basic records support
All checks were successful
/ test (push) Successful in 50s

This commit is contained in:
Neemek 2026-08-17 17:08:24 +02:00
parent db3a3e29eb
commit 42fa039daf
11 changed files with 522 additions and 138 deletions

View file

@ -76,9 +76,13 @@ func (p ParsingError) Format() string {
}
type Parser struct {
source string
trace []string
tokens []Token
source string
trace []string
tokens []Token
state ParserState
}
type ParserState struct {
prev *Token
curr *Token
pos Pos
@ -90,7 +94,9 @@ func NewParser(source string, trace []string, tokens []Token) *Parser {
source: source,
trace: trace,
tokens: tokens,
pos: 0,
state: ParserState{
pos: 0,
},
}
}
@ -116,11 +122,11 @@ func (p *Parser) Parse(path string) (*Program, error) {
// initialize current
p.advance()
for int(p.pos) < len(p.tokens) && p.curr.Type != TokenEOF {
for int(p.state.pos) < len(p.tokens) && p.state.curr.Kind != TokenEOF {
for p.accept(TokenNewLine) {
}
if p.curr.Type == TokenEOF {
if p.state.curr.Kind == TokenEOF {
break
}
@ -139,25 +145,25 @@ func (p *Parser) Parse(path string) (*Program, error) {
&BlockNode{
statements,
0,
p.curr.End,
p.state.curr.End,
},
path,
}, nil
}
func (p *Parser) accept(tokenType TokenKind) bool {
if p.curr == nil {
if p.state.curr == nil {
log.Fatal("unexpected current token nil")
return false
}
if p.ignoreNewLine && tokenType != TokenNewLine {
for p.curr.Type == TokenNewLine {
if p.state.ignoreNewLine && tokenType != TokenNewLine {
for p.state.curr.Kind == TokenNewLine {
p.advance()
}
}
if (*p.curr).Type == tokenType {
if (*p.state.curr).Kind == tokenType {
p.advance()
return true
}
@ -165,46 +171,52 @@ func (p *Parser) accept(tokenType TokenKind) bool {
return false
}
func (p *Parser) acceptAll(tokenTypes ...TokenKind) bool {
if int(p.pos)+len(tokenTypes) > len(p.tokens) {
return false
}
func (p *Parser) getState() ParserState {
return p.state
}
for i, tokenType := range tokenTypes {
if p.tokens[int(p.pos)+i].Type != tokenType {
func (p *Parser) restoreState(state ParserState) {
p.state = state
}
func (p *Parser) acceptSeq(tokenTypes ...TokenKind) bool {
state := p.getState()
for _, t := range tokenTypes {
if !p.accept(t) {
p.restoreState(state)
return false
}
}
p.pos += Pos(len(tokenTypes))
return true
}
func (p *Parser) expect(tokenType TokenKind, reason string) error {
if !p.accept(tokenType) {
return p.error(fmt.Sprintf("Expected token %s, got %s; %s", tokenType, p.curr.Type, reason), p.curr)
return p.error(fmt.Sprintf("Expected token %s, got %s; %s", tokenType, p.state.curr.Kind, reason), p.state.curr)
}
return nil
}
func (p *Parser) peek() (Token, error) {
if p.pos >= Pos(len(p.tokens)) {
if p.state.pos >= Pos(len(p.tokens)) {
return Token{}, errors.New("cannot peek beyond tokens")
}
return p.tokens[p.pos], nil
return p.tokens[p.state.pos], nil
}
func (p *Parser) advance() {
p.prev = p.curr
p.state.prev = p.state.curr
if p.pos < Pos(len(p.tokens)) {
p.curr = &p.tokens[p.pos]
if p.state.pos < Pos(len(p.tokens)) {
p.state.curr = &p.tokens[p.state.pos]
} else {
p.curr = nil
p.state.curr = nil
}
p.pos++
p.state.pos++
}
func (p *Parser) error(error string, causer *Token) error {
@ -224,10 +236,10 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
}
}
oldIgnoreNewline := p.ignoreNewLine
p.ignoreNewLine = false
oldIgnoreNewline := p.state.ignoreNewLine
p.state.ignoreNewLine = false
start := p.prev.Start
start := p.state.prev.Start
var statements []Node
