diff options
Diffstat (limited to 'compiler/parser.nim')
| -rw-r--r-- | compiler/parser.nim | 1376 |
1 files changed, 799 insertions, 577 deletions
diff --git a/compiler/parser.nim b/compiler/parser.nim index 4ea751057..747505097 100644 --- a/compiler/parser.nim +++ b/compiler/parser.nim @@ -16,79 +16,112 @@ # In fact the grammar is generated from this file: -when isMainModule: +when isMainModule or defined(nimTestGrammar): # Leave a note in grammar.txt that it is generated: #| # This file is generated by compiler/parser.nim. - import pegs - var outp = open("doc/grammar.txt", fmWrite) - for line in lines("compiler/parser.nim"): - if line =~ peg" \s* '#| ' {.*}": - outp.write matches[0], "\L" - outp.close - - import ".." / tools / grammar_nanny - checkGrammarFile() + import std/pegs + when defined(nimPreviewSlimSystem): + import std/syncio + + proc writeGrammarFile(x: string) = + var outp = open(x, fmWrite) + for line in lines("compiler/parser.nim"): + if line =~ peg" \s* '#| ' {.*}": + outp.write matches[0], "\L" + outp.close + + when defined(nimTestGrammar): + import std/os + from ../testament/lib/stdtest/specialpaths import buildDir + const newGrammarText = buildDir / "grammar.txt" + + if not dirExists(buildDir): + createDir(buildDir) + + writeGrammarFile(newGrammarText) + + proc checkSameGrammar*() = + doAssert sameFileContent(newGrammarText, "doc/grammar.txt"), + "execute 'nim r compiler/parser.nim' to keep grammar.txt up-to-date" + else: + writeGrammarFile("doc/grammar.txt") + import ".." / tools / grammar_nanny + checkGrammarFile() import - llstream, lexer, idents, strutils, ast, msgs, options, lineinfos, + llstream, lexer, idents, msgs, options, lineinfos, pathutils +when not defined(nimCustomAst): + import ast +else: + import plugins / customast + +import std/strutils + when defined(nimpretty): import layouter +when defined(nimPreviewSlimSystem): + import std/assertions + type - TParser* = object # A TParser object represents a file that - # is being parsed - currInd: int # current indentation level - firstTok: bool # Has the first token been read? - hasProgress: bool # some while loop requires progress ensurance - lex*: TLexer # The lexer that is used for parsing - tok*: TToken # The current token - inPragma*: int # Pragma level + Parser* = object # A Parser object represents a file that + # is being parsed + currInd: int # current indentation level + firstTok: bool # Has the first token been read? + hasProgress: bool # some while loop requires progress ensurance + lex*: Lexer # The lexer that is used for parsing + tok*: Token # The current token + lineStartPrevious*: int + lineNumberPrevious*: int + bufposPrevious*: int + inPragma*: int # Pragma level inSemiStmtList*: int - emptyNode: PNode + when not defined(nimCustomAst): + emptyNode: PNode when defined(nimpretty): em*: Emitter SymbolMode = enum smNormal, smAllowNil, smAfterDot - TPrimaryMode = enum - pmNormal, pmTypeDesc, pmTypeDef, pmSkipSuffix + PrimaryMode = enum + pmNormal, pmTypeDesc, pmTypeDef, pmTrySimple -proc parseAll*(p: var TParser): PNode -proc closeParser*(p: var TParser) -proc parseTopLevelStmt*(p: var TParser): PNode +when defined(nimCustomAst): + # For the `customast` version we cannot share nodes, not even empty nodes: + template emptyNode(p: Parser): PNode = newNode(nkEmpty) # helpers for the other parsers -proc isOperator*(tok: TToken): bool -proc getTok*(p: var TParser) -proc parMessage*(p: TParser, msg: TMsgKind, arg: string = "") -proc skipComment*(p: var TParser, node: PNode) -proc newNodeP*(kind: TNodeKind, p: TParser): PNode -proc newIntNodeP*(kind: TNodeKind, intVal: BiggestInt, p: TParser): PNode -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 parLineInfo*(p: TParser): TLineInfo -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, mode = smNormal): PNode -proc parseTry(p: var TParser; isExpr: bool): PNode -proc parseCase(p: var TParser): PNode -proc parseStmtPragma(p: var TParser): PNode -proc parsePragma(p: var TParser): PNode -proc postExprBlocks(p: var TParser, x: PNode): PNode -proc parseExprStmt(p: var TParser): PNode -proc parseBlock(p: var TParser): PNode -proc primary(p: var TParser, mode: TPrimaryMode): PNode -proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode +proc isOperator*(tok: Token): bool +proc getTok*(p: var Parser) +proc parMessage*(p: Parser, msg: TMsgKind, arg: string = "") +proc skipComment*(p: var Parser, node: PNode) +proc newNodeP*(kind: TNodeKind, p: Parser): PNode +proc newIntNodeP*(kind: TNodeKind, intVal: BiggestInt, p: Parser): PNode +proc newFloatNodeP*(kind: TNodeKind, floatVal: BiggestFloat, p: Parser): PNode +proc newStrNodeP*(kind: TNodeKind, strVal: sink string, p: Parser): PNode +proc newIdentNodeP*(ident: PIdent, p: Parser): PNode +proc expectIdentOrKeyw*(p: Parser) +proc expectIdent*(p: Parser) +proc parLineInfo*(p: Parser): TLineInfo +proc eat*(p: var Parser, tokType: TokType) +proc skipInd*(p: var Parser) +proc optPar*(p: var Parser) +proc optInd*(p: var Parser, n: PNode) +proc indAndComment*(p: var Parser, n: PNode, maybeMissEquals = false) +proc setBaseFlags*(n: PNode, base: NumericalBase) +proc parseSymbol*(p: var Parser, mode = smNormal): PNode +proc parseTry(p: var Parser; isExpr: bool): PNode +proc parseCase(p: var Parser): PNode +proc parseStmtPragma(p: var Parser): PNode +proc parsePragma(p: var Parser): PNode +proc postExprBlocks(p: var Parser, x: PNode): PNode +proc parseExprStmt(p: var Parser): PNode +proc parseBlock(p: var Parser): PNode +proc primary(p: var Parser, mode: PrimaryMode): PNode +proc simpleExprAux(p: var Parser, limit: int, mode: PrimaryMode): PNode # implementation @@ -97,9 +130,12 @@ template prettySection(body) = body when defined(nimpretty): endSection(p.em) -proc getTok(p: var TParser) = +proc getTok(p: var Parser) = ## Get the next token from the parser's lexer, and store it in the parser's ## `tok` member. + p.lineNumberPrevious = p.lex.lineNumber + p.lineStartPrevious = p.lex.lineStart + p.bufposPrevious = p.lex.bufpos rawGetTok(p.lex, p.tok) p.hasProgress = true when defined(nimpretty): @@ -110,37 +146,36 @@ proc getTok(p: var TParser) = rawGetTok(p.lex, p.tok) emitTok(p.em, p.lex, p.tok) -proc openParser*(p: var TParser, fileIdx: FileIndex, inputStream: PLLStream, +proc openParser*(p: var Parser, fileIdx: FileIndex, inputStream: PLLStream, cache: IdentCache; config: ConfigRef) = ## Open a parser, using the given arguments to set up its internal state. ## - initToken(p.tok) + reset(p.tok) openLexer(p.lex, fileIdx, inputStream, cache, config) when defined(nimpretty): openEmitter(p.em, cache, config, fileIdx) getTok(p) # read the first token p.firstTok = true - p.emptyNode = newNode(nkEmpty) + when not defined(nimCustomAst): + p.emptyNode = newNode(nkEmpty) -proc openParser*(p: var TParser, filename: AbsoluteFile, inputStream: PLLStream, +proc openParser*(p: var Parser, filename: AbsoluteFile, inputStream: PLLStream, cache: IdentCache; config: ConfigRef) = openParser(p, fileInfoIdx(config, filename), inputStream, cache, config) -proc closeParser(p: var TParser) = +proc closeParser*(p: var Parser) = ## Close a parser, freeing up its resources. closeLexer(p.lex) - when defined(nimpretty): - closeEmitter(p.em) -proc parMessage(p: TParser, msg: TMsgKind, arg = "") = +proc parMessage(p: Parser, msg: TMsgKind, arg = "") = ## Produce and emit the parser message `arg` to output. lexMessageTok(p.lex, msg, p.tok, arg) -proc parMessage(p: TParser, msg: string, tok: TToken) = +proc parMessage(p: Parser, msg: string, tok: Token) = ## Produce and emit a parser message to output about the token `tok` parMessage(p, errGenerated, msg % prettyTok(tok)) -proc parMessage(p: TParser, arg: string) = +proc parMessage(p: Parser, arg: string) = ## Produce and emit the parser message `arg` to output. lexMessageTok(p.lex, errGenerated, p.tok, arg) @@ -150,7 +185,7 @@ template withInd(p, body: untyped) = body p.currInd = oldInd -template newlineWasSplitting(p: var TParser) = +template newlineWasSplitting(p: var Parser) = when defined(nimpretty): layouter.newlineWasSplitting(p.em) @@ -158,134 +193,127 @@ template realInd(p): bool = p.tok.indent > p.currInd template sameInd(p): bool = p.tok.indent == p.currInd template sameOrNoInd(p): bool = p.tok.indent == p.currInd or p.tok.indent < 0 -proc validInd(p: var TParser): bool {.inline.} = +proc validInd(p: var Parser): bool {.inline.} = result = p.tok.indent < 0 or p.tok.indent > p.currInd -proc rawSkipComment(p: var TParser, node: PNode) = +proc rawSkipComment(p: var Parser, node: PNode) = if p.tok.tokType == tkComment: if node != nil: - when not defined(nimNoNilSeqs): - if node.comment == nil: node.comment = "" + var rhs = node.comment when defined(nimpretty): if p.tok.commentOffsetB > p.tok.commentOffsetA: - node.comment.add fileSection(p.lex.config, p.lex.fileIdx, p.tok.commentOffsetA, p.tok.commentOffsetB) + rhs.add fileSection(p.lex.config, p.lex.fileIdx, p.tok.commentOffsetA, p.tok.commentOffsetB) else: - node.comment.add p.tok.literal + rhs.add p.tok.literal else: - node.comment.add p.tok.literal + rhs.add p.tok.literal + node.comment = move rhs else: parMessage(p, errInternal, "skipComment") getTok(p) -proc skipComment(p: var TParser, node: PNode) = +proc skipComment(p: var Parser, node: PNode) = if p.tok.indent < 0: rawSkipComment(p, node) -proc flexComment(p: var TParser, node: PNode) = +proc flexComment(p: var Parser, node: PNode) = if p.tok.indent < 0 or realInd(p): rawSkipComment(p, node) const errInvalidIndentation = "invalid indentation" errIdentifierExpected = "identifier expected, but got '$1'" errExprExpected = "expression expected, but found '$1'" - errTokenExpected = "'$1' expected" -proc skipInd(p: var TParser) = +proc skipInd(p: var Parser) = if p.tok.indent >= 0: if not realInd(p): parMessage(p, errInvalidIndentation) -proc optPar(p: var TParser) = +proc optPar(p: var Parser) = if p.tok.indent >= 0: if p.tok.indent < p.currInd: parMessage(p, errInvalidIndentation) -proc optInd(p: var TParser, n: PNode) = +proc optInd(p: var Parser, n: PNode) = skipComment(p, n) skipInd(p) -proc getTokNoInd(p: var TParser) = +proc getTokNoInd(p: var Parser) = getTok(p) if p.tok.indent >= 0: parMessage(p, errInvalidIndentation) -proc expectIdentOrKeyw(p: TParser) = +proc expectIdentOrKeyw(p: Parser) = if