#include <u.h>
#include <libc.h>
#include <bio.h>

#include "apl9.h"

Statement *
lexline(Rune *line, int toplevel)
{
	int offset = 0;
	int len = runestrlen(line);
	Statement *stmt = emalloc(sizeof(Statement));
	stmt->ntoks = 0;
	stmt->toks = mallocz(sizeof(Datum) * MAX_LINE_TOKENS, 1);
	stmt->guard = nil;
	stmt->next = nil;

	while(offset < len){
		Rune *p;
		if(isspacerune(line[offset])){
			offset++;
			continue;
		}else if(runestrchr(L"←⋄⍝⍬", line[offset])){
			switch(line[offset]){
			case L'←': stmt->toks[stmt->ntoks].tag = ArrowTag; break;
			case L'⋄': stmt->next = lexline(&line[offset+1], toplevel); goto end;
			case L'⍝': goto end;
			case L'⍬':
				stmt->toks[stmt->ntoks].tag = ArrayTag;
				stmt->toks[stmt->ntoks].array = allocarray(AtypeInt, 1, 0);
				stmt->toks[stmt->ntoks].array->shape[0] = 0;
				break;
			}
			offset++;
		}else if(!toplevel && line[offset] == ':'){
			Rune buf[MAX_LINE_LENGTH];
			Rune *p = buf;
			offset++;
			while(line[offset] != L'⋄' && offset < len){
				*p = line[offset];
				p++;
				offset++;
			}
			*p = 0;
			stmt->guard = lexline(buf, toplevel);
			stmt->ntoks--;
		}else if(line[offset] == '{'){
			Rune buf[MAX_LINE_LENGTH];
			Rune *p = buf;
			offset++;
			while(line[offset] != '}' && offset < len){
				*p = line[offset];
				p++;
				offset++;
			}
			if(line[offset] != '}')
				goto syntax_error;
			*p = 0;
			offset++;
			stmt->toks[stmt->ntoks].tag = FunctionTag;
			stmt->toks[stmt->ntoks].func.type = FunctypeDfn;
			stmt->toks[stmt->ntoks].func.dfn = runestrdup(buf);
		}else if(line[offset] == '('){
			int unclosed = 1;
			Rune buf[MAX_LINE_LENGTH];
			Rune *p = buf;
			offset++;
			while((line[offset] != ')' || unclosed > 1) && offset < len){
				if(line[offset] == '(')
					unclosed++;
				else if(line[offset] == ')')
					unclosed--;
				*p = line[offset];
				p++;
				offset++;
			}
			if(line[offset] != ')')
				goto syntax_error;
			*p = 0;
			offset++;
			stmt->toks[stmt->ntoks].tag = LParTag;
			stmt->toks[stmt->ntoks].stmt = *lexline(buf, toplevel);
			stmt->ntoks++;
			stmt->toks[stmt->ntoks].tag = RParTag;
		}else if(p = runestrchr(primfuncnames, line[offset])){
			stmt->toks[stmt->ntoks].tag = FunctionTag;
			stmt->toks[stmt->ntoks].func.type = FunctypePrim;
			stmt->toks[stmt->ntoks].func.code = p-primfuncnames;
			offset++;
		}else if(p = runestrchr(primmonopnames, line[offset])){
			stmt->toks[stmt->ntoks].tag = MonadicOpTag;
			stmt->toks[stmt->ntoks].operator.type = OperatortypePrim;
			stmt->toks[stmt->ntoks].operator.dyadic = 0;
			stmt->toks[stmt->ntoks].operator.code = p-primmonopnames;
			offset++;
		}else if(p = runestrchr(primdyadopnames, line[offset])){
			stmt->toks[stmt->ntoks].tag = DyadicOpTag;
			stmt->toks[stmt->ntoks].operator.type = OperatortypePrim;
			stmt->toks[stmt->ntoks].operator.dyadic = 1;
			stmt->toks[stmt->ntoks].operator.code = p-primdyadopnames;
			offset++;
		}else if(p = runestrchr(primhybridnames, line[offset])){
			stmt->toks[stmt->ntoks].tag = HybridTag;
			stmt->toks[stmt->ntoks].hybrid = p-primhybridnames;
			offset++;
		}else if(isdigitrune(line[offset]) || (line[offset] == L'¯' && isdigitrune(line[offset+1]))){
			char buf[64];
			char *p = buf;
			int floating = 0;
			if(line[offset] == L'¯'){
				*p++ = '-';
				offset++;
			}
get_digits:
			while(isdigitrune(line[offset]))
				p += runetochar(p, &line[offset++]);
			if(!floating && line[offset] == '.'){
				p += runetochar(p, &line[offset++]);
				floating = 1;
				goto get_digits;
			}
			*p = 0;
			stmt->toks[stmt->ntoks].tag = ArrayTag;
			stmt->toks[stmt->ntoks].array = floating ? mkscalarfloat(atof(buf)) : mkscalarint(atoll(buf));
		}else if(runestrchr(L"⍺⍵", line[offset])){
			Rune *name = L"?";
			name[0] = line[offset];
			stmt->toks[stmt->ntoks].tag = NameTag;
			stmt->toks[stmt->ntoks].symbol = getsym(currentsymtab, name);
			offset++;
		}else if(isalpharune(line[offset])){
			Rune buf[64];
			Rune *p = buf;
			while(isalpharune(line[offset]) || isdigitrune(line[offset])){
				*p = line[offset];
				p++;
				offset++;
			}
			*p = 0;
			stmt->toks[stmt->ntoks].tag = NameTag;
			stmt->toks[stmt->ntoks].symbol = getsym(currentsymtab, buf);
		}else if(runestrchr(L"⎕⍞", line[offset])){
			/* quad names */
			Rune buf[64];
			Rune *p = buf;
			*p++ = line[offset++];
			while(isalpharune(line[offset]))
				*p++ = toupperrune(line[offset++]);
			*p = 0;
			int valid = 0;
			for(int i = 0; quadnames[i].name != nil && !valid; i++){
				if(runestrcmp(buf, quadnames[i].name) != 0)
					continue;
				valid = 1;
				stmt->toks[stmt->ntoks] = quadnamedatum(quadnames[i]);
			}
			if(!valid){
				offset -= runestrlen(buf);
				goto syntax_error;
			}
		}else if(line[offset] == '\''){
			Rune buf[1024]; /* stupid limit on literal string lengths */
			Rune *b = buf;
			int done = 0;
			offset++;
			while(!done && offset < len){
				if(line[offset] == '\'' && line[offset+1] != '\''){
					*b = 0;
					done = 1;
				}else if(line[offset] == '\'' && line[offset+1] == '\''){
					*b++ = '\'';
					offset++;
				}else
					*b++ = line[offset];
				offset++;
			}
			if(!done){
				offset = offset - (b-buf);
				goto syntax_error;
			}
			stmt->toks[stmt->ntoks].tag = ArrayTag;
			if(runestrlen(buf) == 1)
				stmt->toks[stmt->ntoks].array = mkscalarrune(buf[0]);
			else
				stmt->toks[stmt->ntoks].array = mkrunearray(buf);
		}else{
			Rune *err;
syntax_error:
			err = runesmprint("Can't lex: %S", &line[offset]);
			free(stmt->toks);
			free(stmt);
			throwerror(err, ESyntax);
		}
		stmt->ntoks++;
	}
end:
	stmt->toks = realloc(stmt->toks, sizeof(Datum) * stmt->ntoks);
	return stmt;
}