for !p.accept(TokenCloseBrace) {
@ -251,22 +263,22 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
}
}
p.ignoreNewLine = oldIgnoreNewline
p.state.ignoreNewLine = oldIgnoreNewline
return &BlockNode{statements, start, p.prev.End}, nil
return &BlockNode{statements, start, p.state.prev.End}, nil
}
t := p.curr
switch t.Type {
t := p.state.curr
switch t.Kind {
case TokenType:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
if err := p.expect(TokenName, "types must have a name"); err != nil {
return nil, err
}
name := p.prev
name := p.state.prev
if err := p.expect(TokenAssign, "type aliases must be defined with an assign"); err != nil {
return nil, err
@ -282,7 +294,7 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
sig,
start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenIf:
@ -300,7 +312,7 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
var otherwise Node
if p.accept(TokenElse) {
otherwise, err = p.expression(p.curr.Type != TokenIf)
otherwise, err = p.expression(p.state.curr.Kind != TokenIf)
if err != nil {
return nil, err
}
@ -316,7 +328,7 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
case TokenReturn:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
v, err := p.expression(false)
if err != nil {
@ -326,12 +338,12 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
return &ReturnNode{
v,
start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenWhile:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
cond, err := p.expression(false)
if err != nil {
@ -347,12 +359,12 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
cond,
logic,
start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenFor:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
counter, err := p.expression(false)
if err != nil {
@ -379,12 +391,12 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
logic,
start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenInclude:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
if err := p.expect(TokenString, "import requires a path/name to include"); err != nil {
return nil, err
@ -392,13 +404,13 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
return &IncludeNode{
&StringNode{
p.prev.Lexeme[1 : len(p.prev.Lexeme)-1],
p.prev.Lexeme,
p.prev.Start,
p.prev.End,
p.state.prev.Lexeme[1 : len(p.state.prev.Lexeme)-1],
p.state.prev.Lexeme,
p.state.prev.Start,
p.state.prev.End,
},
start,
p.prev.End,
p.state.prev.End,
}, nil
default:
@ -408,7 +420,7 @@ func (p *Parser) expression(mustBeBlock bool) (Node, error) {
}
if p.accept(TokenDeclare) || p.accept(TokenAssign) {
isDeclaration := p.prev.Type == TokenDeclare
isDeclaration := p.state.prev.Kind == TokenDeclare
// possibly assign tuples; not implemented yet
v, err := p.expression(false)
@ -509,7 +521,7 @@ func (p *Parser) binary() (Node, error) {
r := values.Pop()
l := values.Pop()
opToken := ops.Pop()
op := tokenToBinaryOperation(opToken.Type)
op := tokenToBinaryOperation(opToken.Kind)
start, _ := l.Bounds()
_, end := r.Bounds()
@ -524,12 +536,12 @@ func (p *Parser) binary() (Node, error) {
})
}
for isBinaryOperator(p.curr.Type) {
for ops.Current > 0 && binaryPrecedence(p.curr.Type) <= binaryPrecedence(ops.Peek().Type) {
for isBinaryOperator(p.state.curr.Kind) {
for ops.Current > 0 && binaryPrecedence(p.state.curr.Kind) <= binaryPrecedence(ops.Peek().Kind) {
reduce()
}
ops.Push(p.curr)
ops.Push(p.state.curr)
p.advance()
v, err := p.chain()