p.tok.tokType != tkSymbol and not isKeyword(p.tok.tokType): lexMessage(p.lex, errGenerated, errIdentifierExpected % prettyTok(p.tok)) -proc expectIdent(p: TParser) = +proc expectIdent(p: Parser) = if p.tok.tokType != tkSymbol: lexMessage(p.lex, errGenerated, errIdentifierExpected % prettyTok(p.tok)) -proc eat(p: var TParser, tokType: TTokType) = +proc eat(p: var Parser, tokType: TokType) = ## Move the parser to the next token if the current token is of type ## `tokType`, otherwise error. if p.tok.tokType == tokType: getTok(p) else: lexMessage(p.lex, errGenerated, - "expected: '" & TokTypeToStr[tokType] & "', but got: '" & prettyTok(p.tok) & "'") + "expected: '" & $tokType & "', but got: '" & prettyTok(p.tok) & "'") -proc parLineInfo(p: TParser): TLineInfo = +proc parLineInfo(p: Parser): TLineInfo = ## Retrieve the line information associated with the parser's current state. result = getLineInfo(p.lex, p.tok) -proc indAndComment(p: var TParser, n: PNode) = +proc indAndComment(p: var Parser, n: PNode, maybeMissEquals = false) = if p.tok.indent > p.currInd: if p.tok.tokType == tkComment: rawSkipComment(p, n) + elif maybeMissEquals: + let col = p.bufposPrevious - p.lineStartPrevious + var info = newLineInfo(p.lex.fileIdx, p.lineNumberPrevious, col) + parMessage(p, "invalid indentation, maybe you forgot a '=' at $1 ?" % [p.lex.config$info]) else: parMessage(p, errInvalidIndentation) else: skipComment(p, n) -proc newNodeP(kind: TNodeKind, p: TParser): PNode = - result = newNodeI(kind, parLineInfo(p)) +proc newNodeP(kind: TNodeKind, p: Parser): PNode = + result = newNode(kind, parLineInfo(p)) -proc newIntNodeP(kind: TNodeKind, intVal: BiggestInt, p: TParser): PNode = - result = newNodeP(kind, p) - result.intVal = intVal +proc newIntNodeP(kind: TNodeKind, intVal: BiggestInt, p: Parser): PNode = + result = newAtom(kind, intVal, parLineInfo(p)) proc newFloatNodeP(kind: TNodeKind, floatVal: BiggestFloat, - p: TParser): PNode = - result = newNodeP(kind, p) - result.floatVal = floatVal + p: Parser): PNode = + result = newAtom(kind, floatVal, parLineInfo(p)) -proc newStrNodeP(kind: TNodeKind, strVal: string, p: TParser): PNode = - result = newNodeP(kind, p) - result.strVal = strVal +proc newStrNodeP(kind: TNodeKind, strVal: sink string, p: Parser): PNode = + result = newAtom(kind, strVal, parLineInfo(p)) -proc newIdentNodeP(ident: PIdent, p: TParser): PNode = - result = newNodeP(nkIdent, p) - result.ident = ident +proc newIdentNodeP(ident: PIdent, p: Parser): PNode = + result = newAtom(ident, parLineInfo(p)) -proc parseExpr(p: var TParser): PNode -proc parseStmt(p: var TParser): PNode -proc parseTypeDesc(p: var TParser): PNode -proc parseParamList(p: var TParser, retColon = true): PNode +proc parseExpr(p: var Parser): PNode +proc parseStmt(p: var Parser): PNode +proc parseTypeDesc(p: var Parser, fullExpr = false): PNode +proc parseTypeDefValue(p: var Parser): PNode +proc parseParamList(p: var Parser, retColon = true): PNode -proc isSigilLike(tok: TToken): bool {.inline.} = +proc isSigilLike(tok: Token): bool {.inline.} = result = tok.tokType == tkOpr and tok.ident.s[0] == '@' -proc isRightAssociative(tok: TToken): bool {.inline.} = +proc isRightAssociative(tok: Token): bool {.inline.} = ## Determines whether the token is right assocative. result = tok.tokType == tkOpr and tok.ident.s[0] == '^' # or (tok.ident.s.len > 1 and tok.ident.s[^1] == '>') -proc isOperator(tok: TToken): bool = - ## Determines if the given token is an operator type token. - tok.tokType in {tkOpr, tkDiv, tkMod, tkShl, tkShr, tkIn, tkNotin, tkIs, - tkIsnot, tkNot, tkOf, tkAs, tkDotDot, tkAnd, tkOr, tkXor} - -proc isUnary(p: TParser): bool = - ## Check if the current parser token is a unary operator - if p.tok.tokType in {tkOpr, tkDotDot} and - p.tok.strongSpaceB == 0 and - p.tok.strongSpaceA > 0: - result = true +proc isUnary(tok: Token): bool = + ## Check if the given token is a unary operator + tok.tokType in {tkOpr, tkDotDot} and + tok.spacing == {tsLeading} -proc checkBinary(p: TParser) {.inline.} = +proc checkBinary(p: Parser) {.inline.} = ## Check if the current parser token is a binary operator. # we don't check '..' here as that's too annoying if p.tok.tokType == tkOpr: - if p.tok.strongSpaceB > 0 and p.tok.strongSpaceA == 0: + if p.tok.spacing == {tsTrailing}: parMessage(p, warnInconsistentSpacing, prettyTok(p.tok)) -#| module = stmt ^* (';' / IND{=}) +#| module = complexOrSimpleStmt ^* (';' / IND{=}) #| #| comma = ',' COMMENT? #| semicolon = ';' COMMENT? @@ -294,8 +322,8 @@ proc checkBinary(p: TParser) {.inline.} = #| #| operator = OP0 | OP1 | OP2 | OP3 | OP4 | OP5 | OP6 | OP7 | OP8 | OP9 #| | 'or' | 'xor' | 'and' -#| | 'is' | 'isnot' | 'in' | 'notin' | 'of' -#| | 'div' | 'mod' | 'shl' | 'shr' | 'not' | 'static' | '..' +#| | 'is' | 'isnot' | 'in' | 'notin' | 'of' | 'as' | 'from' +#| | 'div' | 'mod' | 'shl' | 'shr' | 'not' | '..' #| #| prefixOperator = operator #| @@ -314,15 +342,30 @@ proc checkBinary(p: TParser) {.inline.} = #| mulExpr = dollarExpr (OP9 optInd dollarExpr)* #| dollarExpr = primary (OP10 optInd primary)* -proc colcom(p: var TParser, n: PNode) = +proc isOperator(tok: Token): bool = + #| operatorB = OP0 | OP1 | OP2 | OP3 | OP4 | OP5 | OP6 | OP7 | OP8 | OP9 | + #| 'div' | 'mod' | 'shl' | 'shr' | 'in' | 'notin' | + #| 'is' | 'isnot' | 'not' | 'of' | 'as' | 'from' | '..' | 'and' | 'or' | 'xor' + tok.tokType in {tkOpr, tkDiv, tkMod, tkShl, tkShr, tkIn, tkNotin, tkIs, + tkIsnot, tkNot, tkOf, tkAs, tkFrom, tkDotDot, tkAnd, + tkOr, tkXor} + +proc colcom(p: var Parser, n: PNode) = eat(p, tkColon) skipComment(p, n) const tkBuiltInMagics = {tkType, tkStatic, tkAddr} -proc parseSymbol(p: var TParser, mode = smNormal): PNode = +template setEndInfo() = + when defined(nimsuggest): + result.endInfo = TLineInfo(fileIndex: p.lex.fileIdx, + line: p.lex.previousTokenEnd.line, + col: p.lex.previousTokenEnd.col) + +proc parseSymbol(p: var Parser, mode = smNormal): PNode = #| symbol = '`' (KEYW|IDENT|literal|(operator|'('|')'|'['|']'|'{'|'}'|'=')+)+ '`' - #| | IDENT | KEYW + #| | IDENT | 'addr' | 'type' | 'static' + #| symbolOrKeyword = symbol | KEYW case p.tok.tokType of tkSymbol: result = newIdentNodeP(p.tok.ident, p) @@ -345,7 +388,7 @@ proc parseSymbol(p: var TParser, mode = smNormal): PNode = while true: case p.tok.tokType of tkAccent: - if result.len == 0: + if not result.hasSon: parMessage(p, errIdentifierExpected, p.tok) break of tkOpr, tkDot, tkDotDot, tkEquals, tkParLe..tkParDotRi: @@ -355,10 +398,9 @@ proc parseSymbol(p: var TParser, mode = smNormal): PNode = tkParLe..tkParDotRi}: accm.add($p.tok) getTok(p) - let node = newNodeI(nkIdent, lineinfo) - node.ident = p.lex.cache.getIdent(accm) + let node = newAtom(p.lex.cache.getIdent(accm), lineinfo) result.add(node) - of tokKeywordLow..tokKeywordHigh, tkSymbol, tkIntLit..tkCharLit: + of tokKeywordLow..tokKeywordHigh, tkSymbol, tkIntLit..tkCustomLit: result.add(newIdentNodeP(p.lex.cache.getIdent($p.tok), p)) getTok(p) else: @@ -372,39 +414,72 @@ proc parseSymbol(p: var TParser, mode = smNormal): PNode = # if it is a keyword: #if not isKeyword(p.tok.tokType): getTok(p) result = p.emptyNode + setEndInfo() -proc colonOrEquals(p: var TParser, a: PNode): PNode = - if p.tok.tokType == tkColon: - result = newNodeP(nkExprColonExpr, p) +proc equals(p: var Parser, a: PNode): PNode = + if p.tok.tokType == tkEquals: + result = newNodeP(nkExprEqExpr, p) getTok(p) - newlineWasSplitting(p) #optInd(p, result) result.add(a) result.add(parseExpr(p)) - elif p.tok.tokType == tkEquals: - result = newNodeP(nkExprEqExpr, p) + else: + result = a + +proc colonOrEquals(p: var Parser, a: PNode): PNode = + if p.tok.tokType == tkColon: + result = newNodeP(nkExprColonExpr, p) getTok(p) + newlineWasSplitting(p) #optInd(p, result) result.add(a) result.add(parseExpr(p)) else: - result = a + result = equals(p, a) -proc exprColonEqExpr(p: var TParser): PNode = - #| exprColonEqExpr = expr (':'|'=' expr)? +proc exprColonEqExpr(p: var Parser): PNode = + #| exprColonEqExpr = expr ((':'|'=') expr + #| / doBlock extraPostExprBlock*)? var a = parseExpr(p) if p.tok.tokType == tkDo: result = postExprBlocks(p, a) else: result = colonOrEquals(p, a) -proc exprList(p: var TParser, endTok: TTokType, result: PNode) = +proc exprEqExpr(p: var Parser): PNode = + #| exprEqExpr = expr ('=' expr + #| / doBlock extraPostExprBlock*)? + var a = parseExpr(p) + if p.tok.tokType == tkDo: + result = postExprBlocks(p, a) + else: + result = equals(p, a) + +proc exprList(p: var Parser, endTok: TokType, result: PNode) = #| exprList = expr ^+ comma when defined(nimpretty): inc p.em.doIndentMore getTok(p) optInd(p, result) # progress guaranteed + var a = parseExpr(p) + result.add(a) + while (p.tok.tokType != endTok) and (p.tok.tokType != tkEof): + if p.tok.tokType != tkComma: break + getTok(p) + optInd(p, a) + var a = parseExpr(p) + result.add(a) + when defined(nimpretty): + dec p.em.doIndentMore + +proc optionalExprList(p: var Parser, endTok: TokType, result: PNode) = + #| optionalExprList = expr ^* comma + when defined(nimpretty): + inc p.em.doIndentMore + getTok(p) + optInd(p, result) + # progress guaranteed while (p.tok.tokType != endTok) and (p.tok.tokType != tkEof): var a = parseExpr(p) result.add(a) @@ -414,7 +489,7 @@ proc exprList(p: var TParser, endTok: TTokType, result: PNode) = when defined(nimpretty): dec p.em.doIndentMore -proc exprColonEqExprListAux(p: var TParser, endTok: TTokType, result: PNode) = +proc exprColonEqExprListAux(p: var Parser, endTok: TokType, result: PNode) = assert(endTok in {tkCurlyRi, tkCurlyDotRi, tkBracketRi, tkParRi}) getTok(p) flexComment(p, result) @@ -424,48 +499,55 @@ proc exprColonEqExprListAux(p: var TParser, endTok: TTokType, result: PNode) = var a = exprColonEqExpr(p) result.add(a) if p.tok.tokType != tkComma: break - getTok(p) - # (1,) produces a tuple expression - if endTok == tkParRi and p.tok.tokType == tkParRi and result.kind == nkPar: + elif result.kind == nkPar: result.transitionSonsKind(nkTupleConstr) + getTok(p) skipComment(p, a) optPar(p) eat(p, endTok) -proc exprColonEqExprList(p: var TParser, kind: TNodeKind, - endTok: TTokType): PNode = +proc exprColonEqExprList(p: var Parser, kind: TNodeKind, + endTok: TokType): PNode = #| exprColonEqExprList = exprColonEqExpr (comma exprColonEqExpr)* (comma)? result = newNodeP(kind, p) exprColonEqExprListAux(p, endTok, result) -proc dotExpr(p: var TParser, a: PNode): PNode = - #| dotExpr = expr '.' optInd (symbol | '[:' exprList ']') - #| explicitGenericInstantiation = '[:' exprList ']' ( '(' exprColonEqExpr ')' )? +proc dotExpr(p: var Parser, a: PNode): PNode = var info = p.parLineInfo getTok(p) - result = newNodeI(nkDotExpr, info) + result = newNode(nkDotExpr, info) optInd(p, result) result.add(a) result.add(parseSymbol(p, smAfterDot)) - if p.tok.tokType == tkBracketLeColon and p.tok.strongSpaceA <= 0: - var x = newNodeI(nkBracketExpr, p.parLineInfo) + if p.tok.tokType == tkBracketLeColon and tsLeading notin p.tok.spacing: + var x = newNode(nkBracketExpr, p.parLineInfo) # rewrite 'x.y[:z]()' to 'y[z](x)' - x.add result[1] + x.add result.secondSon exprList(p, tkBracketRi, x) eat(p, tkBracketRi) - var y = newNodeI(nkCall, p.parLineInfo) + var y = newNode(nkCall, p.parLineInfo) y.add x - y.add result[0] - if p.tok.tokType == tkParLe and p.tok.strongSpaceA <= 0: + y.add result.firstSon + if p.tok.tokType == tkParLe and tsLeading notin p.tok.spacing: exprColonEqExprListAux(p, tkParRi, y) result = y -proc qualifiedIdent(p: var TParser): PNode = - #| qualifiedIdent = symbol ('.' optInd symbol)? +proc dotLikeExpr(p: var Parser, a: PNode): PNode = + var info = p.parLineInfo + result = newNode(nkInfix, info) + optInd(p, result) + var opNode = newIdentNodeP(p.tok.ident, p) + getTok(p) + result.add(opNode) + result.add(a) + result.add(parseSymbol(p, smAfterDot)) + +proc qualifiedIdent(p: var Parser): PNode = + #| qualifiedIdent = symbol ('.' optInd symbolOrKeyword)? result = parseSymbol(p) if p.tok.tokType == tkDot: result = dotExpr(p, result) -proc setOrTableConstr(p: var TParser): PNode = +proc setOrTableConstr(p: var Parser): PNode = #| setOrTableConstr = '{' ((exprColonEqExpr comma)* | ':' ) '}' result = newNodeP(nkCurly, p) getTok(p) # skip '{' @@ -485,29 +567,43 @@ proc setOrTableConstr(p: var TParser): PNode = optPar(p) eat(p, tkCurlyRi) # skip '}' -proc parseCast(p: var TParser): PNode = - #| castExpr = 'cast' '[' optInd typeDesc optPar ']' '(' optInd expr optPar ')' +proc parseCast(p: var Parser): PNode = + #| castExpr = 'cast' ('[' optInd typeDesc optPar ']' '(' optInd expr optPar ')') / + # ('(' optInd exprColonEqExpr optPar ')') result = newNodeP(nkCast, p) getTok(p) - eat(p, tkBracketLe) - optInd(p, result) - result.add(parseTypeDesc(p)) - optPar(p) - eat(p, tkBracketRi) - eat(p, tkParLe) - optInd(p, result) - result.add(parseExpr(p)) + if p.tok.tokType == tkBracketLe: + getTok(p) + optInd(p, result) + result.add(parseTypeDesc(p)) + optPar(p) + eat(p, tkBracketRi) + eat(p, tkParLe) + optInd(p, result) + result.add(parseExpr(p)) + else: + result.add p.emptyNode + eat(p, tkParLe) + optInd(p, result) + result.add(exprColonEqExpr(p)) optPar(p) eat(p, tkParRi) + setEndInfo() + +template setNodeFlag(n: PNode; f: untyped) = + when defined(nimCustomAst): + discard + else: + incl n.flags, f -proc setBaseFlags(n: PNode, base: TNumericalBase) = +proc setBaseFlags(n: PNode, base: NumericalBase) = case base of base10: discard - of base2: incl(n.flags, nfBase2) - of base8: incl(n.flags, nfBase8) - of base16: incl(n.flags, nfBase16) + of base2: setNodeFlag(n, nfBase2) + of base8: setNodeFlag(n, nfBase8) + of base16: setNodeFlag(n, nfBase16) -proc parseGStrLit(p: var TParser, a: PNode): PNode = +proc parseGStrLit(p: var Parser, a: PNode): PNode = case p.tok.tokType of tkGStrLit: result = newNodeP(nkCallStrLit, p) @@ -521,30 +617,48 @@ proc parseGStrLit(p: var TParser, a: PNode): PNode = getTok(p) else: result = a + setEndInfo() -proc complexOrSimpleStmt(p: var TParser): PNode -proc simpleExpr(p: var TParser, mode = pmNormal): PNode +proc complexOrSimpleStmt(p: var Parser): PNode +proc simpleExpr(p: var Parser, mode = pmNormal): PNode +proc parseIfOrWhenExpr(p: var Parser, kind: TNodeKind): PNode -proc semiStmtList(p: var TParser, result: PNode) = +proc semiStmtList(p: var Parser, result: PNode) = inc p.inSemiStmtList - result.add(complexOrSimpleStmt(p)) - # progress guaranteed - while p.tok.tokType == tkSemiColon: - getTok(p) - optInd(p, result) - result.add(complexOrSimpleStmt(p)) + withInd(p): + # Be lenient with the first stmt/expr + let a = case p.tok.tokType + of tkIf: parseIfOrWhenExpr(p, nkIfStmt) + of tkWhen: parseIfOrWhenExpr(p, nkWhenStmt) + else: complexOrSimpleStmt(p) + result.add a + + while p.tok.tokType != tkEof: + if p.tok.tokType == tkSemiColon: + getTok(p) + if p.tok.tokType == tkParRi: + break + elif not (sameInd(p) or realInd(p)): + parMessage(p, errInvalidIndentation) + let a = complexOrSimpleStmt(p) + if a.kind == nkEmpty: + parMessage(p, errExprExpected, p.tok) + getTok(p) + else: + result.add a dec p.inSemiStmtList result.transitionSonsKind(nkStmtListExpr) -proc parsePar(p: var TParser): PNode = +proc parsePar(p: var Parser): PNode = #| parKeyw = 'discard' | 'include' | 'if' | 'while' | 'case' | 'try' #| | 'finally' | 'except' | 'for' | 'block' | 'const' | 'let' #| | 'when' | 'var' | 'mixin' #| par = '(' optInd - #| ( &parKeyw complexOrSimpleStmt ^+ ';' - #| | ';' complexOrSimpleStmt ^+ ';' + #| ( &parKeyw (ifExpr / complexOrSimpleStmt) ^+ ';' + #| | ';' (ifExpr / complexOrSimpleStmt) ^+ ';' #| | pragmaStmt - #| | simpleExpr ( ('=' expr (';' complexOrSimpleStmt ^+ ';' )? ) + #| | simpleExpr ( (doBlock extraPostExprBlock*) + #| | ('=' expr (';' (ifExpr / complexOrSimpleStmt) ^+ ';' )? ) #| | (':' expr (',' exprColonEqExpr ^+ ',' )? ) ) ) #| optPar ')' # @@ -555,7 +669,7 @@ proc parsePar(p: var TParser): PNode = optInd(p, result) flexComment(p, result) if p.tok.tokType in {tkDiscard, tkInclude, tkIf, tkWhile, tkCase, - tkTry, tkDefer, tkFinally, tkExcept, tkFor, tkBlock, + tkTry, tkDefer, tkFinally, tkExcept, tkBlock, tkConst, tkLet, tkWhen, tkVar, tkFor, tkMixin}: # XXX 'bind' used to be an expression, so we exclude it here; @@ -568,7 +682,10 @@ proc parsePar(p: var TParser): PNode = semiStmtList(p, result) elif p.tok.tokType == tkCurlyDotLe: result.add(parseStmtPragma(p)) - elif p.tok.tokType != tkParRi: + elif p.tok.tokType == tkParRi: + # Empty tuple '()' + result.transitionSonsKind(nkTupleConstr) + else: var a = simpleExpr(p) if p.tok.tokType == tkDo: result = postExprBlocks(p, a) @@ -589,13 +706,14 @@ proc parsePar(p: var TParser): PNode = semiStmtList(p, result) else: a = colonOrEquals(p, a) + if a.kind == nkExprColonExpr: + result.transitionSonsKind(nkTupleConstr) result.add(a) if p.tok.tokType == tkComma: getTok(p) skipComment(p, a) # (1,) produces a tuple expression: - if p.tok.tokType == tkParRi: - result.transitionSonsKind(nkTupleConstr) + result.transitionSonsKind(nkTupleConstr) # progress guaranteed while p.tok.tokType != tkParRi and p.tok.tokType != tkEof: var a = exprColonEqExpr(p) @@ -605,17 +723,18 @@ proc parsePar(p: var TParser): PNode = skipComment(p, a) optPar(p) eat(p, tkParRi) + setEndInfo() -proc identOrLiteral(p: var TParser, mode: TPrimaryMode): PNode = +proc identOrLiteral(p: var Parser, mode: PrimaryMode): 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 + #| | CHAR_LIT | CUSTOM_NUMERIC_LIT #| | NIL #| generalizedLit = GENERALIZED_STR_LIT | GENERALIZED_TRIPLESTR_LIT #| identOrLiteral = generalizedLit | symbol | literal - #| | par | arrayConstr | setOrTableConstr + #| | par | arrayConstr | setOrTableConstr | tupleConstr #| | castExpr #| tupleConstr = '(' optInd (exprColonEqExpr comma?)* optPar ')' #| arrayConstr = '[' optInd (exprColonEqExpr comma?)* optPar ']' @@ -694,6 +813,14 @@ proc identOrLiteral(p: var TParser, mode: TPrimaryMode): PNode = of tkCharLit: result = newIntNodeP(nkCharLit, ord(p.tok.literal[0]), p) getTok(p) + of tkCustomLit: + let splitPos = p.tok.iNumber.int + let str = newStrNodeP(nkRStrLit, p.tok.literal.substr(0, splitPos-1), p) + let callee = newIdentNodeP(getIdent(p.lex.cache, p.tok.literal.substr(splitPos)), p) + result = newNodeP(nkDotExpr, p) + result.add str + result.add callee + getTok(p) of tkNil: result = newNodeP(nkNilLit, p) getTok(p) @@ -716,48 +843,48 @@ proc identOrLiteral(p: var TParser, mode: TPrimaryMode): PNode = getTok(p) # we must consume a token here to prevent endless loops! result = p.emptyNode -proc namedParams(p: var TParser, callee: PNode, - kind: TNodeKind, endTok: TTokType): PNode = +proc namedParams(p: var Parser, callee: PNode, + kind: TNodeKind, endTok: TokType): PNode = let a = callee result = newNodeP(kind, p) result.add(a) # progress guaranteed exprColonEqExprListAux(p, endTok, result) -proc commandParam(p: var TParser, isFirstParam: var bool; mode: TPrimaryMode): PNode = +proc commandParam(p: var Parser, isFirstParam: var bool; mode: PrimaryMode): PNode = if mode == pmTypeDesc: result = simpleExpr(p, mode) + elif not isFirstParam: + result = exprEqExpr(p) else: result = parseExpr(p) - if p.tok.tokType == tkDo: - result = postExprBlocks(p, result) - elif p.tok.tokType == tkEquals and not isFirstParam: - let lhs = result - result = newNodeP(nkExprEqExpr, p) - getTok(p) - result.add(lhs) - result.add(parseExpr(p)) + if p.tok.tokType == tkDo: + result = postExprBlocks(p, result) isFirstParam = false -const - tkTypeClasses = {tkRef, tkPtr, tkVar, tkStatic, tkType, - tkEnum, tkTuple, tkObject, tkProc} - -proc commandExpr(p: var TParser; r: PNode; mode: TPrimaryMode): PNode = - result = newNodeP(nkCommand, p) - result.add(r) - var isFirstParam = true - # progress NOT guaranteed - p.hasProgress = false - result.add commandParam(p, isFirstParam, mode) - -proc primarySuffix(p: var TParser, r: PNode, - baseIndent: int, mode: TPrimaryMode): PNode = +proc commandExpr(p: var Parser; r: PNode; mode: PrimaryMode): PNode = + if mode == pmTrySimple: + result = r + else: + result = newNodeP(nkCommand, p) + result.add(r) + var isFirstParam = true + # progress NOT guaranteed + p.hasProgress = false + result.add commandParam(p, isFirstParam, mode) + +proc isDotLike(tok: Token): bool = + result = tok.tokType == tkOpr and tok.ident.s.len > 1 and + tok.ident.s[0] == '.' and tok.ident.s[1] != '.' + +proc primarySuffix(p: var Parser, r: PNode, + baseIndent: int, mode: PrimaryMode): PNode = #| primarySuffix = '(' (exprColonEqExpr comma?)* ')' - #| | '.' optInd symbol generalizedLit? + #| | '.' optInd symbolOrKeyword ('[:' exprList ']' ( '(' exprColonEqExpr ')' )?)