/** @file LexRust.cxx ** Lexer for Rust. ** ** Copyright (c) 2013 by SiegeLord ** Converted to lexer object and added further folding features/properties by "Udo Lechner" **/ // Copyright 1998-2005 by Neil Hodgson // The License.txt file describes the conditions under which this software may be distributed. #include #include #include #include #include #include #include #include #include "ILexer.h" #include "Scintilla.h" #include "SciLexer.h" #include "PropSetSimple.h" #include "WordList.h" #include "LexAccessor.h" #include "Accessor.h" #include "StyleContext.h" #include "CharacterSet.h" #include "LexerModule.h" #include "OptionSet.h" #ifdef SCI_NAMESPACE using namespace Scintilla; #endif static const int NUM_RUST_KEYWORD_LISTS = 7; static const int MAX_RUST_IDENT_CHARS = 1023; static bool IsStreamCommentStyle(int style) { return style == SCE_RUST_COMMENTBLOCK || style == SCE_RUST_COMMENTBLOCKDOC; } // Options used for LexerRust struct OptionsRust { bool fold; bool foldSyntaxBased; bool foldComment; bool foldCommentMultiline; bool foldCommentExplicit; std::string foldExplicitStart; std::string foldExplicitEnd; bool foldExplicitAnywhere; bool foldCompact; int foldAtElseInt; bool foldAtElse; OptionsRust() { fold = false; foldSyntaxBased = true; foldComment = false; foldCommentMultiline = true; foldCommentExplicit = true; foldExplicitStart = ""; foldExplicitEnd = ""; foldExplicitAnywhere = false; foldCompact = true; foldAtElseInt = -1; foldAtElse = false; } }; static const char * const rustWordLists[NUM_RUST_KEYWORD_LISTS + 1] = { "Primary keywords and identifiers", "Built in types", "Other keywords", "Keywords 4", "Keywords 5", "Keywords 6", "Keywords 7", 0, }; struct OptionSetRust : public OptionSet { OptionSetRust() { DefineProperty("fold", &OptionsRust::fold); DefineProperty("fold.comment", &OptionsRust::foldComment); DefineProperty("fold.compact", &OptionsRust::foldCompact); DefineProperty("fold.at.else", &OptionsRust::foldAtElse); DefineProperty("fold.rust.syntax.based", &OptionsRust::foldSyntaxBased, "Set this property to 0 to disable syntax based folding."); DefineProperty("fold.rust.comment.multiline", &OptionsRust::foldCommentMultiline, "Set this property to 0 to disable folding multi-line comments when fold.comment=1."); DefineProperty("fold.rust.comment.explicit", &OptionsRust::foldCommentExplicit, "Set this property to 0 to disable folding explicit fold points when fold.comment=1."); DefineProperty("fold.rust.explicit.start", &OptionsRust::foldExplicitStart, "The string to use for explicit fold start points, replacing the standard //{."); DefineProperty("fold.rust.explicit.end", &OptionsRust::foldExplicitEnd, "The string to use for explicit fold end points, replacing the standard //}."); DefineProperty("fold.rust.explicit.anywhere", &OptionsRust::foldExplicitAnywhere, "Set this property to 1 to enable explicit fold points anywhere, not just in line comments."); DefineProperty("lexer.rust.fold.at.else", &OptionsRust::foldAtElseInt, "This option enables Rust folding on a \"} else {\" line of an if statement."); DefineWordListSets(rustWordLists); } }; class LexerRust : public ILexer { WordList keywords[NUM_RUST_KEYWORD_LISTS]; OptionsRust options; OptionSetRust osRust; public: virtual ~LexerRust() { } void