@ -558,16 +570,16 @@ func (p *Parser) chain() (Node, error) {
if err = p.expect(TokenName, "can only access properties by name"); err != nil {
return nil, err
}
name := p.prev
name := p.state.prev
f = &AccessNode{
f,
p.prev,
p.state.prev,
name.Start,
name.End,
}
if p.curr.Type == TokenOpenParenthesis {
if p.state.curr.Kind == TokenOpenParenthesis {
args, err := p.parseArgs()
if err != nil {
return nil, err
@ -577,11 +589,11 @@ func (p *Parser) chain() (Node, error) {
f,
args,
name.Start,
p.prev.End,
p.state.prev.End,
}
}
} else if p.curr.Type == TokenOpenParenthesis {
start := p.curr.Start
} else if p.state.curr.Kind == TokenOpenParenthesis {
start := p.state.curr.Start
args, err := p.parseArgs()
if err != nil {
return nil, err
@ -592,10 +604,10 @@ func (p *Parser) chain() (Node, error) {
args,
start,
p.prev.End,
p.state.prev.End,
}
} else if p.accept(TokenOpenBracket) {
start := p.prev.Start
start := p.state.prev.Start
index, err := p.expression(false)
if err != nil {
@ -610,7 +622,7 @@ func (p *Parser) chain() (Node, error) {
f,
index,
start,
p.prev.End,
p.state.prev.End,
}
} else {
break
@ -621,71 +633,71 @@ func (p *Parser) chain() (Node, error) {
}
func (p *Parser) factor() (Node, error) {
switch (*p.curr).Type {
switch (*p.state.curr).Kind {
case TokenString:
p.advance()
return &StringNode{
(*p.prev).Lexeme[1 : len((*p.prev).Lexeme)-1],
(*p.prev).Lexeme,
p.prev.Start,
p.prev.End,
(*p.state.prev).Lexeme[1 : len((*p.state.prev).Lexeme)-1],
(*p.state.prev).Lexeme,
p.state.prev.Start,
p.state.prev.End,
}, nil
case TokenInteger:
p.advance()
num, success := new(big.Int).SetString(p.prev.Lexeme, 10)
num, success := new(big.Int).SetString(p.state.prev.Lexeme, 10)
if !success {
return nil, p.error(fmt.Sprintf("cannot parse integer base 10: %s", p.prev.Lexeme), p.prev)
return nil, p.error(fmt.Sprintf("cannot parse integer base 10: %s", p.state.prev.Lexeme), p.state.prev)
}
return &IntegerNode{
num,
p.prev.Start,
p.prev.End,
p.state.prev.Start,
p.state.prev.End,
}, nil
case TokenFloat:
p.advance()
num, err := strconv.ParseFloat((*p.prev).Lexeme, FloatSize)
num, err := strconv.ParseFloat((*p.state.prev).Lexeme, FloatSize)
if err != nil {
return nil, p.error(fmt.Sprintf("Error parsing number: %v", err), p.prev)
return nil, p.error(fmt.Sprintf("Error parsing number: %v", err), p.state.prev)
}
return &FloatNode{
num,
p.prev.Start,
p.prev.End,
p.state.prev.Start,
p.state.prev.End,
}, nil
case TokenHexadecimal:
p.advance()
start := (*p.prev).Start
num, ok := new(big.Int).SetString(p.prev.Lexeme[2:], 16)
start := (*p.state.prev).Start
num, ok := new(big.Int).SetString(p.state.prev.Lexeme[2:], 16)
if !ok {
return nil, p.error(fmt.Sprintf("cannot parse hexadecimal: %v", p.prev.Lexeme), p.prev)
return nil, p.error(fmt.Sprintf("cannot parse hexadecimal: %v", p.state.prev.Lexeme), p.state.prev)
}
return &IntegerNode{
num,
start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenTrue:
p.advance()
return &BooleanNode{
true,
p.prev.Start,
p.prev.End,
p.state.prev.Start,
p.state.prev.End,
}, nil
case TokenFalse:
p.advance()
return &BooleanNode{
false,
p.prev.Start,
p.prev.End,
p.state.prev.Start,
p.state.prev.End,
}, nil
case TokenNil:
@ -694,7 +706,7 @@ func (p *Parser) factor() (Node, error) {
case TokenOpenBracket:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
// TODO: find better solution; current one is messy
// Maybe perform better analysis to determine the kind of the list...