? generalizedLit? + #| | DOTLIKEOP optInd symbolOrKeyword generalizedLit? #| | '[' optInd exprColonEqExprList optPar ']' #| | '{' optInd exprColonEqExprList optPar '}' - #| | &( '`'|IDENT|literal|'cast'|'addr'|'type') expr # command syntax + # XXX strong spaces need to be reflected above result = r # progress guaranteed @@ -766,18 +893,11 @@ proc primarySuffix(p: var TParser, r: PNode, case p.tok.tokType of tkParLe: # progress guaranteed - if p.tok.strongSpaceA > 0: - # inside type sections, expressions such as `ref (int, bar)` - # are parsed as a nkCommand with a single tuple argument (nkPar) - if mode == pmTypeDef: - result = newNodeP(nkCommand, p) - result.add r - result.add primary(p, pmNormal) - else: - result = commandExpr(p, result, mode) + if tsLeading in p.tok.spacing: + result = commandExpr(p, result, mode) break result = namedParams(p, result, nkCall, tkParRi) - if result.len > 1 and result[1].kind == nkExprColonExpr: + if result.has2Sons and result.secondSon.kind == nkExprColonExpr: result.transitionSonsKind(nkObjConstr) of tkDot: # progress guaranteed @@ -785,42 +905,51 @@ proc primarySuffix(p: var TParser, r: PNode, result = parseGStrLit(p, result) of tkBracketLe: # progress guaranteed - if p.tok.strongSpaceA > 0: + if tsLeading in p.tok.spacing: result = commandExpr(p, result, mode) break result = namedParams(p, result, nkBracketExpr, tkBracketRi) of tkCurlyLe: # progress guaranteed - if p.tok.strongSpaceA > 0: + if tsLeading in p.tok.spacing: result = commandExpr(p, result, mode) break result = namedParams(p, result, nkCurlyExpr, tkCurlyRi) - of tkSymbol, tkAccent, tkIntLit..tkCharLit, tkNil, tkCast, - tkOpr, tkDotDot, tkTypeClasses - {tkRef, tkPtr}: + of tkSymbol, tkAccent, tkIntLit..tkCustomLit, tkNil, tkCast, + tkOpr, tkDotDot, tkVar, tkOut, tkStatic, tkType, tkEnum, tkTuple, + tkObject, tkProc: # XXX: In type sections we allow the free application of the # command syntax, with the exception of expressions such as # `foo ref` or `foo ptr`. Unfortunately, these two are also # used as infix operators for the memory regions feature and # the current parsing rules don't play well here. - if p.inPragma == 0 and (isUnary(p) or p.tok.tokType notin {tkOpr, tkDotDot}): - # actually parsing {.push hints:off.} as {.push(hints:off).} is a sweet - # solution, but pragmas.nim can't handle that - result = commandExpr(p, result, mode) - break + let isDotLike2 = p.tok.isDotLike + if isDotLike2 and p.lex.config.isDefined("nimPreviewDotLikeOps"): + # synchronize with `tkDot` branch + result = dotLikeExpr(p, result) + result = parseGStrLit(p, result) + else: + if isDotLike2: + parMessage(p, warnDotLikeOps, "dot-like operators will be parsed differently with `-d:nimPreviewDotLikeOps`") + if p.inPragma == 0 and (isUnary(p.tok) or p.tok.tokType notin {tkOpr, tkDotDot}): + # actually parsing {.push hints:off.} as {.push(hints:off).} is a sweet + # solution, but pragmas.nim can't handle that + result = commandExpr(p, result, mode) + break else: break -proc parseOperators(p: var TParser, headNode: PNode, - limit: int, mode: TPrimaryMode): PNode = +proc parseOperators(p: var Parser, headNode: PNode, + limit: int, mode: PrimaryMode): PNode = result = headNode # expand while operators have priorities higher than 'limit' - var opPrec = getPrecedence(p.tok, false) + var opPrec = getPrecedence(p.tok) let modeB = if mode == pmTypeDef: pmTypeDesc else: mode # the operator itself must not start on a new line: # progress guaranteed - while opPrec >= limit and p.tok.indent < 0 and not isUnary(p): + while opPrec >= limit and p.tok.indent < 0 and not isUnary(p.tok): checkBinary(p) - var leftAssoc = 1-ord(isRightAssociative(p.tok)) + let leftAssoc = ord(not isRightAssociative(p.tok)) var a = newNodeP(nkInfix, p) var opNode = newIdentNodeP(p.tok.ident, p) # skip operator: getTok(p) @@ -832,10 +961,14 @@ proc parseOperators(p: var TParser, headNode: PNode, a.add(result) a.add(b) result = a - opPrec = getPrecedence(p.tok, false) + opPrec = getPrecedence(p.tok) + setEndInfo() -proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode = +proc simpleExprAux(p: var Parser, limit: int, mode: PrimaryMode): PNode = + var mode = mode result = primary(p, mode) + if mode == pmTrySimple: + mode = pmNormal if p.tok.tokType == tkCurlyDotLe and (p.tok.indent < 0 or realInd(p)) and mode == pmNormal: var pragmaExp = newNodeP(nkPragmaExpr, p) @@ -844,72 +977,14 @@ proc simpleExprAux(p: var TParser, limit: int, mode: TPrimaryMode): PNode = result = pragmaExp result = parseOperators(p, result, limit, mode) -proc simpleExpr(p: var TParser, mode = pmNormal): PNode = +proc simpleExpr(p: var Parser, mode = pmNormal): PNode = when defined(nimpretty): inc p.em.doIndentMore result = simpleExprAux(p, -1, mode) when defined(nimpretty): dec p.em.doIndentMore -proc parseIfExpr(p: var TParser, kind: TNodeKind): PNode = - #| condExpr = expr colcom expr optInd - #| ('elif' expr colcom expr optInd)* - #| 'else' colcom expr - #| ifExpr = 'if' condExpr - #| whenExpr = 'when' condExpr - when true: - result = newNodeP(kind, p) - while true: - getTok(p) # skip `if`, `when`, `elif` - var branch = newNodeP(nkElifExpr, p) - optInd(p, branch) - branch.add(parseExpr(p)) - colcom(p, branch) - branch.add(parseStmt(p)) - skipComment(p, branch) - result.add(branch) - if p.tok.tokType != tkElif: break # or not sameOrNoInd(p): break - if p.tok.tokType == tkElse: # and sameOrNoInd(p): - var branch = newNodeP(nkElseExpr, p) - eat(p, tkElse) - colcom(p, branch) - branch.add(parseStmt(p)) - result.add(branch) - else: - var - b: PNode - wasIndented = false - result = newNodeP(kind, p) - - getTok(p) - let branch = newNodeP(nkElifExpr, p) - branch.add(parseExpr(p)) - colcom(p, branch) - let oldInd = p.currInd - if realInd(p): - p.currInd = p.tok.indent - wasIndented = true - branch.add(parseExpr(p)) - result.add branch - while sameInd(p) or not wasIndented: - case p.tok.tokType - of tkElif: - b = newNodeP(nkElifExpr, p) - getTok(p) - optInd(p, b) - b.add(parseExpr(p)) - of tkElse: - b = newNodeP(nkElseExpr, p) - getTok(p) - else: break - colcom(p, b) - b.add(parseStmt(p)) - result.add(b) - if b.kind == nkElseExpr: break - if wasIndented: - p.currInd = oldInd - -proc parsePragma(p: var TParser): PNode = +proc parsePragma(p: var Parser): PNode = #| pragma = '{.' optInd (exprColonEqExpr comma?)* optPar ('.}' | '}') result = newNodeP(nkPragma, p) inc p.inPragma @@ -937,10 +1012,11 @@ proc parsePragma(p: var TParser): PNode = when defined(nimpretty): dec p.em.doIndentMore dec p.em.keepIndents + setEndInfo() -proc identVis(p: var TParser; allowDot=false): PNode = +proc identVis(p: var Parser; allowDot=false): PNode = #| identVis = symbol OPR? # postfix position - #| identVisDot = symbol '.' optInd symbol OPR? + #| identVisDot = symbol '.' optInd symbolOrKeyword OPR? var a = parseSymbol(p) if p.tok.tokType == tkOpr: when defined(nimpretty): @@ -954,7 +1030,7 @@ proc identVis(p: var TParser; allowDot=false): PNode = else: result = a -proc identWithPragma(p: var TParser; allowDot=false): PNode = +proc identWithPragma(p: var Parser; allowDot=false): PNode = #| identWithPragma = identVis pragma? #| identWithPragmaDot = identVisDot pragma? var a = identVis(p, allowDot) @@ -966,17 +1042,17 @@ proc identWithPragma(p: var TParser; allowDot=false): PNode = result = a type - TDeclaredIdentFlag = enum + DeclaredIdentFlag = enum withPragma, # identifier may have pragma withBothOptional # both ':' and '=' parts are optional withDot # allow 'var ident.ident = value' - TDeclaredIdentFlags = set[TDeclaredIdentFlag] + DeclaredIdentFlags = set[DeclaredIdentFlag] -proc parseIdentColonEquals(p: var TParser, flags: TDeclaredIdentFlags): PNode = +proc parseIdentColonEquals(p: var Parser, flags: DeclaredIdentFlags): PNode = #| declColonEquals = identWithPragma (comma identWithPragma)* comma? - #| (':' optInd typeDesc)? ('=' optInd expr)? + #| (':' optInd typeDescExpr)? ('=' optInd expr)? #| identColonEquals = IDENT (comma IDENT)* comma? - #| (':' optInd typeDesc)? ('=' optInd expr)?) + #| (':' optInd typeDescExpr)? ('=' optInd expr)?) var a: PNode result = newNodeP(nkIdentDefs, p) # progress guaranteed @@ -994,7 +1070,7 @@ proc parseIdentColonEquals(p: var TParser, flags: TDeclaredIdentFlags): PNode = if p.tok.tokType == tkColon: getTok(p) optInd(p, result) - result.add(parseTypeDesc(p)) + result.add(parseTypeDesc(p, fullExpr = true)) else: result.add(newNodeP(nkEmpty, p)) if p.tok.tokType != tkEquals and withBothOptional notin flags: @@ -1005,13 +1081,13 @@ proc parseIdentColonEquals(p: var TParser, flags: TDeclaredIdentFlags): PNode = result.add(parseExpr(p)) else: result.add(newNodeP(nkEmpty, p)) + setEndInfo() -proc parseTuple(p: var TParser, indentAllowed = false): PNode = - #| inlTupleDecl = 'tuple' - #| '[' optInd (identColonEquals (comma/semicolon)?)* optPar ']' - #| extTupleDecl = 'tuple' - #| COMMENT? (IND{>} identColonEquals (IND{=} identColonEquals)*)? - #| tupleClass = 'tuple' +proc parseTuple(p: var Parser, indentAllowed = false): PNode = + #| tupleTypeBracket = '[' optInd (identColonEquals (comma/semicolon)?)