SCI_METHOD Release() override { delete this; } int SCI_METHOD Version() const override { return lvOriginal; } const char * SCI_METHOD PropertyNames() override { return osRust.PropertyNames(); } int SCI_METHOD PropertyType(const char *name) override { return osRust.PropertyType(name); } const char * SCI_METHOD DescribeProperty(const char *name) override { return osRust.DescribeProperty(name); } Sci_Position SCI_METHOD PropertySet(const char *key, const char *val) override; const char * SCI_METHOD DescribeWordListSets() override { return osRust.DescribeWordListSets(); } Sci_Position SCI_METHOD WordListSet(int n, const char *wl) override; void SCI_METHOD Lex(Sci_PositionU startPos, Sci_Position length, int initStyle, IDocument *pAccess) override; void SCI_METHOD Fold(Sci_PositionU startPos, Sci_Position length, int initStyle, IDocument *pAccess) override; void * SCI_METHOD PrivateCall(int, void *) override { return 0; } static ILexer *LexerFactoryRust() { return new LexerRust(); } }; Sci_Position SCI_METHOD LexerRust::PropertySet(const char *key, const char *val) { if (osRust.PropertySet(&options, key, val)) { return 0; } return -1; } Sci_Position SCI_METHOD LexerRust::WordListSet(int n, const char *wl) { Sci_Position firstModification = -1; if (n < NUM_RUST_KEYWORD_LISTS) { WordList *wordListN = &keywords[n]; WordList wlNew; wlNew.Set(wl); if (*wordListN != wlNew) { wordListN->Set(wl); firstModification = 0; } } return firstModification; } static bool IsWhitespace(int c) { return c == ' ' || c == '\t' || c == '\r' || c == '\n'; } /* This isn't quite right for Unicode identifiers */ static bool IsIdentifierStart(int ch) { return (IsASCII(ch) && (isalpha(ch) || ch == '_')) || !IsASCII(ch); } /* This isn't quite right for Unicode identifiers */ static bool IsIdentifierContinue(int ch) { return (IsASCII(ch) && (isalnum(ch) || ch == '_')) || !IsASCII(ch); } static void ScanWhitespace(Accessor& styler, Sci_Position& pos, Sci_Position max) { while (IsWhitespace(styler.SafeGetCharAt(pos, '\0')) && pos < max) { if (pos == styler.LineEnd(styler.GetLine(pos))) styler.SetLineState(styler.GetLine(pos), 0); pos++; } styler.ColourTo(pos-1, SCE_RUST_DEFAULT); } static void GrabString(char* s, Accessor& styler, Sci_Position start, Sci_Position len) { for (Sci_Position ii = 0; ii < len; ii++) s[ii] = styler[ii + start]; s[len] = '\0'; } static void ScanIdentifier(Accessor& styler, Sci_Position& pos, WordList *keywords) { Sci_Position start = pos; while (IsIdentifierContinue(styler.SafeGetCharAt(pos, '\0'))) pos++; if (styler.SafeGetCharAt(pos, '\0') == '!') { pos++; styler.ColourTo(pos - 1, SCE_RUST_MACRO); } else { char s[MAX_RUST_IDENT_CHARS + 1]; int len = pos - start; len = len > MAX_RUST_IDENT_CHARS ? MAX_RUST_IDENT_CHARS : len; GrabString(s, styler, start, len); bool keyword = false; for (int ii = 0; ii < NUM_RUST_KEYWORD_LISTS; ii++) { if (keywords[ii].InList(s)) { styler.ColourTo(pos - 1, SCE_RUST_WORD + ii); keyword = true; break; } } if (!keyword) { styler.ColourTo(pos - 1, SCE_RUST_IDENTIFIER); } } } /* Scans a sequence of digits, returning true if it found any. */ static bool ScanDigits(Accessor& styler, Sci_Position& pos, int base) { Sci_Position old_pos = pos; for (;;) { int c = styler.SafeGetCharAt(pos, '\0'); if (IsADigit(c, base) || c == '_') pos++; else break; } return old_pos != pos; } /* Scans an integer and floating point literals. */ static void ScanNumber(Accessor& styler, Sci_Position& pos) { int base = 10; int c = styler.SafeGetCharAt(pos, '\0'); int n = styler.SafeGetCharAt(pos + 1, '\0'); bool error = false; /* Scan the prefix, thus determining the base. * 10 is default if there's no prefix. */ if (c == '0' && n == 'x') { pos += 2; base = 16; } else if (c == '0' && n == 'b') { pos += 2; base = 2; } else if (c == '0' && n == 'o') { pos += 2; base = 8; } /* Scan initial digits. The literal is malformed if there are none. */ error |= !ScanDigits(styler, pos, base); /* See if there's an integer suffix. We mimic the Rust's lexer * and munch it even if there was an error above. */ c = styler.SafeGetCharAt(pos, '\0'); if (c == 'u' || c == 'i') { pos++; c = styler.SafeGetCharAt(pos, '\0'); n = styler.SafeGetCharAt(pos + 1, '\0'); if (c == '8' || c == 's') { pos++; } else if (c == '1' && n == '6') { pos += 2; } else if (c == '3' && n == '2') { pos += 2; } else if (c == '6' && n == '4') { pos += 2; } else { error = true; } /* See if it's a floating point literal. These literals have to be base 10. */ } else if (!error) { /* If there's a period, it's a floating point literal unless it's * followed by an identifier (meaning this is a method call, e.g. * `1.foo()`) or another period, in which case it's a range (e.g. 1..2) */ n = styler.SafeGetCharAt(pos + 1, '\0'); if (c == '.' && !(IsIdentifierStart(n) || n == '.')) { error |= base != 10; pos++; /* It's ok to have no digits after the period. */ ScanDigits(styler, pos, 10); } /* Look for the exponentiation. */ c = styler.SafeGetCharAt(pos, '\0'); if (c == 'e' || c == 'E') { error |= base != 10; pos++; c = styler.SafeGetCharAt(pos, '\0'); if (c == '-' || c == '+') pos++; /* It is invalid to have no digits in the exponent. */ error |= !ScanDigits(styler, pos, 10); } /* Scan the floating point suffix. */ c = styler.SafeGetCharAt(pos, '\0'); if (c == 'f') { error |= base != 10; pos++; c = styler.SafeGetCharAt(pos, '\0'); n = styler.SafeGetCharAt(pos + 1, '\0'); if (c == '3' && n == '2') { pos += 2; } else if (c == '6' && n == '4') { pos += 2; } else { error = true; } } } if (error) styler.ColourTo(pos - 1, SCE_RUST_LEXERROR); else styler.ColourTo(pos - 1, SCE_RUST_NUMBER); } static bool IsOneCharOperator(int c) { return c == ';' || c == ',' || c == '(' || c == ')' || c == '{' || c == '}' || c == '[' || c == ']' || c == '@' || c == '#' || c == '~' || c == '+' || c == '*' || c == '/' || c == '^' || c == '%' || c == '.' || c == ':' || c == '!' || c == '<' || c == '>' || c == '=' || c == '-' || c == '&' || c == '|' || c == '$' || c == '?'; } static bool IsTwoCharOperator(int c, int n) { return (c == '.' && n == '.') || (c == ':' && n == ':') || (c == '!' && n == '=') || (c == '<' && n == '<') || (c == '<' && n == '=') || (c == '>' && n == '>') || (c == '>' && n == '=') || (c == '=' && n == '=') || (c == '=' && n == '>') || (c == '-' && n == '>') || (c == '&' && n == '&') || (c == '|' && n == '|') || (c == '-' && n == '=') || (c == '&' && n == '=') || (c == '|' && n == '=') || (c == '+' && n == '=') || (c == '*' && n == '=') || (c == '/' && n == '=') || (c == '^' && n == '=') || (c == '%' && n == '='); } static bool IsThreeCharOperator(int c, int n, int n2) { return (c == '<' && n == '<' && n2 == '=') || (c == '>' && n == '>' && n2 == '='); } static bool IsValidCharacterEscape(int c) { return c == 'n' || c == 'r' || c == 't' || c == '\\' || c == '\'' || c == '"' || c == '0'; } static bool IsValidStringEscape(int c) { return IsValidCharacterEscape(c) || c == '\n' || c == '\r'; } static bool ScanNumericEscape(Accessor &styler, Sci_Position& pos, Sci_Position num_digits, bool stop_asap) { for (;;) { int c = styler.SafeGetCharAt(pos, '\0'); if (!IsADigit(c, 16)) break; num_digits--; pos++; if (num_digits == 0 && stop_asap) return true; } if (num_digits == 0) { return true; } else { return false; } } /* This is overly permissive for character literals in order to accept UTF-8 encoded * character literals. */ static void ScanCharacterLiteralOrLifetime(Accessor &styler, Sci_Position& pos, bool ascii_only) { pos++; int c = styler.SafeGetCharAt(pos, '\0'); int n = styler.SafeGetCharAt(pos + 1, '\0'); bool done = false; bool valid_lifetime = !ascii_only && IsIdentifierStart(c); bool valid_char = true; bool first = true; while (!done) { switch (c) { case '\\': done = true; if (IsValidCharacterEscape(n)) { pos += 2; } else if (n == 'x') { pos += 2; valid_char = ScanNumericEscape(styler, pos, 2, false); } else if (n == 'u' && !ascii_only) { pos += 2; if (styler.SafeGetCharAt(pos, '\0') != '{') { // old-style valid_char = ScanNumericEscape(styler, pos, 4, false); } else { int n_digits = 0; while (IsADigit(styler.SafeGetCharAt(++pos, '\0'), 16) && n_digits++ < 6) { } if (n_digits > 0 && styler.SafeGetCharAt(pos, '\0') == '}') pos++; else valid_char = false; } } else if (n == 'U' && !ascii_only) { pos += 2; valid_char = ScanNumericEscape(styler, pos, 8, false); } else { valid_char = false; } break; case '\'': valid_char = !first; done = true; break; case '\t': case '\n': case '\r': case '\0': valid_char = false; done = true; break; default: if (ascii_only && !IsASCII((char)c)) { done = true; valid_char = false; } else if (!IsIdentifierContinue(c) && !first) { done = true; } else { pos++; } break; } c = styler.SafeGetCharAt(pos, '\0'); n = styler.SafeGetCharAt(pos + 1, '\0'); first = false; } if (styler.SafeGetCharAt(pos, '\0') == '\'') { valid_lifetime = false; } else { valid_char = false; } if (valid_lifetime) { styler.ColourTo(pos - 1, SCE_RUST_LIFETIME); } else if (valid_char) { pos++; styler.ColourTo(pos - 1, ascii_only ? SCE_RUST_BYTECHARACTER : SCE_RUST_CHARACTER); } else { styler.ColourTo(pos - 1, SCE_RUST_LEXERROR); } } enum CommentState { UnknownComment, DocComment, NotDocComment }; /* * The rule for block-doc comments is as follows: /xxN and /x! (where x is an asterisk, N is a non-asterisk) start doc comments. * Otherwise it's a regular comment. */ static void ResumeBlockComment(Accessor &styler, Sci_Position& pos, Sci_Position max, CommentState state, int level) { int c = styler.SafeGetCharAt(pos, '\0'); bool maybe_doc_comment = false; if (c == '*') { int n = styler.SafeGetCharAt(pos + 1, '\0'); if (n != '*' && n != '/') { maybe_doc_comment = true; } } else if (c == '!') { maybe_doc_comment = true; } for (;;) { int n = styler.SafeGetCharAt(pos + 1, '\0'); if (pos == styler.LineEnd(styler.GetLine(pos))) styler.SetLineState(styler.GetLine(pos), level); if (c == '*') { pos++; if (n == '/') { pos++; level--; if (level == 0) { styler.SetLineState(styler.GetLine(pos), 0); if (state == DocComment || (state == UnknownComment && maybe_doc_comment)) styler.ColourTo(pos - 1, SCE_RUST_COMMENTBLOCKDOC); else styler.ColourTo(pos - 1, SCE_RUST_COMMENTBLOCK); break; } } } else if (c == '/') { pos++; if (n == '*') { pos++; level++; } } else { pos++; } if (pos >= max) { if (state == DocComment || (state == UnknownComment && maybe_doc_comment)) styler.ColourTo(pos - 1, SCE_RUST_COMMENTBLOCKDOC); else styler.ColourTo(pos - 1, SCE_RUST_COMMENTBLOCK); break; } c = styler.SafeGetCharAt(pos, '\0'); } } /* * The rule for line-doc comments is as follows... ///N and //! (where N is a non slash) start doc comments. * Otherwise it's a normal line comment. */ static void ResumeLineComment(Accessor &styler, Sci_Position& pos, Sci_Position max, CommentState state) { bool maybe_doc_comment = false; int c = styler.SafeGetCharAt(pos, '\0'); if (c == '/') { if (pos < max) { pos++; c = styler.SafeGetCharAt(pos, '\0'); if (c != '/') { maybe_doc_comment = true; } } } else if (c == '!') { maybe_doc_comment = true; } while (pos < max && c != '\n') { if (pos == styler.LineEnd(styler.GetLine(pos))) styler.SetLineState(styler.GetLine(pos), 0); pos++; c = styler.SafeGetCharAt(pos, '\0'); } if (state == DocComment || (state == UnknownComment && maybe_doc_comment)) styler.ColourTo(pos - 1, SCE_RUST_COMMENTLINEDOC); else styler.ColourTo(pos - 1, SCE_RUST_COMMENTLINE); } static void ScanComments(Accessor &styler, Sci_Position& pos, Sci_Position max) { pos++; int c = styler.SafeGetCharAt(pos, '\0'); pos++; if (c == '/') ResumeLineComment(styler, pos, max, UnknownComment); else if (c == '*') ResumeBlockComment(styler, pos, max, UnknownComment, 1); } static void ResumeString(Accessor &styler, Sci_Position& pos, Sci_Position max, bool ascii_only) { int c = styler.SafeGetCharAt(pos, '\0'); bool error = false; while (c != '"' && !error) { if (pos >= max) { error = true; break; } if (pos == styler.LineEnd(styler.GetLine(pos))) styler.SetLineState(styler.GetLine(pos), 0); if (c == '\\') { int n = styler.SafeGetCharAt(pos + 1, '\0'); if (IsValidStringEscape(n)) { pos += 2; } else if (n == 'x') { pos += 2; error = !ScanNumericEscape(styler, pos, 2, true); } else if (n == 'u' && !ascii_only) { pos += 2; if (styler.SafeGetCharAt(pos, '\0') != '{') { // old-style error = !ScanNumericEscape(styler, pos, 4, true); } else { int n_digits = 0; while (IsADigit(styler.SafeGetCharAt(++pos, '\0'), 16) && n_digits++ < 6) { } if (n_digits > 0 && styler.SafeGetCharAt(pos, '\0') == '}') pos++; else error = true; } } else if (n == 'U' && !ascii_only) { pos += 2; error = !ScanNumericEscape(styler, pos, 8, true); } else { pos += 1; error = true; } } else { if (ascii_only && !IsASCII((char)c)) error = true; else pos++; } c = styler.SafeGetCharAt(pos, '\0'); } if (!error) pos++; styler.ColourTo(pos - 1, ascii_only ? SCE_RUST_BYTESTRING : SCE_RUST_STRING); } static void ResumeRawString(Accessor &styler, Sci_Position& pos, Sci_Position max, int num_hashes, bool ascii_only) { for (;;) { if (pos == styler.LineEnd(styler.GetLine(pos))) styler.SetLineState(styler.GetLine(pos), num_hashes); int