@ -708,12 +720,12 @@ func (p *Parser) factor() (Node, error) {
[]Node{},
s,
start,
p.prev.End,
p.state.prev.End,
}, nil
}
oldIgnoreNewline := p.ignoreNewLine
p.ignoreNewLine = true
oldIgnoreNewline := p.state.ignoreNewLine
p.state.ignoreNewLine = true
var values []Node
for !p.accept(TokenCloseBracket) {
@ -731,19 +743,19 @@ func (p *Parser) factor() (Node, error) {
values = append(values, value)
}
p.ignoreNewLine = oldIgnoreNewline
p.state.ignoreNewLine = oldIgnoreNewline
return &ListNode{
values,
nil,
start,
p.prev.End,
p.state.prev.End,
}, nil
// unary minus
case TokenMinus:
p.advance()
op := p.prev
op := p.state.prev
f, err := p.factor()
if err != nil {
@ -754,12 +766,12 @@ func (p *Parser) factor() (Node, error) {
f,
op,
op.Start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenBang:
p.advance()
op := p.prev
op := p.state.prev
v, err := p.factor()
if err != nil {
@ -771,16 +783,16 @@ func (p *Parser) factor() (Node, error) {
v,
op,
op.Start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenName:
p.advance()
name := (*p.prev).Lexeme
start := p.prev.Start
nameEnd := p.prev.End
name := (*p.state.prev).Lexeme
start := p.state.prev.Start
nameEnd := p.state.prev.End
if p.curr.Type == TokenOpenParenthesis {
if p.state.curr.Kind == TokenOpenParenthesis {
args, err := p.parseArgs()
if err != nil {
return nil, err
@ -794,7 +806,7 @@ func (p *Parser) factor() (Node, error) {
},
args,
start,
p.prev.End,
p.state.prev.End,
}, nil
}
@ -806,11 +818,11 @@ func (p *Parser) factor() (Node, error) {
case TokenFunc:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
var name *Token
if p.accept(TokenName) { // can be unnamed, but accept name if it is named
name = p.prev
name = p.state.prev
}
params, err := p.parseParams()
@ -842,7 +854,7 @@ func (p *Parser) factor() (Node, error) {
yield,
logic,
start,
p.prev.End,
p.state.prev.End,
}
if name != nil {
@ -851,7 +863,7 @@ func (p *Parser) factor() (Node, error) {
fn,
true,
start,
p.prev.End,
p.state.prev.End,
}, nil
}
@ -859,11 +871,67 @@ func (p *Parser) factor() (Node, error) {
case TokenOpenParenthesis:
p.advance()
start := p.prev.Start
start := p.state.prev.Start
oldCare := p.state.ignoreNewLine
p.state.ignoreNewLine = true
p.skipNewLines()
// we're inside an object
if p.acceptAll(TokenName, TokenColon) {
return nil, p.error("objects are not implemented yet (TBD)", p.prev)
key := p.state.curr
if p.acceptSeq(TokenName, TokenColon) {
entries := map[string]Node{}
for len(entries) == 0 || !p.accept(TokenCloseParenthesis) {
if len(entries) != 0 {
p.skipNewLines()
key = p.state.curr
if !p.acceptSeq(TokenName, TokenColon) {
return nil, p.error("expected a record name", key)
}
}
name := key.Lexeme
if _, ok := entries[name]; ok {
return nil, p.error("duplicate key; already defined.", key)
}
if p.accept(TokenComma) || p.accept(TokenCloseParenthesis) {
entries[name] = &ReferenceNode{
name,
key.Start,
key.End,
}
if p.state.prev.Kind == TokenCloseParenthesis {
break
}
continue
} else {
x, err := p.expression(false)
if err != nil {
return nil, err
}
entries[name] = x
}
if !p.accept(TokenComma) {
if err := p.expect(TokenCloseParenthesis, "record must be closed"); err != nil {
