$2) : $1); }
-	| simple redir			{ $$ = mk(nArgs,$1,$2); }
+	| first args			{ $$ = ($2 != NULL ? mk(nArgs,$1,$2) : $1); }
+
+args	: arg
+	| args arg			{ $$ = ($2 != NULL ? mk(nArgs,$1,$2) : $1); }
+
+arg	: word
+	| redir
 
 first	: comword
 	| first '^' sword		{ $$ = mk(nConcat,$1,$3); }
@@ -154,6 +161,7 @@ keyword	: FOR		{ $$ = "for"; }
 	| TWIDDLE	{ $$ = "~"; }
 	| BANG		{ $$ = "!"; }
 	| SUBSHELL	{ $$ = "@"; }
+	| '='		{ $$ = "="; }
 
 words	:		{ $$ = NULL; }
 	| words word	{ $$ = ($1 != NULL ? ($2 != NULL ? mk(nLappend,$1,$2) : $1) : $2); }
diff --git a/rc.1 b/rc.1
index 3a21bf5..3f16c74 100644
--- a/rc.1
+++ b/rc.1
@@ -395,7 +395,7 @@ interprets several characters specially; special characters
 automatically terminate words.
 The following characters are special:
 .Ds
-.Cr "# ; & | ^ $ = \` ' { } ( ) < >"
+.Cr "# ; & | ^ $ \` ' { } ( ) < >"
 .De
 .PP
 The single quote
@@ -1936,6 +1936,7 @@ grammar, edited to remove semantic actions.
 %term ANDAND BACKBACK BANG CASE COUNT DUP ELSE END FLAT FN FOR IF IN
 %term OROR PIPE REDIR SUB SUBSHELL SWITCH TWIDDLE WHILE WORD HUH
 
+%left '^' '='
 %left WHILE ')' ELSE
 %left ANDAND OROR '\en'
 %left BANG SUBSHELL
@@ -1964,7 +1965,7 @@ brace: '{' body '}'
 
 paren: '(' body ')'
 
-assign: first '=' word
+assign: first optcaret '=' optcaret word
 
 epilog: /* empty */ | redir epilog
 
@@ -1996,9 +1997,13 @@ cmd	: /* empty */	%prec WHILE
 	| FN words brace
 	| FN words
 
-optcaret: /* empty */ | '^'
+optcaret: /* empty */ %prec '^' | '^'
 
-simple: first | simple word | simple redir
+simple: first | first args
+
+args: arg | args arg
+
+arg: word | redir
 
 first: comword | first '^' sword
 
@@ -2018,7 +2023,7 @@ comword: '$' sword
 	| WORD
 
 keyword: FOR | IN | WHILE | IF | SWITCH
-	| FN | ELSE | CASE | TWIDDLE | BANG | SUBSHELL
+	| FN | ELSE | CASE | TWIDDLE | BANG | SUBSHELL | '='
 
 words: /* empty */ | words word