* optPar ']' + #| tupleType = 'tuple' tupleTypeBracket + #| tupleDecl = 'tuple' (tupleTypeBracket / + #| COMMENT? (IND{>} identColonEquals (IND{=} identColonEquals)*)?) result = newNodeP(nkTupleTy, p) getTok(p) if p.tok.tokType == tkBracketLe: @@ -1050,8 +1126,9 @@ proc parseTuple(p: var TParser, indentAllowed = false): PNode = parMessage(p, errGenerated, "the syntax for tuple types is 'tuple[...]', not 'tuple(...)'") else: result = newNodeP(nkTupleClassTy, p) + setEndInfo() -proc parseParamList(p: var TParser, retColon = true): PNode = +proc parseParamList(p: var Parser, retColon = true): PNode = #| paramList = '(' declColonEquals ^* (comma/semicolon) ')' #| paramListArrow = paramList? ('->' optInd typeDesc)? #| paramListColon = paramList? (':' optInd typeDesc)? @@ -1091,63 +1168,74 @@ proc parseParamList(p: var TParser, retColon = true): PNode = if hasRet and p.tok.indent < 0: getTok(p) optInd(p, result) - result[0] = parseTypeDesc(p) + result.replaceFirstSon parseTypeDesc(p) elif not retColon and not hasParLe: # Mark as "not there" in order to mark for deprecation in the semantic pass: result = p.emptyNode when defined(nimpretty): dec p.em.doIndentMore dec p.em.keepIndents + setEndInfo() -proc optPragmas(p: var TParser): PNode = +proc optPragmas(p: var Parser): PNode = if p.tok.tokType == tkCurlyDotLe and (p.tok.indent < 0 or realInd(p)): result = parsePragma(p) else: result = p.emptyNode -proc parseDoBlock(p: var TParser; info: TLineInfo): PNode = +proc parseDoBlock(p: var Parser; info: TLineInfo): PNode = #| doBlock = 'do' paramListArrow pragma? colcom stmt - let params = parseParamList(p, retColon=false) + result = nil + var params = parseParamList(p, retColon=false) let pragmas = optPragmas(p) colcom(p, result) result = parseStmt(p) - if params.kind != nkEmpty: + if params.kind != nkEmpty or pragmas.kind != nkEmpty: + if params.kind == nkEmpty: + params = newNodeP(nkFormalParams, p) + params.add(p.emptyNode) # return type result = newProcNode(nkDo, info, body = result, params = params, name = p.emptyNode, pattern = p.emptyNode, genericParams = p.emptyNode, pragmas = pragmas, exceptions = p.emptyNode) + setEndInfo() -proc parseProcExpr(p: var TParser; isExpr: bool; kind: TNodeKind): PNode = - #| procExpr = 'proc' paramListColon pragma? ('=' COMMENT? stmt)? +proc parseProcExpr(p: var Parser; isExpr: bool; kind: TNodeKind): PNode = + #| routineExpr = ('proc' | 'func' | 'iterator') paramListColon pragma? ('=' COMMENT? stmt)? + #| routineType = ('proc' | 'iterator') paramListColon pragma? # either a proc type or a anonymous proc let info = parLineInfo(p) - getTok(p) let hasSignature = p.tok.tokType in {tkParLe, tkColon} and p.tok.indent < 0 let params = parseParamList(p) let pragmas = optPragmas(p) if p.tok.tokType == tkEquals and isExpr: getTok(p) - skipComment(p, result) - result = newProcNode(kind, info, body = parseStmt(p), + result = newProcNode(kind, info, body = p.emptyNode, params = params, name = p.emptyNode, pattern = p.emptyNode, genericParams = p.emptyNode, pragmas = pragmas, exceptions = p.emptyNode) + skipComment(p, result) + result.replaceSon bodyPos, parseStmt(p) else: - result = newNodeI(nkProcTy, info) - if hasSignature: - result.add(params) + result = newNode(if kind == nkIteratorDef: nkIteratorTy else: nkProcTy, info) + if hasSignature or pragmas.kind != nkEmpty: + if hasSignature: + result.add(params) + else: # pragmas but no param list, implies typeclass with pragmas + result.add(p.emptyNode) if kind == nkFuncDef: parMessage(p, "func keyword is not allowed in type descriptions, use proc with {.noSideEffect.} pragma instead") result.add(pragmas) + setEndInfo() -proc isExprStart(p: TParser): bool = +proc isExprStart(p: Parser): bool = case p.tok.tokType of tkSymbol, tkAccent, tkOpr, tkNot, tkNil, tkCast, tkIf, tkFor, tkProc, tkFunc, tkIterator, tkBind, tkBuiltInMagics, - tkParLe, tkBracketLe, tkCurlyLe, tkIntLit..tkCharLit, tkVar, tkRef, tkPtr, - tkTuple, tkObject, tkWhen, tkCase, tkOut: + tkParLe, tkBracketLe, tkCurlyLe, tkIntLit..tkCustomLit, tkVar, tkRef, tkPtr, + tkEnum, tkTuple, tkObject, tkWhen, tkCase, tkOut, tkTry, tkBlock: result = true else: result = false -proc parseSymbolList(p: var TParser, result: PNode) = +proc parseSymbolList(p: var Parser, result: PNode) = # progress guaranteed while true: var s = parseSymbol(p, smAllowNil) @@ -1156,16 +1244,20 @@ proc parseSymbolList(p: var TParser, result: PNode) = if p.tok.tokType != tkComma: break getTok(p) optInd(p, s) + setEndInfo() -proc parseTypeDescKAux(p: var TParser, kind: TNodeKind, - mode: TPrimaryMode): PNode = - #| distinct = 'distinct' optInd typeDesc +proc parseTypeDescKAux(p: var Parser, kind: TNodeKind, + mode: PrimaryMode): PNode = result = newNodeP(kind, p) getTok(p) if p.tok.indent != -1 and p.tok.indent <= p.currInd: return optInd(p, result) + let isTypedef = mode == pmTypeDef and p.tok.tokType in {tkObject, tkTuple} if not isOperator(p.tok) and isExprStart(p): - result.add(primary(p, mode)) + if isTypedef: + result.add(parseTypeDefValue(p)) + else: + result.add(primary(p, mode)) if kind == nkDistinctTy and p.tok.tokType == tkSymbol: # XXX document this feature! var nodeKind: TNodeKind @@ -1179,11 +1271,14 @@ proc parseTypeDescKAux(p: var TParser, kind: TNodeKind, let list = newNodeP(nodeKind, p) result.add list parseSymbolList(p, list) + if mode == pmTypeDef and not isTypedef: + result = parseOperators(p, result, -1, mode) + setEndInfo() -proc parseVarTuple(p: var TParser): PNode +proc parseVarTuple(p: var Parser): PNode -proc parseFor(p: var TParser): PNode = - #| forStmt = 'for' (identWithPragma ^+ comma) 'in' expr colcom stmt +proc parseFor(p: var Parser): PNode = + #| forStmt = 'for' ((varTuple / identWithPragma) ^+ comma) 'in' expr colcom stmt #| forExpr = forStmt getTokNoInd(p) result = newNodeP(nkForStmt, p) @@ -1204,6 +1299,7 @@ proc parseFor(p: var TParser): PNode = result.add(parseExpr(p)) colcom(p, result) result.add(parseStmt(p)) + setEndInfo() template nimprettyDontTouch(body) = when defined(nimpretty): @@ -1212,7 +1308,7 @@ template nimprettyDontTouch(body) = when defined(nimpretty): dec p.em.keepIndents -proc parseExpr(p: var TParser): PNode = +proc parseExpr(p: var Parser): PNode = #| expr = (blockExpr #| | ifExpr #| | whenExpr @@ -1226,13 +1322,13 @@ proc parseExpr(p: var TParser): PNode = result = parseBlock(p) of tkIf: nimprettyDontTouch: - result = parseIfExpr(p, nkIfExpr) + result = parseIfOrWhenExpr(p, nkIfExpr) of tkFor: nimprettyDontTouch: result = parseFor(p) of tkWhen: nimprettyDontTouch: - result = parseIfExpr(p, nkWhenExpr) + result = parseIfOrWhenExpr(p, nkWhenStmt) of tkCase: # Currently we think nimpretty is good enough with case expressions, # so it is allowed to touch them: @@ -1242,81 +1338,74 @@ proc parseExpr(p: var TParser): PNode = nimprettyDontTouch: result = parseTry(p, isExpr=true) else: result = simpleExpr(p) - -proc parseEnum(p: var TParser): PNode -proc parseObject(p: var TParser): PNode -proc parseTypeClass(p: var TParser): PNode - -proc primary(p: var TParser, mode: TPrimaryMode): PNode = - #| typeKeyw = 'var' | 'out' | 'ref' | 'ptr' | 'shared' | 'tuple' - #| | 'proc' | 'iterator' | 'distinct' | 'object' | 'enum' - #| primary = typeKeyw optInd typeDesc - #| / prefixOperator* identOrLiteral primarySuffix* - #| / 'bind' primary + setEndInfo() + +proc parseEnum(p: var Parser): PNode +proc parseObject(p: var Parser): PNode +proc parseTypeClass(p: var Parser): PNode + +proc primary(p: var Parser, mode: PrimaryMode): PNode = + #| simplePrimary = SIGILLIKEOP? identOrLiteral primarySuffix* + #| commandStart = &('`'|IDENT|literal|'cast'|'addr'|'type'|'var'|'out'| + #| 'static'|'enum'|'tuple'|'object'|'proc') + #| primary = simplePrimary (commandStart expr (doBlock extraPostExprBlock*)?)? + #| / operatorB primary + #| / routineExpr + #| / rawTypeDesc + #| / prefixOperator primary + # XXX strong spaces need to be reflected in commandStart + # command part is handled in the primarySuffix proc + + # prefix operators: if isOperator(p.tok): + # Note 'sigil like' operators are currently not reflected in the grammar + # and should be removed for Nim 2.0, I don't think anybody uses them. let isSigil = isSigilLike(p.tok) result = newNodeP(nkPrefix, p) var a = newIdentNodeP(p.tok.ident, p) result.add(a) getTok(p) optInd(p, a) - if isSigil: - #XXX prefix operators + const identOrLiteralKinds = tkBuiltInMagics + {tkSymbol, tkAccent, tkNil, + tkIntLit..tkCustomLit, tkCast, tkOut, tkParLe, tkBracketLe, tkCurlyLe} + if isSigil and p.tok.tokType in identOrLiteralKinds: let baseInd = p.lex.currLineIndent - result.add(primary(p, pmSkipSuffix)) + result.add(identOrLiteral(p, mode)) result = primarySuffix(p, result, baseInd, mode) else: result.add(primary(p, pmNormal)) return - case p.tok.tokType: - of tkTuple: result = parseTuple(p, mode == pmTypeDef) - of tkProc: result = parseProcExpr(p, mode notin {pmTypeDesc, pmTypeDef}, nkLambda) - of tkFunc: result = parseProcExpr(p, mode notin {pmTypeDesc, pmTypeDef}, nkFuncDef) + case p.tok.tokType + of tkProc: + getTok(p) + result = parseProcExpr(p, mode != pmTypeDesc, nkLambda) + of tkFunc: + getTok(p) + result = parseProcExpr(p, mode != pmTypeDesc, nkFuncDef) of tkIterator: - result = parseProcExpr(p, mode notin {pmTypeDesc, pmTypeDef}, nkLambda) - if result.kind == nkLambda: result.transitionSonsKind(nkIteratorDef) - else: result.transitionSonsKind(nkIteratorTy) - of tkEnum: - if mode == pmTypeDef: - prettySection: - result = parseEnum(p) - else: - result = newNodeP(nkEnumTy, p) - getTok(p) - of tkObject: - if mode == pmTypeDef: - prettySection: - result = parseObject(p) - else: - result = newNodeP(nkObjectTy, p) - getTok(p) - of tkConcept: - if mode == pmTypeDef: - result = parseTypeClass(p) - else: - parMessage(p, "the 'concept' keyword is only valid in 'type' sections") + getTok(p) + result = parseProcExpr(p, mode != pmTypeDesc, nkIteratorDef) of tkBind: + # legacy syntax, no-op in current nim result = newNodeP(nkBind, p) getTok(p) optInd(p, result) result.add(primary(p, pmNormal)) - of tkVar: result = parseTypeDescKAux(p, nkVarTy, mode) - of tkRef: result = parseTypeDescKAux(p, nkRefTy, mode) - of tkPtr: result = parseTypeDescKAux(p, nkPtrTy, mode) - of tkDistinct: result = parseTypeDescKAux(p, nkDistinctTy, mode) + of tkTuple, tkEnum, tkObject, tkConcept, + tkVar, tkOut, tkRef, tkPtr, tkDistinct: + result = parseTypeDesc(p) else: let baseInd = p.lex.currLineIndent result = identOrLiteral(p, mode) - if mode != pmSkipSuffix: - result = primarySuffix(p, result, baseInd, mode) + result = primarySuffix(p, result, baseInd, mode) -proc binaryNot(p: var TParser; a: PNode): PNode = - if p.tok.tokType == tkNot: +proc binaryNot(p: var Parser; a: PNode): PNode = + if p.tok.tokType == tkNot and p.tok.indent < 0: let notOpr = newIdentNodeP(p.tok.ident, p) getTok(p) optInd(p, notOpr) - let b = parseExpr(p) + let b = primary(p, pmTypeDesc) result = newNodeP(nkInfix, p) result.add notOpr result.add a @@ -1324,32 +1413,91 @@ proc binaryNot(p: var TParser; a: PNode): PNode = else: result = a -proc parseTypeDesc(p: var TParser): PNode = - #| typeDesc = simpleExpr ('not' expr)? +proc parseTypeDesc(p: var Parser, fullExpr = false): PNode = + #| rawTypeDesc = (tupleType | routineType | 'enum' | 'object' | + #| ('var' | 'out' | 'ref' | 'ptr' | 'distinct') typeDesc?) + #| ('not' primary)? + #| typeDescExpr = (routineType / simpleExpr) ('not' primary)? + #| typeDesc = rawTypeDesc / typeDescExpr newlineWasSplitting(p) - result = simpleExpr(p, pmTypeDesc) + if fullExpr: + result = simpleExpr(p, pmTypeDesc) + else: + case p.tok.tokType + of tkTuple: + result = parseTuple(p, false) + of tkProc: + getTok(p) + result = parseProcExpr(p, false, nkLambda) + of tkIterator: + getTok(p) + result = parseProcExpr(p, false, nkIteratorDef) + of tkEnum: + result = newNodeP(nkEnumTy, p) + getTok(p) + of tkObject: + result = newNodeP(nkObjectTy, p) + getTok(p) + of tkConcept: + result = p.emptyNode + parMessage(p, "the 'concept' keyword is only valid in 'type' sections") + of tkVar: result = parseTypeDescKAux(p, nkVarTy, pmTypeDesc) + of tkOut: result = parseTypeDescKAux(p, nkOutTy, pmTypeDesc) + of tkRef: result = parseTypeDescKAux(p, nkRefTy, pmTypeDesc) + of tkPtr: result = parseTypeDescKAux(p, nkPtrTy, pmTypeDesc) + of tkDistinct: result = parseTypeDescKAux(p, nkDistinctTy, pmTypeDesc) + else: + result = simpleExpr(p, pmTypeDesc) result = binaryNot(p, result) + setEndInfo() -proc parseTypeDefAux(p: var TParser): PNode = - #| typeDefAux = simpleExpr ('not' expr)? - #| | 'concept' typeClass - result = simpleExpr(p, pmTypeDef) +proc parseTypeDefValue(p: var Parser): PNode = + #| typeDefValue = ((tupleDecl | enumDecl | objectDecl | conceptDecl | + #| ('ref' | 'ptr' | 'distinct') (tupleDecl | objectDecl)) + #| / (simpleExpr (exprEqExpr ^+ comma postExprBlocks?)