c = styler.SafeGetCharAt(pos, '\0'); if (c == '"') { pos++; int trailing_num_hashes = 0; while (styler.SafeGetCharAt(pos, '\0') == '#' && trailing_num_hashes < num_hashes) { trailing_num_hashes++; pos++; } if (trailing_num_hashes == num_hashes) { styler.SetLineState(styler.GetLine(pos), 0); break; } } else if (pos >= max) { break; } else { if (ascii_only && !IsASCII((char)c)) break; pos++; } } styler.ColourTo(pos - 1, ascii_only ? SCE_RUST_BYTESTRINGR : SCE_RUST_STRINGR); } static void ScanRawString(Accessor &styler, Sci_Position& pos, Sci_Position max, bool ascii_only) { pos++; int num_hashes = 0; while (styler.SafeGetCharAt(pos, '\0') == '#') { num_hashes++; pos++; } if (styler.SafeGetCharAt(pos, '\0') != '"') { styler.ColourTo(pos - 1, SCE_RUST_LEXERROR); } else { pos++; ResumeRawString(styler, pos, max, num_hashes, ascii_only); } } void SCI_METHOD LexerRust::Lex(Sci_PositionU startPos, Sci_Position length, int initStyle, IDocument *pAccess) { PropSetSimple props; Accessor styler(pAccess, &props); Sci_Position pos = startPos; Sci_Position max = pos + length; styler.StartAt(pos); styler.StartSegment(pos); if (initStyle == SCE_RUST_COMMENTBLOCK || initStyle == SCE_RUST_COMMENTBLOCKDOC) { ResumeBlockComment(styler, pos, max, initStyle == SCE_RUST_COMMENTBLOCKDOC ? DocComment : NotDocComment, styler.GetLineState(styler.GetLine(pos) - 1)); } else if (initStyle == SCE_RUST_COMMENTLINE || initStyle == SCE_RUST_COMMENTLINEDOC) { ResumeLineComment(styler, pos, max, initStyle == SCE_RUST_COMMENTLINEDOC ? DocComment : NotDocComment); } else if (initStyle == SCE_RUST_STRING) { ResumeString(styler, pos, max, false); } else if (initStyle == SCE_RUST_BYTESTRING) { ResumeString(styler, pos, max, true); } else if (initStyle == SCE_RUST_STRINGR) { ResumeRawString(styler, pos, max, styler.GetLineState(styler.GetLine(pos) - 1), false); } else if (initStyle == SCE_RUST_BYTESTRINGR) { ResumeRawString(styler, pos, max, styler.GetLineState(styler.GetLine(pos) - 1), true); } while (pos < max) { int c = styler.SafeGetCharAt(pos, '\0'); int n = styler.SafeGetCharAt(pos + 1, '\0'); int n2 = styler.SafeGetCharAt(pos + 2, '\0'); if (pos == 0 && c == '#' && n == '!' && n2 != '[') { pos += 2; ResumeLineComment(styler, pos, max, NotDocComment); } else if (IsWhitespace(c)) { ScanWhitespace(styler, pos, max); } else if (c == '/' && (n == '/' || n == '*')) { ScanComments(styler, pos, max); } else if (c == 'r' && (n == '#' || n == '"')) { ScanRawString(styler, pos, max, false); } else if (c == 'b' && n == 'r' && (n2 == '#' || n2 == '"')) { pos++; ScanRawString(styler, pos, max, true); } else if (c == 'b' && n == '"') { pos += 2; ResumeString(styler, pos, max, true); } else if (c == 'b' && n == '\'') { pos++; ScanCharacterLiteralOrLifetime(styler, pos, true); } else if (IsIdentifierStart(c)) { ScanIdentifier(styler, pos, keywords); } else if (IsADigit(c)) { ScanNumber(styler, pos); } else if (IsThreeCharOperator(c, n, n2)) { pos += 3; styler.ColourTo(pos - 1, SCE_RUST_OPERATOR); } else if (IsTwoCharOperator(c, n)) { pos += 2; styler.ColourTo(pos - 1, SCE_RUST_OPERATOR); } else if (IsOneCharOperator(c)) { pos++; styler.ColourTo(pos - 1, SCE_RUST_OPERATOR); } else if (c == '\'') { ScanCharacterLiteralOrLifetime(styler, pos, false); } else