diff options
Diffstat (limited to 'compiler/parser.nim')
| -rw-r--r-- | compiler/parser.nim | 396 |
1 files changed, 253 insertions, 143 deletions
diff --git a/compiler/parser.nim b/compiler/parser.nim index fd51b04ec..07f5c9de9 100644 --- a/compiler/parser.nim +++ b/compiler/parser.nim @@ -1,7 +1,7 @@ # # # The Nimrod Compiler -# (c) Copyright 2013 Andreas Rumpf +# (c) Copyright 2014 Andreas Rumpf # # See the file "copying.txt", included in this # distribution, for details about the copyright. @@ -18,10 +18,10 @@ # In fact the grammar is generated from this file: when isMainModule: import pegs - var outp = open("compiler/grammar.txt", fmWrite) + var outp = open("doc/grammar.txt", fmWrite) for line in lines("compiler/parser.nim"): if line =~ peg" \s* '#| ' {.*}": - outp.writeln matches[0] + outp.write matches[0], "\L" outp.close import @@ -31,12 +31,13 @@ type TParser*{.final.} = object # a TParser object represents a module that # is being parsed currInd: int # current indentation - firstTok: bool + firstTok, strongSpaces: bool lex*: TLexer # the lexer that is used for parsing tok*: TToken # the current token + inPragma: int + inSemiStmtList: int -proc ParseAll*(p: var TParser): PNode -proc openParser*(p: var TParser, filename: string, inputstream: PLLStream) +proc parseAll*(p: var TParser): PNode proc closeParser*(p: var TParser) proc parseTopLevelStmt*(p: var TParser): PNode # implements an iterator. Returns the next top-level statement or @@ -48,7 +49,6 @@ proc parseString*(s: string, filename: string = "", line: int = 0): PNode # correct error messages referring to the original source. # helpers for the other parsers -proc getPrecedence*(tok: TToken): int proc isOperator*(tok: TToken): bool proc getTok*(p: var TParser) proc parMessage*(p: TParser, msg: TMsgKind, arg: string = "") @@ -59,15 +59,15 @@ proc newFloatNodeP*(kind: TNodeKind, floatVal: BiggestFloat, p: TParser): PNode proc newStrNodeP*(kind: TNodeKind, strVal: string, p: TParser): PNode proc newIdentNodeP*(ident: PIdent, p: TParser): PNode proc expectIdentOrKeyw*(p: TParser) -proc ExpectIdent*(p: TParser) +proc expectIdent*(p: TParser) proc parLineInfo*(p: TParser): TLineInfo -proc Eat*(p: var TParser, TokType: TTokType) +proc eat*(p: var TParser, tokType: TTokType) proc skipInd*(p: var TParser) proc optPar*(p: var TParser) proc optInd*(p: var TParser, n: PNode) proc indAndComment*(p: var TParser, n: PNode) proc setBaseFlags*(n: PNode, base: TNumericalBase) -proc parseSymbol*(p: var TParser): PNode +proc parseSymbol*(p: var TParser, allowNil = false): PNode proc parseTry(p: var TParser): PNode proc parseCase(p: var TParser): PNode # implementation @@ -75,17 +75,20 @@ proc parseCase(p: var TParser): PNode proc getTok(p: var TParser) = rawGetTok(p.lex, p.tok) -proc OpenParser*(p: var TParser, fileIdx: int32, inputStream: PLLStream) = +proc openParser*(p: var TParser, fileIdx: int32, inputStream: PLLStream, + strongSpaces=false) = initToken(p.tok) - OpenLexer(p.lex, fileIdx, inputstream) + openLexer(p.lex, fileIdx, inputStream) getTok(p) # read the first token p.firstTok = true + p.strongSpaces = strongSpaces -proc OpenParser*(p: var TParser, filename: string, inputStream: PLLStream) = - openParser(p, filename.fileInfoIdx, inputStream) +proc openParser*(p: var TParser, filename: string, inputStream: PLLStream, + strongSpaces=false) = + openParser(p, filename.fileInfoIdx, inputstream, strongSpaces) -proc CloseParser(p: var TParser) = - CloseLexer(p.lex) +proc closeParser(p: var TParser) = + closeLexer(p.lex) proc parMessage(p: TParser, msg: TMsgKind, arg: string = "") = lexMessage(p.lex, msg, arg) @@ -135,12 +138,12 @@ proc expectIdentOrKeyw(p: TParser) = if p.tok.tokType != tkSymbol and not isKeyword(p.tok.tokType): lexMessage(p.lex, errIdentifierExpected, prettyTok(p.tok)) -proc ExpectIdent(p: TParser) = +proc expectIdent(p: TParser) = if p.tok.tokType != tkSymbol: lexMessage(p.lex, errIdentifierExpected, prettyTok(p.tok)) -proc Eat(p: var TParser, TokType: TTokType) = - if p.tok.TokType == TokType: getTok(p) +proc eat(p: var TParser, tokType: TTokType) = + if p.tok.tokType == tokType: getTok(p) else: lexMessage(p.lex, errTokenExpected, TokTypeToStr[tokType]) proc parLineInfo(p: TParser): TLineInfo = @@ -185,40 +188,58 @@ proc relevantOprChar(ident: PIdent): char {.inline.} = if result == '\\' and L > 1: result = ident.s[1] -proc IsSigilLike(tok: TToken): bool {.inline.} = +proc isSigilLike(tok: TToken): bool {.inline.} = result = tok.tokType == tkOpr and relevantOprChar(tok.ident) == '@' -proc IsLeftAssociative(tok: TToken): bool {.inline.} = +proc isLeftAssociative(tok: TToken): bool {.inline.} = result = tok.tokType != tkOpr or relevantOprChar(tok.ident) != '^' -proc getPrecedence(tok: TToken): int = +proc getPrecedence(tok: TToken, strongSpaces: bool): int = + template considerStrongSpaces(x): expr = + x + (if strongSpaces: 100 - tok.strongSpaceA.int*10 else: 0) + case tok.tokType of tkOpr: let L = tok.ident.s.len let relevantChar = relevantOprChar(tok.ident) - template considerAsgn(value: expr) = - result = if tok.ident.s[L-1] == '=': 1 else: value + template considerAsgn(value: expr) = + result = if tok.ident.s[L-1] == '=': 1 else: considerStrongSpaces(value) case relevantChar of '$', '^': considerAsgn(10) of '*', '%', '/', '\\': considerAsgn(9) - of '~': result = 8 + of '~': result = considerStrongSpaces(8) of '+', '-', '|': considerAsgn(8) of '&': considerAsgn(7) - of '=', '<', '>', '!': result = 5 + of '=', '<', '>', '!': result = considerStrongSpaces(5) of '.': considerAsgn(6) - of '?': result = 2 + of '?': result = considerStrongSpaces(2) else: considerAsgn(2) of tkDiv, tkMod, tkShl, tkShr: result = 9 - of tkIn, tkNotIn, tkIs, tkIsNot, tkNot, tkOf, tkAs: result = 5 - of tkDotDot: result = 6 + of tkIn, tkNotin, tkIs, tkIsnot, tkNot, tkOf, tkAs: result = 5 + of tkDotDot: result = considerStrongSpaces(6) of tkAnd: result = 4 of tkOr, tkXor: result = 3 - else: result = - 10 - -proc isOperator(tok: TToken): bool = - result = getPrecedence(tok) >= 0 + else: result = -10 + +proc isOperator(tok: TToken): bool = + tok.tokType in {tkOpr, tkDiv, tkMod, tkShl, tkShr, tkIn, tkNotin, tkIs, + tkIsnot, tkNot, tkOf, tkAs, tkDotDot, tkAnd, tkOr, tkXor} + +proc isUnary(p: TParser): bool = + p.strongSpaces and p.tok.tokType in {tkOpr, tkDotDot} and + p.tok.strongSpaceB == 0 and + p.tok.strongSpaceA > 0 + +proc checkBinary(p: TParser) {.inline.