?)) + #| ('not' primary)? + case p.tok.tokType + of tkTuple: result = parseTuple(p, true) + of tkRef: result = parseTypeDescKAux(p, nkRefTy, pmTypeDef) + of tkPtr: result = parseTypeDescKAux(p, nkPtrTy, pmTypeDef) + of tkDistinct: result = parseTypeDescKAux(p, nkDistinctTy, pmTypeDef) + of tkEnum: + prettySection: + result = parseEnum(p) + of tkObject: + prettySection: + result = parseObject(p) + of tkConcept: + result = parseTypeClass(p) + else: + result = simpleExpr(p, pmTypeDef) + if p.tok.tokType != tkNot: + if result.kind == nkCommand: + var isFirstParam = false + while p.tok.tokType == tkComma: + getTok(p) + optInd(p, result) + result.add(commandParam(p, isFirstParam, pmTypeDef)) + result = postExprBlocks(p, result) result = binaryNot(p, result) + setEndInfo() proc makeCall(n: PNode): PNode = ## Creates a call if the given node isn't already a call. if n.kind in nkCallKinds: result = n else: - result = newNodeI(nkCall, n.info) + result = newNode(nkCall, n.info) result.add n -proc postExprBlocks(p: var TParser, x: PNode): PNode = - #| postExprBlocks = ':' stmt? ( IND{=} doBlock - #| | IND{=} 'of' exprList ':' stmt - #| | IND{=} 'elif' expr ':' stmt - #| | IND{=} 'except' exprList ':' stmt - #| | IND{=} 'else' ':' stmt )* +proc postExprBlocks(p: var Parser, x: PNode): PNode = + #| extraPostExprBlock = ( IND{=} doBlock + #| | IND{=} 'of' exprList ':' stmt + #| | IND{=} 'elif' expr ':' stmt + #| | IND{=} 'except' optionalExprList ':' stmt + #| | IND{=} 'finally' ':' stmt + #| | IND{=} 'else' ':' stmt ) + #| postExprBlocks = (doBlock / ':' (extraPostExprBlock / stmt)) extraPostExprBlock* result = x if p.tok.indent >= 0: return @@ -1366,14 +1514,17 @@ proc postExprBlocks(p: var TParser, x: PNode): PNode = result = makeCall(result) getTok(p) skipComment(p, result) - if p.tok.tokType notin {tkOf, tkElif, tkElse, tkExcept}: + if not (p.tok.tokType in {tkOf, tkElif, tkElse, tkExcept, tkFinally} and sameInd(p)): var stmtList = newNodeP(nkStmtList, p) stmtList.add parseStmt(p) # to keep backwards compatibility (see tests/vm/tstringnil) - if stmtList[0].kind == nkStmtList: stmtList = stmtList[0] + if stmtList.firstSon.kind == nkStmtList: stmtList = stmtList.firstSon - stmtList.flags.incl nfBlockArg - if openingParams.kind != nkEmpty: + setNodeFlag stmtList, nfBlockArg + if openingParams.kind != nkEmpty or openingPragmas.kind != nkEmpty: + if openingParams.kind == nkEmpty: + openingParams = newNodeP(nkFormalParams, p) + openingParams.add(p.emptyNode) # return type result.add newProcNode(nkDo, stmtList.info, body = stmtList, params = openingParams, name = p.emptyNode, pattern = p.emptyNode, @@ -1391,7 +1542,7 @@ proc postExprBlocks(p: var TParser, x: PNode): PNode = getTok(p) nextBlock = parseDoBlock(p, info) else: - case nextToken: + case nextToken of tkOf: nextBlock = newNodeP(nkOfBranch, p) exprList(p, tkColon, nextBlock) @@ -1402,7 +1553,10 @@ proc postExprBlocks(p: var TParser, x: PNode): PNode = nextBlock.add parseExpr(p) of tkExcept: nextBlock = newNodeP(nkExceptBranch, p) - exprList(p, tkColon, nextBlock) + optionalExprList(p, tkColon, nextBlock) + of tkFinally: + nextBlock = newNodeP(nkFinally, p) + getTok(p) of tkElse: nextBlock = newNodeP(nkElse, p) getTok(p) @@ -1410,21 +1564,19 @@ proc postExprBlocks(p: var TParser, x: PNode): PNode = eat(p, tkColon) nextBlock.add parseStmt(p) - nextBlock.flags.incl nfBlockArg + setNodeFlag nextBlock, nfBlockArg result.add nextBlock - if nextBlock.kind == nkElse: break + if nextBlock.kind in {nkElse, nkFinally}: break else: if openingParams.kind != nkEmpty: parMessage(p, "expected ':'") -proc parseExprStmt(p: var TParser): PNode = - #| exprStmt = simpleExpr - #| (( '=' optInd expr colonBody? ) - #| / ( expr ^+ comma - #| postExprBlocks - #| ))? - var a = simpleExpr(p) +proc parseExprStmt(p: var Parser): PNode = + #| exprStmt = simpleExpr postExprBlocks? + #| / simplePrimary (exprEqExpr ^+ comma) postExprBlocks? + #| / simpleExpr '=' optInd (expr postExprBlocks?) + var a = simpleExpr(p, pmTrySimple) if p.tok.tokType == tkEquals: result = newNodeP(nkAsgn, p) getTok(p) @@ -1434,27 +1586,25 @@ proc parseExprStmt(p: var TParser): PNode = result.add(a) result.add(b) else: - # simpleExpr parsed 'p a' from 'p a, b'? var isFirstParam = false - if p.tok.indent < 0 and p.tok.tokType == tkComma and a.kind == nkCommand: - result = a + # if an expression is starting here, a simplePrimary was parsed and + # this is the start of a command + if p.tok.indent < 0 and isExprStart(p): + result = newTree(nkCommand, a.info, a) + let baseIndent = p.currInd while true: - getTok(p) - optInd(p, result) result.add(commandParam(p, isFirstParam, pmNormal)) - if p.tok.tokType != tkComma: break - elif p.tok.indent < 0 and isExprStart(p): - result = newNode(nkCommand, a.info, @[a]) - while true: - result.add(commandParam(p, isFirstParam, pmNormal)) - if p.tok.tokType != tkComma: break + if p.tok.tokType != tkComma or + (p.tok.indent >= 0 and p.tok.indent < baseIndent): + break getTok(p) optInd(p, result) else: result = a result = postExprBlocks(p, result) + setEndInfo() -proc parseModuleName(p: var TParser, kind: TNodeKind): PNode = +proc parseModuleName(p: var Parser, kind: TNodeKind): PNode = result = parseExpr(p) when false: # parseExpr already handles 'as' syntax ... @@ -1464,8 +1614,9 @@ proc parseModuleName(p: var TParser, kind: TNodeKind): PNode = getTok(p) result.add(a) result.add(parseExpr(p)) + setEndInfo() -proc parseImport(p: var TParser, kind: TNodeKind): PNode = +proc parseImport(p: var Parser, kind: TNodeKind): PNode = #| importStmt = 'import' optInd expr #| ((comma expr)* #| / 'except' optInd (expr ^+ comma)) @@ -1492,8 +1643,9 @@ proc parseImport(p: var TParser, kind: TNodeKind): PNode = getTok(p) optInd(p, a) #expectNl(p) + setEndInfo() -proc parseIncludeStmt(p: var TParser): PNode = +proc parseIncludeStmt(p: var Parser): PNode = #| includeStmt = 'include' optInd expr ^+ comma result = newNodeP(nkIncludeStmt, p) getTok(p) # skip `import` or `include` @@ -1508,8 +1660,9 @@ proc parseIncludeStmt(p: var TParser): PNode = getTok(p) optInd(p, a) #expectNl(p) + setEndInfo() -proc parseFromStmt(p: var TParser): PNode = +proc parseFromStmt(p: var Parser): PNode = #| fromStmt = 'from' expr 'import' optInd expr (comma expr)* result = newNodeP(nkFromStmt, p) getTok(p) # skip `from` @@ -1528,14 +1681,15 @@ proc parseFromStmt(p: var TParser): PNode = getTok(p) optInd(p, a) #expectNl(p) + setEndInfo() -proc parseReturnOrRaise(p: var TParser, kind: TNodeKind): PNode = +proc parseReturnOrRaise(p: var Parser, kind: TNodeKind): PNode = #| returnStmt = 'return' optInd expr? #| raiseStmt = 'raise' optInd expr? #| yieldStmt = 'yield' optInd expr? #| discardStmt = 'discard' optInd expr? #| breakStmt = 'break' optInd expr? - #| continueStmt = 'break' optInd expr? + #| continueStmt = 'continue' optInd expr? result = newNodeP(kind, p) getTok(p) if p.tok.tokType == tkComment: @@ -1549,8 +1703,9 @@ proc parseReturnOrRaise(p: var TParser, kind: TNodeKind): PNode = var e = parseExpr(p) e = postExprBlocks(p, e) result.add(e) + setEndInfo() -proc parseIfOrWhen(p: var TParser, kind: TNodeKind): PNode = +proc parseIfOrWhen(p: var Parser, kind: TNodeKind): PNode = #| condStmt = expr colcom stmt COMMENT? #| (IND{=} 'elif' expr colcom stmt)* #| (IND{=} 'else' colcom stmt)? @@ -1573,8 +1728,34 @@ proc parseIfOrWhen(p: var TParser, kind: TNodeKind): PNode = colcom(p, branch) branch.add(parseStmt(p)) result.add(branch) + setEndInfo() -proc parseWhile(p: var TParser): PNode = +proc parseIfOrWhenExpr(p: var Parser, kind: TNodeKind): PNode = + #| condExpr = expr colcom stmt optInd + #| ('elif' expr colcom stmt optInd)* + #| 'else' colcom stmt + #| ifExpr = 'if' condExpr + #| whenExpr = 'when' condExpr + result = newNodeP(kind, p) + while true: + getTok(p) # skip `if`, `when`, `elif` + var branch = newNodeP(nkElifExpr, p) + optInd(p, branch) + branch.add(parseExpr(p)) + colcom(p, branch) + branch.add(parseStmt(p)) + skipComment(p, branch) + result.add(branch) + if p.tok.tokType != tkElif: break + if p.tok.tokType == tkElse: + var branch = newNodeP(nkElseExpr, p) + eat(p, tkElse) + colcom(p, branch) + branch.add(parseStmt(p)) + result.add(branch) + setEndInfo() + +proc parseWhile(p: var Parser): PNode = #| whileStmt = 'while' expr colcom stmt result = newNodeP(nkWhileStmt, p) getTok(p) @@ -1582,8 +1763,9 @@ proc parseWhile(p: var TParser): PNode = result.add(parseExpr(p)) colcom(p, result) result.add(parseStmt(p)) + setEndInfo() -proc parseCase(p: var TParser): PNode = +proc parseCase(p: var Parser): PNode = #| ofBranch = 'of' exprList colcom stmt #| ofBranches = ofBranch (IND{=} ofBranch)* #| (IND{=} 'elif' expr colcom stmt)* @@ -1629,24 +1811,27 @@ proc parseCase(p: var TParser): PNode = if wasIndented: p.currInd = oldInd + setEndInfo() -proc parseTry(p: var TParser; isExpr: bool): PNode = +proc parseTry(p: var Parser; isExpr: bool): PNode = #| tryStmt = 'try' colcom stmt &(IND{=}? 'except'|'finally') - #| (IND{=}? 'except' exprList colcom stmt)* + #| (IND{=}? 