if (c == '"') { pos++; ResumeString(styler, pos, max, false); } else { pos++; styler.ColourTo(pos - 1, SCE_RUST_LEXERROR); } } styler.ColourTo(pos - 1, SCE_RUST_DEFAULT); styler.Flush(); } void SCI_METHOD LexerRust::Fold(Sci_PositionU startPos, Sci_Position length, int initStyle, IDocument *pAccess) { if (!options.fold) return; LexAccessor styler(pAccess); Sci_PositionU endPos = startPos + length; int visibleChars = 0; bool inLineComment = false; Sci_Position lineCurrent = styler.GetLine(startPos); int levelCurrent = SC_FOLDLEVELBASE; if (lineCurrent > 0) levelCurrent = styler.LevelAt(lineCurrent-1) >> 16; Sci_PositionU lineStartNext = styler.LineStart(lineCurrent+1); int levelMinCurrent = levelCurrent; int levelNext = levelCurrent; char chNext = styler[startPos]; int styleNext = styler.StyleAt(startPos); int style = initStyle; const bool userDefinedFoldMarkers = !options.foldExplicitStart.empty() && !options.foldExplicitEnd.empty(); for (Sci_PositionU i = startPos; i < endPos; i++) { char ch = chNext; chNext = styler.SafeGetCharAt(i + 1); int stylePrev = style; style = styleNext; styleNext = styler.StyleAt(i + 1); bool atEOL = i == (lineStartNext-1); if ((style == SCE_RUST_COMMENTLINE) || (style == SCE_RUST_COMMENTLINEDOC)) inLineComment = true; if (options.foldComment && options.foldCommentMultiline && IsStreamCommentStyle(style) && !inLineComment) { if (!IsStreamCommentStyle(stylePrev)) { levelNext++; } else if (!IsStreamCommentStyle(styleNext) && !atEOL) { // Comments don't end at end of line and the next character may be unstyled. levelNext--; } } if (options.foldComment && options.foldCommentExplicit && ((style == SCE_RUST_COMMENTLINE) || options.foldExplicitAnywhere)) { if (userDefinedFoldMarkers) { if (styler.Match(i, options.foldExplicitStart.c_str())) { levelNext++; } else if (styler.Match(i, options.foldExplicitEnd.c_str())) { levelNext--; } } else { if ((ch == '/') && (chNext == '/')) { char chNext2 = styler.SafeGetCharAt(i + 2); if (chNext2 == '{') { levelNext++; } else if (chNext2 == '}') { levelNext--; } } } } if (options.foldSyntaxBased && (style == SCE_RUST_OPERATOR)) { if (ch == '{') { // Measure the minimum before a '{' to allow // folding on "} else {" if (levelMinCurrent > levelNext) { levelMinCurrent = levelNext; } levelNext++; } else if (ch == '}') { levelNext--; } } if (!IsASpace(ch)) visibleChars++; if (atEOL || (i == endPos-1)) { int levelUse = levelCurrent; if (options.foldSyntaxBased && options.foldAtElse) { levelUse = levelMinCurrent; } int lev = levelUse | levelNext << 16; if (visibleChars == 0 && options.foldCompact) lev |= SC_FOLDLEVELWHITEFLAG; if (levelUse < levelNext) lev |= SC_FOLDLEVELHEADERFLAG; if (lev != styler.LevelAt(lineCurrent)) { styler.SetLevel(lineCurrent, lev); } lineCurrent++; lineStartNext = styler.LineStart(lineCurrent+1); levelCurrent = levelNext; levelMinCurrent = levelCurrent; if (atEOL && (i == static_cast(styler.Length()-1))) { // There is an empty line at end of file so give it same level and empty styler.SetLevel(lineCurrent, (levelCurrent | levelCurrent << 16) | SC_FOLDLEVELWHITEFLAG); } visibleChars = 0; inLineComment = false; } } } LexerModule lmRust(SCLEX_RUST, LexerRust::LexerFactoryRust, "rust", rustWordLists);