} = + # we don't check '..' here as that's too annoying + if p.strongSpaces and p.tok.tokType == tkOpr: + if p.tok.strongSpaceB > 0 and p.tok.strongSpaceA != p.tok.strongSpaceB: + parMessage(p, errGenerated, "number of spaces around '$#' not consistent"% + prettyTok(p.tok)) + elif p.tok.strongSpaceA notin {0,1,2,4,8}: + parMessage(p, errGenerated, "number of spaces must be 0,1,2,4 or 8") #| module = stmt ^* (';' / IND{=}) #| @@ -252,7 +273,7 @@ proc colcom(p: var TParser, n: PNode) = eat(p, tkColon) skipComment(p, n) -proc parseSymbol(p: var TParser): PNode = +proc parseSymbol(p: var TParser, allowNil = false): PNode = #| symbol = '`' (KEYW|IDENT|operator|'(' ')'|'[' ']'|'{' '}'|'='|literal)+ '`' #| | IDENT case p.tok.tokType @@ -279,7 +300,7 @@ proc parseSymbol(p: var TParser): PNode = add(result, newIdentNodeP(getIdent"{}", p)) getTok(p) eat(p, tkCurlyRi) - of tokKeywordLow..tokKeywordHigh, tkSymbol, tkOpr, tkDotDot: + of tokKeywordLow..tokKeywordHigh, tkSymbol, tkOpr, tkDot, tkDotDot: add(result, newIdentNodeP(p.tok.ident, p)) getTok(p) of tkIntLit..tkCharLit: @@ -291,9 +312,13 @@ proc parseSymbol(p: var TParser): PNode = break eat(p, tkAccent) else: - parMessage(p, errIdentifierExpected, p.tok) - getTok(p) # BUGFIX: We must consume a token here to prevent endless loops! - result = ast.emptyNode + if allowNil and p.tok.tokType == tkNil: + result = newNodeP(nkNilLit, p) + getTok(p) + else: + parMessage(p, errIdentifierExpected, p.tok) + getTok(p) # BUGFIX: We must consume a token here to prevent endless loops! + result = ast.emptyNode proc indexExpr(p: var TParser): PNode = #| indexExpr = expr @@ -427,7 +452,7 @@ proc parseCast(p: var TParser): PNode = proc setBaseFlags(n: PNode, base: TNumericalBase) = case base - of base10: nil + of base10: discard of base2: incl(n.flags, nfBase2) of base8: incl(n.flags, nfBase8) of base16: incl(n.flags, nfBase16) @@ -454,11 +479,13 @@ proc complexOrSimpleStmt(p: var TParser): PNode proc simpleExpr(p: var TParser, mode = pmNormal): PNode proc semiStmtList(p: var TParser, result: PNode) = + inc p.inSemiStmtList result.add(complexOrSimpleStmt(p)) - while p.tok.tokType == tkSemicolon: + while p.tok.tokType == tkSemiColon: getTok(p) optInd(p, result) result.add(complexOrSimpleStmt(p)) + dec p.inSemiStmtList result.kind = nkStmtListExpr proc parsePar(p: var TParser): PNode = @@ -482,7 +509,7 @@ proc parsePar(p: var TParser): PNode = # XXX 'bind' used to be an expression, so we exclude it here; # tests/reject/tbind2 fails otherwise. semiStmtList(p, result) - elif p.tok.tokType == tkSemicolon: + elif p.tok.tokType == tkSemiColon: # '(;' enforces 'stmt' context: getTok(p) optInd(p, result) @@ -498,7 +525,9 @@ proc parsePar(p: var TParser): PNode = asgn.sons[0] = a asgn.sons[1] = b result.add(asgn) - elif p.tok.tokType == tkSemicolon: + if p.tok.tokType == tkSemiColon: + semiStmtList(p, result) + elif p.tok.tokType == tkSemiColon: # stmt context: result.add(a) semiStmtList(p, result) @@ -518,14 +547,14 @@ proc parsePar(p: var TParser): PNode = eat(p, tkParRi) proc identOrLiteral(p: var TParser, mode: TPrimaryMode): PNode = + #| literal = | INT_LIT | INT8_LIT | INT16_LIT | INT32_LIT | INT64_LIT + #| | UINT_LIT | UINT8_LIT | UINT16_LIT | UINT32_LIT | UINT64_LIT + #| | FLOAT_LIT | FLOAT32_LIT | FLOAT64_LIT + #| | STR_LIT | RSTR_LIT | TRIPLESTR_LIT + #| | CHAR_LIT + #| | NIL #| generalizedLit = GENERALIZED_STR_LIT | GENERALIZED_TRIPLESTR_LIT - #| identOrLiteral = generalizedLit | symbol - #| | INT_LIT | INT8_LIT | INT16_LIT | INT32_LIT | INT64_LIT - #| | UINT_LIT | UINT8_LIT | UINT16_LIT | UINT32_LIT | UINT64_LIT - #| | FLOAT_LIT | FLOAT32_LIT | FLOAT64_LIT - #| | STR_LIT | RSTR_LIT | TRIPLESTR_LIT - #| | CHAR_LIT - #| | NIL + #| identOrLiteral = generalizedLit | symbol | literal #| | par | arrayConstr | setOrTableConstr #| | castExpr #| tupleConstr = '(' optInd (exprColonEqExpr comma?)