'except' optionalExprList colcom stmt)* #| (IND{=}? 'finally' colcom stmt)? #| tryExpr = 'try' colcom stmt &(optInd 'except'|'finally') - #| (optInd 'except' exprList colcom stmt)* + #| (optInd 'except' optionalExprList colcom stmt)* #| (optInd 'finally' colcom stmt)? result = newNodeP(nkTryStmt, p) + let parentIndent = p.currInd # isExpr getTok(p) colcom(p, result) result.add(parseStmt(p)) var b: PNode = nil - while sameOrNoInd(p) or isExpr: + + while sameOrNoInd(p) or (isExpr and parentIndent <= p.tok.indent): case p.tok.tokType of tkExcept: b = newNodeP(nkExceptBranch, p) - exprList(p, tkColon, b) + optionalExprList(p, tkColon, b) of tkFinally: b = newNodeP(nkFinally, p) getTok(p) @@ -1655,15 +1840,16 @@ proc parseTry(p: var TParser; isExpr: bool): PNode = b.add(parseStmt(p)) result.add(b) if b == nil: parMessage(p, "expected 'except'") + setEndInfo() -proc parseExceptBlock(p: var TParser, kind: TNodeKind): PNode = - #| exceptBlock = 'except' colcom stmt +proc parseExceptBlock(p: var Parser, kind: TNodeKind): PNode = result = newNodeP(kind, p) getTok(p) colcom(p, result) result.add(parseStmt(p)) + setEndInfo() -proc parseBlock(p: var TParser): PNode = +proc parseBlock(p: var Parser): PNode = #| blockStmt = 'block' symbol? colcom stmt #| blockExpr = 'block' symbol? colcom stmt result = newNodeP(nkBlockStmt, p) @@ -1672,16 +1858,18 @@ proc parseBlock(p: var TParser): PNode = else: result.add(parseSymbol(p)) colcom(p, result) result.add(parseStmt(p)) + setEndInfo() -proc parseStaticOrDefer(p: var TParser; k: TNodeKind): PNode = +proc parseStaticOrDefer(p: var Parser; k: TNodeKind): PNode = #| staticStmt = 'static' colcom stmt #| deferStmt = 'defer' colcom stmt result = newNodeP(k, p) getTok(p) colcom(p, result) result.add(parseStmt(p)) + setEndInfo() -proc parseAsm(p: var TParser): PNode = +proc parseAsm(p: var Parser): PNode = #| asmStmt = 'asm' pragma? (STR_LIT | RSTR_LIT | TRIPLESTR_LIT) result = newNodeP(nkAsmStmt, p) getTokNoInd(p) @@ -1696,8 +1884,9 @@ proc parseAsm(p: var TParser): PNode = result.add(p.emptyNode) return getTok(p) + setEndInfo() -proc parseGenericParam(p: var TParser): PNode = +proc parseGenericParam(p: var Parser): PNode = #| genericParam = symbol (comma symbol)* (colon expr)? ('=' optInd expr)? var a: PNode result = newNodeP(nkIdentDefs, p) @@ -1731,8 +1920,9 @@ proc parseGenericParam(p: var TParser): PNode = result.add(parseExpr(p)) else: result.add(p.emptyNode) + setEndInfo() -proc parseGenericParamList(p: var TParser): PNode = +proc parseGenericParamList(p: var Parser): PNode = #| genericParamList = '[' optInd #| genericParam ^* (comma/semicolon) optPar ']' result = newNodeP(nkGenericParams, p) @@ -1749,20 +1939,28 @@ proc parseGenericParamList(p: var TParser): PNode = skipComment(p, a) optPar(p) eat(p, tkBracketRi) + setEndInfo() -proc parsePattern(p: var TParser): PNode = +proc parsePattern(p: var Parser): PNode = #| pattern = '{' stmt '}' eat(p, tkCurlyLe) result = parseStmt(p) eat(p, tkCurlyRi) + setEndInfo() -proc parseRoutine(p: var TParser, kind: TNodeKind): PNode = +proc parseRoutine(p: var Parser, kind: TNodeKind): PNode = #| indAndComment = (IND{>} COMMENT)? | COMMENT? #| routine = optInd identVis pattern? genericParamList? #| paramListColon pragma? ('=' COMMENT? stmt)? indAndComment result = newNodeP(kind, p) getTok(p) optInd(p, result) + if kind in {nkProcDef, nkLambda, nkIteratorDef, nkFuncDef} and + p.tok.tokType notin {tkSymbol, tokKeywordLow..tokKeywordHigh, tkAccent}: + # no name; lambda or proc type + # in every context that we can parse a routine, we can also parse these + result = parseProcExpr(p, true, if kind == nkProcDef: nkLambda else: kind) + return result.add(identVis(p)) if p.tok.tokType == tkCurlyLe and p.validInd: result.add(p.parsePattern) else: result.add(p.emptyNode) @@ -1775,25 +1973,34 @@ proc parseRoutine(p: var TParser, kind: TNodeKind): PNode = else: result.add(p.emptyNode) # empty exception tracking: result.add(p.emptyNode) - if p.tok.tokType == tkEquals and p.validInd: + let maybeMissEquals = p.tok.tokType != tkEquals + if (not maybeMissEquals) and p.validInd: getTok(p) skipComment(p, result) result.add(parseStmt(p)) else: result.add(p.emptyNode) - indAndComment(p, result) - -proc newCommentStmt(p: var TParser): PNode = + indAndComment(p, result, maybeMissEquals) + let body = result.lastSon + if body.kind == nkStmtList and body.hasSon and body.firstSon.comment.len > 0 and body.firstSon.kind != nkCommentStmt: + if result.comment.len == 0: + # proc fn*(a: int): int = a ## foo + # => moves comment `foo` to `fn` + result.comment = body.firstSon.comment + body.firstSon.comment = "" + #else: + # assert false, p.lex.config$body.info # avoids hard to track bugs, fail early. + # Yeah, that worked so well. There IS a bug in this logic, now what? + setEndInfo() + +proc newCommentStmt(p: var Parser): PNode = #| commentStmt = COMMENT result = newNodeP(nkCommentStmt, p) result.comment = p.tok.literal getTok(p) -type - TDefParser = proc (p: var TParser): PNode {.nimcall.} - -proc parseSection(p: var TParser, kind: TNodeKind, - defparser: TDefParser): PNode = +proc parseSection(p: var Parser, kind: TNodeKind, + defparser: proc (p: var Parser): PNode {.nimcall.}): PNode = #| section(RULE) = COMMENT? RULE / (IND{>} (RULE / COMMENT)^+IND{=} DED) result = newNodeP(kind, p) if kind != nkTypeSection: getTok(p) @@ -1814,15 +2021,16 @@ proc parseSection(p: var TParser, kind: TNodeKind, else: parMessage(p, errIdentifierExpected, p.tok) break - if result.len == 0: parMessage(p, errIdentifierExpected, p.tok) + if not result.hasSon: parMessage(p, errIdentifierExpected, p.tok) elif p.tok.tokType in {tkSymbol, tkAccent, tkParLe} and p.tok.indent < 0: # tkParLe is allowed for ``var (x, y) = ...`` tuple parsing result.add(defparser(p)) else: parMessage(p, errIdentifierExpected, p.tok) + setEndInfo() -proc parseEnum(p: var TParser): PNode = - #| enum = 'enum' optInd (symbol pragma? optInd ('=' optInd expr COMMENT?)? comma?)+ +proc parseEnum(p: var Parser): PNode = + #| enumDecl = 'enum' optInd (symbol pragma? optInd ('=' optInd expr COMMENT?)? comma?)+ result = newNodeP(nkEnumTy, p) getTok(p) result.add(p.emptyNode) @@ -1835,7 +2043,7 @@ proc parseEnum(p: var TParser): PNode = var symPragma = a var pragma: PNode - if p.tok.tokType == tkCurlyDotLe: + if (p.tok.indent < 0 or p.tok.indent >= p.currInd) and p.tok.tokType == tkCurlyDotLe: pragma = optPragmas(p) symPragma = newNodeP(nkPragmaExpr, p) symPragma.add(a) @@ -1864,11 +2072,12 @@ proc parseEnum(p: var TParser): PNode = if p.tok.indent >= 0 and p.tok.indent <= p.currInd or p.tok.tokType == tkEof: break - if result.len <= 1: + if not result.has2Sons: parMessage(p, errIdentifierExpected, p.tok) + setEndInfo() -proc parseObjectPart(p: var TParser): PNode -proc parseObjectWhen(p: var TParser): PNode = +proc parseObjectPart(p: var Parser): PNode +proc parseObjectWhen(p: var Parser): PNode = #| objectWhen = 'when' expr colcom objectPart COMMENT? #| ('elif' expr colcom objectPart COMMENT?)* #| ('else' colcom objectPart COMMENT?)? @@ -1891,22 +2100,19 @@ proc parseObjectWhen(p: var TParser): PNode = branch.add(parseObjectPart(p)) flexComment(p, branch) result.add(branch) + setEndInfo() -proc parseObjectCase(p: var TParser): PNode = +proc parseObjectCase(p: var Parser): PNode = #| objectBranch = 'of' exprList colcom objectPart #| objectBranches = objectBranch (IND{=} objectBranch)* #| (IND{=} 'elif' expr colcom objectPart)* #| (IND{=} 'else' colcom objectPart)? - #| objectCase = 'case' identWithPragma ':' typeDesc ':'? COMMENT? + #| objectCase = 'case' declColonEquals ':'? COMMENT? #| (IND{>} objectBranches DED #| | IND{=} objectBranches) result = newNodeP(nkRecCase, p) getTokNoInd(p) - var a = newNodeP(nkIdentDefs, p) - a.add(identWithPragma(p)) - eat(p, tkColon) - a.add(parseTypeDesc(p)) - a.add(p.emptyNode) + var a = parseIdentColonEquals(p, {withPragma}) result.add(a) if p.tok.tokType == tkColon: getTok(p) flexComment(p, result) @@ -1936,8 +2142,9 @@ proc parseObjectCase(p: var TParser): PNode = if b.kind == nkElse: break if wasIndented: p.currInd = oldInd + setEndInfo() -proc parseObjectPart(p: var TParser): PNode = +proc parseObjectPart(p: var Parser): PNode = #| objectPart = IND{>} objectPart^+IND{=} DED #| / objectWhen / objectCase / 'nil' / 'discard' / declColonEquals if realInd(p): @@ -1951,7 +2158,7 @@ proc parseObjectPart(p: var TParser): PNode = else: parMessage(p, errIdentifierExpected, p.tok) break - else: + elif sameOrNoInd(p): case p.tok.tokType of tkWhen: result = parseObjectWhen(p) @@ -1966,17 +2173,15 @@ proc parseObjectPart(p: var TParser): PNode = getTok(p) else: result = p.emptyNode + else: + result = p.emptyNode + setEndInfo() -proc parseObject(p: var TParser): PNode = - #| object = 'object' pragma? ('of' typeDesc)? COMMENT? objectPart +proc parseObject(p: var Parser): PNode = + #| objectDecl = 'object' ('of' typeDesc)? COMMENT? objectPart result = newNodeP(nkObjectTy, p) getTok(p) - if p.tok.tokType == tkCurlyDotLe and p.validInd: - # Deprecated since v0.20.0 - parMessage(p, warnDeprecated, "type pragmas follow the type name; this form of writing pragmas is deprecated") - result.add(parsePragma(p)) - else: - result.add(p.emptyNode) + result.add(p.emptyNode) # compatibility with old pragma node if p.tok.tokType == tkOf and p.tok.indent < 0: var a = newNodeP(nkOfInherit, p) getTok(p) @@ -1989,12 +2194,15 @@ proc parseObject(p: var TParser): PNode = # an initial IND{>} HAS to follow: if not realInd(p): result.add(p.emptyNode) - return - result.add(parseObjectPart(p)) + else: + result.add(parseObjectPart(p)) + setEndInfo() -proc parseTypeClassParam(p: var TParser): PNode = - let modifier = case p.tok.tokType - of tkOut, tkVar: nkVarTy +proc parseTypeClassParam(p: var Parser): PNode = + let modifier = + case p.tok.tokType + of tkVar: nkVarTy + of tkOut: nkOutTy of tkPtr: nkPtrTy of tkRef: nkRefTy of tkStatic: nkStaticTy @@ -2007,19 +2215,26 @@ proc parseTypeClassParam(p: var TParser): PNode = result.add(p.parseSymbol) else: result = p.parseSymbol + setEndInfo() -proc parseTypeClass(p: var TParser): PNode = - #| typeClassParam = ('var' | 'out')? symbol - #| typeClass = typeClassParam ^* ',' (pragma)? ('of' typeDesc ^* ',')? +proc parseTypeClass(p: var Parser): PNode = + #| conceptParam = ('var' | 'out' | 'ptr' | 