return nil, err
}
break
}
}
p.state.ignoreNewLine = oldCare
return &RecordNode{
entries,
start,
p.state.prev.End,
}, nil
}
v, err := p.expression(false)
@ -876,6 +944,7 @@ func (p *Parser) factor() (Node, error) {
return nil, err
}
p.state.ignoreNewLine = oldCare
return v, nil
}
@ -897,26 +966,27 @@ func (p *Parser) factor() (Node, error) {
}
}
p.state.ignoreNewLine = oldCare
return &TupleNode{
items,
start,
p.prev.End,
p.state.prev.End,
}, nil
case TokenBreakpoint:
p.advance()
return &BreakpointNode{
p.prev.Start,
p.prev.End,
p.state.prev.Start,
p.state.prev.End,
}, nil
case TokenOpenBrace:
return p.expression(true)
default:
return nil, p.error(fmt.Sprintf("invalid factor %s", p.curr), p.curr)
return nil, p.error(fmt.Sprintf("invalid factor %s", p.state.curr), p.state.curr)
}
}
@ -956,7 +1026,7 @@ func (p *Parser) parseParams() ([]FunctionParameter, error) {
params := make([]FunctionParameter, 0)
if p.accept(TokenName) {
name := (*p.prev).Lexeme
name := (*p.state.prev).Lexeme
if err := p.expect(TokenColon, "parameters must have a type separated by a colon"); err != nil {
return nil, err
}
@ -977,7 +1047,7 @@ func (p *Parser) parseParams() ([]FunctionParameter, error) {
if err := p.expect(TokenName, "parameters must have a name (cannot have trailing comma)"); err != nil {
return nil, err
}
name = (*p.prev).Lexeme
name = (*p.state.prev).Lexeme
if err := p.expect(TokenColon, "parameters must have a type separated by a colon"); err != nil {
return nil, err
}
@ -1005,9 +1075,42 @@ func (p *Parser) parseSignature() (TypeSignature, error) {
var s TypeSignature
if p.accept(TokenOpenParenthesis) {
oldCare := p.state.ignoreNewLine
p.state.ignoreNewLine = true
p.skipNewLines()
// we're inside an object
if p.acceptAll(TokenName, TokenColon) {
return nil, p.error("objects are not implemented yet (TBD)", p.prev)
name := p.state.curr
if p.acceptSeq(TokenName, TokenColon) {
entries := map[string]TypeSignature{}
for len(entries) == 0 || !p.accept(TokenCloseParenthesis) {
if len(entries) != 0 {
p.skipNewLines()
name = p.state.curr
if !p.acceptSeq(TokenName, TokenColon) {
return nil, p.error("expected record member", p.state.curr)
}
}
sig, err := p.parseSignature()
if err != nil {
return nil, err
}
entries[name.Lexeme] = sig
if !p.accept(TokenComma) {
if err := p.expect(TokenCloseParenthesis, "record must be closed"); err != nil {
return nil, err
}
break
}
}
return &RecordSignature{
entries,
}, nil
}
v, err := p.parseSignature()
@ -1043,6 +1146,8 @@ func (p *Parser) parseSignature() (TypeSignature, error) {
items,
}
}
p.state.ignoreNewLine = oldCare
} else if p.accept(TokenFunc) {
if err := p.expect(TokenOpenParenthesis, "func signature must have parentheses for parameters"); err != nil {
return nil, err
@ -1095,7 +1200,7 @@ func (p *Parser) parseSignature() (TypeSignature, error) {
if err := p.expect(TokenName, "type must be a name"); err != nil {
return nil, err
}
name := (*p.prev).Lexeme
name := (*p.state.prev).Lexeme
switch name {
case "str":
@ -1130,3 +1235,8 @@ func (p *Parser) parseSignature() (TypeSignature, error) {
return s, nil
}
func (p *Parser) skipNewLines() {
for p.accept(TokenNewLine) {
}
}