* optPar ')' @@ -634,16 +663,20 @@ proc namedParams(p: var TParser, callee: PNode, addSon(result, a) exprColonEqExprListAux(p, endTok, result) -proc primarySuffix(p: var TParser, r: PNode): PNode = +proc parseMacroColon(p: var TParser, x: PNode): PNode +proc primarySuffix(p: var TParser, r: PNode, baseIndent: int): PNode = #| primarySuffix = '(' (exprColonEqExpr comma?)* ')' doBlocks? #| | doBlocks #| | '.' optInd ('type' | 'addr' | symbol) generalizedLit? #| | '[' optInd indexExprList optPar ']' #| | '{' optInd indexExprList optPar '}' + #| | &( '`'|IDENT|literal|'cast') expr # command syntax result = r - while p.tok.indent < 0: + while p.tok.indent < 0 or + (p.tok.tokType == tkDot and p.tok.indent >= baseIndent): case p.tok.tokType of tkParLe: + if p.strongSpaces and p.tok.strongSpaceA > 0: break result = namedParams(p, result, nkCall, tkParRi) if result.len > 1 and result.sons[1].kind == nkExprColonExpr: result.kind = nkObjConstr @@ -658,21 +691,47 @@ proc primarySuffix(p: var TParser, r: PNode): PNode = result = dotExpr(p, result) result = parseGStrLit(p, result) of tkBracketLe: + if p.strongSpaces and p.tok.strongSpaceA > 0: break result = namedParams(p, result, nkBracketExpr, tkBracketRi) of tkCurlyLe: + if p.strongSpaces and p.tok.strongSpaceA > 0: break result = namedParams(p, result, nkCurlyExpr, tkCurlyRi) - else: break - + of tkSymbol, tkAccent, tkIntLit..tkCharLit, tkNil, tkCast: + if p.inPragma == 0: + # actually parsing {.push hints:off.} as {.push(hints:off).} is a sweet + # solution, but pragmas.nim can't handle that + let a = result + result = newNodeP(nkCommand, p) + addSon(result, a) + addSon result, parseExpr(p) + when false: + while p.tok.tokType != tkEof: + let a = parseExpr(p) + addSon(result, a) + if p.tok.tokType != tkComma: break + getTok(p) + optInd(p, a) + if p.tok.tokType == tkDo: + parseDoBlocks(p, result) + else: + result = parseMacroColon(p, result) + break + else: + break + proc primary(p: var TParser, mode: TPrimaryMode): PNode +proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode -proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode = - result = primary(p, mode) +proc parseOperators(p: var TParser, headNode: PNode, + limit: int, mode: TPrimaryMode): PNode = + result = headNode # expand while operators have priorities higher than 'limit' - var opPrec = getPrecedence(p.tok) + var opPrec = getPrecedence(p.tok, p.strongSpaces) let modeB = if mode == pmTypeDef: pmTypeDesc else: mode # the operator itself must not start on a new line: - while opPrec >= limit and p.tok.indent < 0: - var leftAssoc = ord(IsLeftAssociative(p.tok)) + while opPrec >= limit and p.tok.indent < 0 and not isUnary(p): + checkBinary(p) + var leftAssoc = ord(isLeftAssociative(p.tok)) var a = newNodeP(nkInfix, p) var opNode = newIdentNodeP(p.tok.ident, p) # skip operator: getTok(p) @@ -683,7 +742,11 @@ proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode = addSon(a, result) addSon(a, b) result = a - opPrec = getPrecedence(p.tok) + opPrec = getPrecedence(p.tok, p.strongSpaces) + +proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode = + result = primary(p, mode) + result = parseOperators(p, result, limit, mode) proc simpleExpr(p: var TParser, mode = pmNormal): PNode = result = simpleExprAux(p, -1, mode) @@ -713,6 +776,7 @@ proc parseIfExpr(p: var TParser, kind: TNodeKind): PNode = proc parsePragma(p: var TParser): PNode = #| pragma = '{.' optInd (exprColonExpr comma?)