'ref' | 'static' | 'type')? symbol + #| conceptDecl = 'concept' conceptParam ^* ',' (pragma)? ('of' typeDesc ^* ',')? #| &IND{>} stmt result = newNodeP(nkTypeClassTy, p) getTok(p) - var args = newNodeP(nkArgList, p) - result.add(args) - args.add(p.parseTypeClassParam) - while p.tok.tokType == tkComma: - getTok(p) + if p.tok.tokType == tkComment: + skipComment(p, result) + + if p.tok.indent < 0: + var args = newNodeP(nkArgList, p) + result.add(args) args.add(p.parseTypeClassParam) + while p.tok.tokType == tkComma: + getTok(p) + args.add(p.parseTypeClassParam) + else: + result.add(p.emptyNode) # see ast.isNewStyleConcept if p.tok.tokType == tkCurlyDotLe and p.validInd: result.add(parsePragma(p)) else: @@ -2039,45 +2254,33 @@ proc parseTypeClass(p: var TParser): PNode = skipComment(p, result) # an initial IND{>} HAS to follow: if not realInd(p): + if result.isNewStyleConcept: + parMessage(p, "routine expected, but found '$1' (empty new-styled concepts are not allowed)", p.tok) result.add(p.emptyNode) else: result.add(parseStmt(p)) + setEndInfo() -proc parseTypeDef(p: var TParser): PNode = +proc parseTypeDef(p: var Parser): PNode = #| - #| typeDef = identWithPragmaDot genericParamList? '=' optInd typeDefAux - #| indAndComment? / identVisDot genericParamList? pragma '=' optInd typeDefAux + #| typeDef = identVisDot genericParamList? pragma '=' optInd typeDefValue #| indAndComment? result = newNodeP(nkTypeDef, p) var identifier = identVis(p, allowDot=true) var identPragma = identifier var pragma: PNode var genericParam: PNode - var noPragmaYet = true - - if p.tok.tokType == tkCurlyDotLe: - pragma = optPragmas(p) - identPragma = newNodeP(nkPragmaExpr, p) - identPragma.add(identifier) - identPragma.add(pragma) - noPragmaYet = false if p.tok.tokType == tkBracketLe and p.validInd: - if not noPragmaYet: - # Deprecated since v0.20.0 - parMessage(p, warnDeprecated, "pragma before generic parameter list is deprecated") genericParam = parseGenericParamList(p) else: genericParam = p.emptyNode - if noPragmaYet: - pragma = optPragmas(p) - if pragma.kind != nkEmpty: - identPragma = newNodeP(nkPragmaExpr, p) - identPragma.add(identifier) - identPragma.add(pragma) - elif p.tok.tokType == tkCurlyDotLe: - parMessage(p, errGenerated, "pragma already present") + pragma = optPragmas(p) + if pragma.kind != nkEmpty: + identPragma = newNodeP(nkPragmaExpr, p) + identPragma.add(identifier) + identPragma.add(pragma) result.add(identPragma) result.add(genericParam) @@ -2086,28 +2289,41 @@ proc parseTypeDef(p: var TParser): PNode = result.info = parLineInfo(p) getTok(p) optInd(p, result) - result.add(parseTypeDefAux(p)) + result.add(parseTypeDefValue(p)) else: result.add(p.emptyNode) indAndComment(p, result) # special extension! + setEndInfo() -proc parseVarTuple(p: var TParser): PNode = - #| varTuple = '(' optInd identWithPragma ^+ comma optPar ')' '=' optInd expr +proc parseVarTuple(p: var Parser): PNode = + #| varTupleLhs = '(' optInd (identWithPragma / varTupleLhs) ^+ comma optPar ')' (':' optInd typeDescExpr)? + #| varTuple = varTupleLhs '=' optInd expr result = newNodeP(nkVarTuple, p) getTok(p) # skip '(' optInd(p, result) # progress guaranteed - while p.tok.tokType in {tkSymbol, tkAccent}: - var a = identWithPragma(p, allowDot=true) + while p.tok.tokType in {tkSymbol, tkAccent, tkParLe}: + var a: PNode + if p.tok.tokType == tkParLe: + a = parseVarTuple(p) + a.add(p.emptyNode) + else: + a = identWithPragma(p, allowDot=true) result.add(a) if p.tok.tokType != tkComma: break getTok(p) skipComment(p, a) - result.add(p.emptyNode) # no type desc optPar(p) eat(p, tkParRi) + if p.tok.tokType == tkColon: + getTok(p) + optInd(p, result) + result.add(parseTypeDesc(p, fullExpr = true)) + else: + result.add(p.emptyNode) # no type desc + setEndInfo() -proc parseVariable(p: var TParser): PNode = +proc parseVariable(p: var Parser): PNode = #| colonBody = colcom stmt postExprBlocks? #| variable = (varTuple / identColonEquals) colonBody? indAndComment if p.tok.tokType == tkParLe: @@ -2116,10 +2332,11 @@ proc parseVariable(p: var TParser): PNode = optInd(p, result) result.add(parseExpr(p)) else: result = parseIdentColonEquals(p, {withPragma, withDot}) - result[^1] = postExprBlocks(p, result[^1]) + result.setLastSon postExprBlocks(p, result.lastSon) indAndComment(p, result) + setEndInfo() -proc parseConstant(p: var TParser): PNode = +proc parseConstant(p: var Parser): PNode = #| constant = (varTuple / identWithPragma) (colon typeDesc)? '=' optInd expr indAndComment if p.tok.tokType == tkParLe: result = parseVarTuple(p) else: @@ -2134,11 +2351,12 @@ proc parseConstant(p: var TParser): PNode = eat(p, tkEquals) optInd(p, result) #add(result, parseStmtListExpr(p)) - result.add(parseExpr(p)) - result[^1] = postExprBlocks(p, result[^1]) + let a = parseExpr(p) + result.add postExprBlocks(p, a) indAndComment(p, result) + setEndInfo() -proc parseBind(p: var TParser, k: TNodeKind): PNode = +proc parseBind(p: var Parser, k: TNodeKind): PNode = #| bindStmt = 'bind' optInd qualifiedIdent ^+ comma #| mixinStmt = 'mixin' optInd qualifiedIdent ^+ comma result = newNodeP(k, p) @@ -2152,19 +2370,21 @@ proc parseBind(p: var TParser, k: TNodeKind): PNode = getTok(p) optInd(p, a) #expectNl(p) + setEndInfo() -proc parseStmtPragma(p: var TParser): PNode = +proc parseStmtPragma(p: var Parser): PNode = #| pragmaStmt = pragma (':' COMMENT? stmt)? result = parsePragma(p) if p.tok.tokType == tkColon and p.tok.indent < 0: let a = result - result = newNodeI(nkPragmaBlock, a.info) + result = newNode(nkPragmaBlock, a.info) getTok(p) skipComment(p, result) result.add a result.add parseStmt(p) + setEndInfo() -proc simpleStmt(p: var TParser): PNode = +proc simpleStmt(p: var Parser): PNode = #| simpleStmt = ((returnStmt | raiseStmt | yieldStmt | discardStmt | breakStmt #| | continueStmt | pragmaStmt | importStmt | exportStmt | fromStmt #| | includeStmt | commentStmt) / exprStmt) COMMENT? @@ -2187,12 +2407,13 @@ proc simpleStmt(p: var TParser): PNode = else: result = p.emptyNode if result.kind notin {nkEmpty, nkCommentStmt}: skipComment(p, result) -proc complexOrSimpleStmt(p: var TParser): PNode = +proc complexOrSimpleStmt(p: var Parser): PNode = #| complexOrSimpleStmt = (ifStmt | whenStmt | whileStmt #| | tryStmt | forStmt #| | blockStmt | staticStmt | deferStmt | asmStmt #| | 'proc' routine #| | 'method' routine + #| | 'func' routine #| | 'iterator' routine #| | 'macro' routine #| | 'template' routine @@ -2246,7 +2467,7 @@ proc complexOrSimpleStmt(p: var TParser): PNode = of tkUsing: result = parseSection(p, nkUsingStmt, parseVariable) else: result = simpleStmt(p) -proc parseStmt(p: var TParser): PNode = +proc parseStmt(p: var Parser): PNode = #| stmt = (IND{>} complexOrSimpleStmt^+(IND{=} / ';') DED) #| / simpleStmt ^+ ';' if p.tok.indent > p.currInd: @@ -2269,24 +2490,23 @@ proc parseStmt(p: var TParser): PNode = # deprecate this syntax later break p.hasProgress = false - var a = complexOrSimpleStmt(p) - if a.kind != nkEmpty: - result.add(a) + if p.tok.tokType in {tkElse, tkElif}: + break # Allow this too, see tests/parser/tifexprs + + let a = complexOrSimpleStmt(p) + if a.kind == nkEmpty and not p.hasProgress: + parMessage(p, errExprExpected, p.tok) + break else: - # This is done to make the new 'if' expressions work better. - # XXX Eventually we need to be more strict here. - if p.tok.tokType notin {tkElse, tkElif}: - parMessage(p, errExprExpected, p.tok) - getTok(p) - else: - break + result.add a + if not p.hasProgress and p.tok.tokType == tkEof: break else: # the case statement is only needed for better error messages: case p.tok.tokType of tkIf, tkWhile, tkCase, tkTry, tkFor, tkBlock, tkAsm, tkProc, tkFunc, tkIterator, tkMacro, tkType, tkConst, tkWhen, tkVar: - parMessage(p, "complex statement requires indentation") + parMessage(p, "nestable statement requires indentation") result = p.emptyNode else: if p.inSemiStmtList > 0: @@ -2305,23 +2525,13 @@ proc parseStmt(p: var TParser): PNode = if p.tok.tokType != tkSemiColon: break getTok(p) if err and p.tok.tokType == tkEof: break + setEndInfo() -proc parseAll(p: var TParser): PNode = - ## Parses the rest of the input stream held by the parser into a PNode. - result = newNodeP(nkStmtList, p) - while p.tok.tokType != tkEof: - p.hasProgress = false - var a = complexOrSimpleStmt(p) - if a.kind != nkEmpty and p.hasProgress: - result.add(a) - else: - parMessage(p, errExprExpected, p.tok) - # bugfix: consume a token here to prevent an endless loop: - getTok(p) - if p.tok.indent != 0: - parMessage(p, errInvalidIndentation) +proc checkFirstLineIndentation*(p: var Parser) = + if p.tok.indent != 0 and tsLeading in p.tok.spacing: + parMessage(p, errInvalidIndentation) -proc parseTopLevelStmt(p: var TParser): PNode = +proc parseTopLevelStmt*(p: var Parser): PNode = ## Implements an iterator which, when called repeatedly, returns the next ## top-level statement or emptyNode if end of stream. result = p.emptyNode @@ -2349,10 +2559,21 @@ proc parseTopLevelStmt(p: var TParser): PNode = result = complexOrSimpleStmt(p) if result.kind == nkEmpty: parMessage(p, errExprExpected, p.tok) break + setEndInfo() + +proc parseAll*(p: var Parser): PNode = + ## Parses the rest of the input stream held by the parser into a PNode. + result = newNodeP(nkStmtList, p) + while true: + let nextStmt = p.parseTopLevelStmt() + if nextStmt.kind == nkEmpty: + break + result &= nextStmt + setEndInfo() proc parseString*(s: string; cache: IdentCache; config: ConfigRef; filename: string = ""; line: int = 0; - errorHandler: TErrorHandler = nil): PNode = + errorHandler: ErrorHandler = nil): PNode = ## Parses a string into an AST, returning the top node. ## `filename` and `line`, although optional, provide info so that the ## compiler can generate correct error messages referring to the original @@ -2360,9 +2581,10 @@ proc parseString*(s: string; cache: IdentCache; config: ConfigRef; var stream = llStreamOpen(s) stream.lineOffset = line - var parser: TParser - parser.lex.errorHandler = errorHandler - openParser(parser, AbsoluteFile filename, stream, cache, config) + var p = Parser() + p.lex.errorHandler = errorHandler + openParser(p, AbsoluteFile filename, stream, cache, config) - result = parser.parseAll - closeParser(parser) + result = p.parseAll + closeParser(p) + setEndInfo() |