* optPar ('.}' | '}') result = newNodeP(nkPragma, p) + inc p.inPragma getTok(p) optInd(p, result) while p.tok.tokType notin {tkCurlyDotRi, tkCurlyRi, tkEof}: @@ -724,6 +788,7 @@ proc parsePragma(p: var TParser): PNode = optPar(p) if p.tok.tokType in {tkCurlyDotRi, tkCurlyRi}: getTok(p) else: parMessage(p, errTokenExpected, ".}") + dec p.inPragma proc identVis(p: var TParser): PNode = #| identVis = symbol opr? # postfix position @@ -798,7 +863,7 @@ proc parseTuple(p: var TParser, indentAllowed = false): PNode = while p.tok.tokType in {tkSymbol, tkAccent}: var a = parseIdentColonEquals(p, {}) addSon(result, a) - if p.tok.tokType notin {tkComma, tkSemicolon}: break + if p.tok.tokType notin {tkComma, tkSemiColon}: break getTok(p) skipComment(p, a) optPar(p) @@ -840,13 +905,13 @@ proc parseParamList(p: var TParser, retColon = true): PNode = parMessage(p, errTokenExpected, ")") break addSon(result, a) - if p.tok.tokType notin {tkComma, tkSemicolon}: break + if p.tok.tokType notin {tkComma, tkSemiColon}: break getTok(p) skipComment(p, a) optPar(p) eat(p, tkParRi) let hasRet = if retColon: p.tok.tokType == tkColon - else: p.tok.tokType == tkOpr and IdentEq(p.tok.ident, "->") + else: p.tok.tokType == tkOpr and identEq(p.tok.ident, "->") if hasRet and p.tok.indent < 0: getTok(p) optInd(p, result) @@ -905,14 +970,30 @@ proc isExprStart(p: TParser): bool = tkTuple, tkObject, tkType, tkWhen, tkCase, tkShared: result = true else: result = false - -proc parseTypeDescKAux(p: var TParser, kind: TNodeKind, - mode: TPrimaryMode): PNode = + +proc parseSymbolList(p: var TParser, result: PNode, allowNil = false) = + while true: + var s = parseSymbol(p, allowNil) + if s.kind == nkEmpty: break + addSon(result, s) + if p.tok.tokType != tkComma: break + getTok(p) + optInd(p, s) + +proc parseTypeDescKAux(p: var TParser, kind: TNodeKind, + mode: TPrimaryMode): PNode = result = newNodeP(kind, p) getTok(p) optInd(p, result) if not isOperator(p.tok) and isExprStart(p): addSon(result, primary(p, mode)) + if kind == nkDistinctTy and p.tok.tokType in {tkWith, tkWithout}: + let nodeKind = if p.tok.tokType == tkWith: nkWith + else: nkWithout + getTok(p) + let list = newNodeP(nodeKind, p) + result.addSon list + parseSymbolList(p, list, allowNil = true) proc parseExpr(p: var TParser): PNode = #| expr = (ifExpr @@ -929,7 +1010,6 @@ proc parseExpr(p: var TParser): PNode = proc parseEnum(p: var TParser): PNode proc parseObject(p: var TParser): PNode -proc parseDistinct(p: var TParser): PNode proc parseTypeClass(p: var TParser): PNode proc primary(p: var TParser, mode: TPrimaryMode): PNode = @@ -941,7 +1021,7 @@ proc primary(p: var TParser, mode: TPrimaryMode): PNode = #| / 'static' primary #| / 'bind' primary if isOperator(p.tok): - let isSigil = IsSigilLike(p.tok) + let isSigil = isSigilLike(p.tok) result = newNodeP(nkPrefix, p) var a = newIdentNodeP(p.tok.ident, p) addSon(result, a) @@ -949,8 +1029,9 @@ proc primary(p: var TParser, mode: TPrimaryMode): PNode = optInd(p, a) if isSigil: #XXX prefix operators + let baseInd = p.lex.currLineIndent addSon(result, primary(p, pmSkipSuffix)) - result = primarySuffix(p, result) + result = primarySuffix(p, result, baseInd) else: addSon(result, primary(p, pmNormal)) return @@ -965,14 +1046,19 @@ proc primary(p: var TParser, mode: TPrimaryMode): PNode = of tkTuple: result = parseTuple(p, mode == pmTypeDef) of tkProc: result = parseProcExpr(p, mode notin {pmTypeDesc, pmTypeDef}) of tkIterator: - if mode in {pmTypeDesc, pmTypeDef}: - result = parseProcExpr(p, false) - result.kind = nkIteratorTy + when false: + if mode in {pmTypeDesc, pmTypeDef}: + result = parseProcExpr(p, false) + result.kind = nkIteratorTy + else: + # no anon iterators for now: + parMessage(p, errExprExpected, p.tok) + getTok(p) # we must consume a token here to prevend endless loops! + result = ast.emptyNode else: - # no anon iterators for now: - parMessage(p, errExprExpected, p.tok) - getTok(p) # we must consume a token here to prevend endless loops! - result = ast.emptyNode + result = parseProcExpr(p, mode notin {pmTypeDesc, pmTypeDef}) + if result.kind == nkLambda: result.kind = nkIteratorDef + else: result.kind = nkIteratorTy of tkEnum: if mode == pmTypeDef: result = parseEnum(p) @@ -995,18 +1081,23 @@ proc primary(p: var TParser, mode: TPrimaryMode): PNode = getTokNoInd(p) addSon(result, primary(p, pmNormal)) of tkStatic: - result = newNodeP(nkStaticExpr, p) + let info = parLineInfo(p) getTokNoInd(p) - addSon(result, primary(p, pmNormal)) + let next = primary(p, pmNormal) + if next.kind == nkBracket and next.sonsLen == 1: + result = newNode(nkStaticTy, info, @[next.sons[0]]) + else: + result = newNode(nkStaticExpr, info, @[next]) of tkBind: result = newNodeP(nkBind, p) getTok(p) optInd(p, result) addSon(result, primary(p, pmNormal)) else: + let baseInd = p.lex.currLineIndent result = identOrLiteral(p, mode) if mode != pmSkipSuffix: - result = primarySuffix(p, result) + result = primarySuffix(p, result, baseInd) proc parseTypeDesc(p: var TParser): PNode = #| typeDesc = simpleExpr @@ -1014,6 +1105,7 @@ proc parseTypeDesc(p: var TParser): PNode = proc parseTypeDefAux(p: var TParser): PNode = #| typeDefAux = simpleExpr + #| | 'generic' typeClass result = simpleExpr(p, pmTypeDef) proc makeCall(n: PNode): PNode = @@ -1023,17 +1115,54 @@ proc makeCall(n: PNode): PNode = result = newNodeI(nkCall, n.info) result.add n +proc parseMacroColon(p: var TParser, x: PNode): PNode = + #| macroColon = ':' stmt? ( IND{=} 'of' exprList ':' stmt + #| | IND{=} 'elif' expr ':' stmt + #| | IND{=} 'except' exprList ':' stmt + #| | IND{=} 'else' ':' stmt )* + result = x + if p.tok.tokType == tkColon and p.tok.indent < 0: + result = makeCall(result) + getTok(p) + skipComment(p, result) + if p.tok.tokType notin {tkOf, tkElif, tkElse, tkExcept}: + let body = parseStmt(p) + addSon(result, newProcNode(nkDo, body.info, body)) + while sameInd(p): + var b: PNode + case p.tok.tokType + of tkOf: + b = newNodeP(nkOfBranch, p) + exprList(p, tkColon, b) + of tkElif: + b = newNodeP(nkElifBranch, p) + getTok(p) + optInd(p, b) + addSon(b, parseExpr(p)) + eat(p, tkColon) + of tkExcept: + b = newNodeP(nkExceptBranch, p) + exprList(p, tkColon, b) + skipComment(p, b) + of tkElse: + b = newNodeP(nkElse, p) + getTok(p) + eat(p, tkColon) + else: break + addSon(b, parseStmt(p)) + addSon(result, b) + if b.kind == nkElse: break + proc parseExprStmt(p: var TParser): PNode = #| exprStmt = simpleExpr #| (( '=' optInd expr ) #| / ( expr ^+ comma #| doBlocks - #| / ':' stmt? ( IND{=} 'of' exprList ':' stmt - #| | IND{=} 'elif' expr ':' stmt - #| | IND{=} 'except' exprList ':' stmt - #| | IND{=} 'else' ':' stmt )* + #| / macroColon #| ))? + inc p.inPragma var a = simpleExpr(p) + dec p.inPragma if p.tok.tokType == tkEquals: getTok(p) optInd(p, result) @@ -1056,37 +1185,7 @@ proc parseExprStmt(p: var TParser): PNode = result = makeCall(result) parseDoBlocks(p, result) return result - if p.tok.tokType == tkColon and p.tok.indent < 0: - result = makeCall(result) - getTok(p) - skipComment(p, result) - if p.tok.TokType notin {tkOf, tkElif, tkElse, tkExcept}: - let body = parseStmt(p) - addSon(result, newProcNode(nkDo, body.info, body)) - while sameInd(p): - var b: PNode - case p.tok.tokType - of tkOf: - b = newNodeP(nkOfBranch, p) - exprList(p, tkColon, b) - of tkElif: - b = newNodeP(nkElifBranch, p) - getTok(p) - optInd(p, b) - addSon(b, parseExpr(p)) - eat(p, tkColon) - of tkExcept: - b = newNodeP(nkExceptBranch, p) - exprList(p, tkColon, b) - skipComment(p, b) - of tkElse: - b = newNodeP(nkElse, p) - getTok(p) - eat(p, tkColon) - else: break - addSon(b, parseStmt(p)) - addSon(result, b) - if b.kind == nkElse: break + result = parseMacroColon(p, result) proc parseModuleName(p: var TParser, kind: TNodeKind): PNode = result = parseExpr(p) @@ -1169,8 +1268,7 @@ proc parseReturnOrRaise(p: var TParser, kind: TNodeKind): PNode = if p.tok.tokType == tkComment: skipComment(p, result) addSon(result, ast.emptyNode) - elif p.tok.indent >= 0 and p.tok.indent <= p.currInd or - p.tok.tokType == tkEof: + elif p.tok.indent >= 0 and p.tok.indent <= p.currInd or not isExprStart(p): # NL terminates: addSon(result, ast.emptyNode) else: @@ -1378,7 +1476,7 @@ proc parseGenericParamList(p: var TParser): PNode = while p.tok.tokType in {tkSymbol, tkAccent}: var a = parseGenericParam(p) addSon(result, a) - if p.tok.tokType notin {tkComma, tkSemicolon}: break + if p.tok.tokType notin {tkComma, tkSemiColon}: break getTok(p) skipComment(p, a) optPar(p) @@ -1433,7 +1531,7 @@ proc parseSection(p: var TParser, kind: TNodeKind, defparser: TDefParser): PNode = #| section(p) = COMMENT? p / (IND{>} (p / COMMENT)^+IND{=} DED) result = newNodeP(kind, p) - getTok(p) + if kind != nkTypeSection: getTok(p) skipComment(p, result) if realInd(p): withInd(p): @@ -1633,12 +1731,15 @@ proc parseTypeClassParam(p: var TParser): PNode = result = p.parseSymbol proc parseTypeClass(p: var TParser): PNode = + #| typeClassParam = ('var')? symbol + #| typeClass = typeClassParam ^* ',' (pragma)? ('of' typeDesc ^* ',')? + #| &IND{>} stmt result = newNodeP(nkTypeClassTy, p) getTok(p) var args = newNode(nkArgList) addSon(result, args) addSon(args, p.parseTypeClassParam) - while p.tok.TokType == tkComma: + while p.tok.tokType == tkComma: getTok(p) addSon(args, p.parseTypeClassParam) if p.tok.tokType == tkCurlyDotLe and p.validInd: @@ -1663,13 +1764,6 @@ proc parseTypeClass(p: var TParser): PNode = else: addSon(result, parseStmt(p)) -proc parseDistinct(p: var TParser): PNode = - #| distinct = 'distinct' optInd typeDesc - result = newNodeP(nkDistinctTy, p) - getTok(p) - optInd(p, result) - addSon(result, parseTypeDesc(p)) - proc parseTypeDef(p: var TParser): PNode = #| typeDef = identWithPragma genericParamList? '=' optInd typeDefAux #| indAndComment? @@ -1791,7 +1885,16 @@ proc complexOrSimpleStmt(p: var TParser): PNode = of tkMacro: result = parseRoutine(p, nkMacroDef) of tkTemplate: result = parseRoutine(p, nkTemplateDef) of tkConverter: result = parseRoutine(p, nkConverterDef) - of tkType: result = parseSection(p, nkTypeSection, parseTypeDef) + of tkType: + getTok(p) + if p.tok.tokType == tkParLe: + getTok(p) + result = newNodeP(nkTypeOfExpr, p) + result.addSon(primary(p, pmTypeDesc)) + eat(p, tkParRi) + result = parseOperators(p, result, -1, pmNormal) + else: + result = parseSection(p, nkTypeSection, parseTypeDef) of tkConst: result = parseSection(p, nkConstSection, parseConstant) of tkLet: result = parseSection(p, nkLetSection, parseVariable) of tkWhen: result = parseIfOrWhen(p, nkWhenStmt) @@ -1809,16 +1912,16 @@ proc parseStmt(p: var TParser): PNode = withInd(p): while true: if p.tok.indent == p.currInd: - nil - elif p.tok.tokType == tkSemicolon: + discard + elif p.tok.tokType == tkSemiColon: getTok(p) if p.tok.indent < 0 or p.tok.indent == p.currInd: discard else: break else: - if p.tok.indent > p.currInd: + if p.tok.indent > p.currInd and p.tok.tokType != tkDot: parMessage(p, errInvalidIndentation) break - if p.tok.toktype in {tkCurlyRi, tkParRi, tkCurlyDotRi, tkBracketRi}: + if p.tok.tokType in {tkCurlyRi, tkParRi, tkCurlyDotRi, tkBracketRi}: # XXX this ensures tnamedparamanonproc still compiles; # deprecate this syntax later break @@ -1836,14 +1939,19 @@ proc parseStmt(p: var TParser): PNode = parMessage(p, errComplexStmtRequiresInd) result = ast.emptyNode else: - result = newNodeP(nkStmtList, p) - while true: - if p.tok.indent >= 0: parMessage(p, errInvalidIndentation) - let a = simpleStmt(p) - if a.kind == nkEmpty: parMessage(p, errExprExpected, p.tok) - result.add(a) - if p.tok.tokType != tkSemicolon: break - getTok(p) + if p.inSemiStmtList > 0: + result = simpleStmt(p) + if result.kind == nkEmpty: parMessage(p, errExprExpected, p.tok) + else: + result = newNodeP(nkStmtList, p) + while true: + if p.tok.indent >= 0: + parMessage(p, errInvalidIndentation) + let a = simpleStmt(p) + if a.kind == nkEmpty: parMessage(p, errExprExpected, p.tok) + result.add(a) + if p.tok.tokType != tkSemiColon: break + getTok(p) proc parseAll(p: var TParser): PNode = result = newNodeP(nkStmtList, p) @@ -1862,11 +1970,11 @@ proc parseTopLevelStmt(p: var TParser): PNode = result = ast.emptyNode while true: if p.tok.indent != 0: - if p.firstTok and p.tok.indent < 0: nil + if p.firstTok and p.tok.indent < 0: discard else: parMessage(p, errInvalidIndentation) p.firstTok = false case p.tok.tokType - of tkSemicolon: + of tkSemiColon: getTok(p) if p.tok.indent <= 0: discard else: parMessage(p, errInvalidIndentation) @@ -1877,11 +1985,13 @@ proc parseTopLevelStmt(p: var TParser): PNode = break proc parseString(s: string, filename: string = "", line: int = 0): PNode = - var stream = LLStreamOpen(s) + var stream = llStreamOpen(s) stream.lineOffset = line var parser: TParser - OpenParser(parser, filename, stream) + # XXX for now the builtin 'parseStmt/Expr' functions do not know about strong + # spaces... + openParser(parser, filename, stream, false) result = parser.parseAll - CloseParser(parser) + closeParser(parser) |