You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
375 lines
11 KiB
375 lines
11 KiB
// Scintilla source code edit control |
|
/** @file LexTCL.cxx |
|
** Lexer for TCL language. |
|
**/ |
|
// Copyright 1998-2001 by Andre Arpin <arpin@kingston.net> |
|
// The License.txt file describes the conditions under which this software may be distributed. |
|
|
|
#include <stdlib.h> |
|
#include <string.h> |
|
#include <stdio.h> |
|
#include <stdarg.h> |
|
#include <assert.h> |
|
#include <ctype.h> |
|
|
|
#include <string> |
|
#include <string_view> |
|
|
|
#include "ILexer.h" |
|
#include "Scintilla.h" |
|
#include "SciLexer.h" |
|
|
|
#include "WordList.h" |
|
#include "LexAccessor.h" |
|
#include "Accessor.h" |
|
#include "StyleContext.h" |
|
#include "CharacterSet.h" |
|
#include "LexerModule.h" |
|
|
|
using namespace Lexilla; |
|
|
|
// Extended to accept accented characters |
|
static inline bool IsAWordChar(int ch) { |
|
return ch >= 0x80 || |
|
(isalnum(ch) || ch == '_' || ch ==':' || ch=='.'); // : name space separator |
|
} |
|
|
|
static inline bool IsAWordStart(int ch) { |
|
return ch >= 0x80 || (ch ==':' || isalpha(ch) || ch == '_'); |
|
} |
|
|
|
static inline bool IsANumberChar(int ch) { |
|
// Not exactly following number definition (several dots are seen as OK, etc.) |
|
// but probably enough in most cases. |
|
return (ch < 0x80) && |
|
(IsADigit(ch, 0x10) || toupper(ch) == 'E' || |
|
ch == '.' || ch == '-' || ch == '+'); |
|
} |
|
|
|
static void ColouriseTCLDoc(Sci_PositionU startPos, Sci_Position length, int , WordList *keywordlists[], Accessor &styler) { |
|
#define isComment(s) (s==SCE_TCL_COMMENT || s==SCE_TCL_COMMENTLINE || s==SCE_TCL_COMMENT_BOX || s==SCE_TCL_BLOCK_COMMENT) |
|
const bool foldComment = styler.GetPropertyInt("fold.comment") != 0; |
|
const bool foldCompact = styler.GetPropertyInt("fold.compact", 1) != 0; |
|
bool commentLevel = false; |
|
bool subBrace = false; // substitution begin with a brace ${.....} |
|
enum tLineState {LS_DEFAULT, LS_OPEN_COMMENT, LS_OPEN_DOUBLE_QUOTE, LS_COMMENT_BOX, LS_MASK_STATE = 0xf, |
|
LS_COMMAND_EXPECTED = 16, LS_BRACE_ONLY = 32 |
|
} lineState = LS_DEFAULT; |
|
bool prevSlash = false; |
|
int currentLevel = 0; |
|
bool expected = 0; |
|
bool subParen = 0; |
|
|
|
Sci_Position currentLine = styler.GetLine(startPos); |
|
if (currentLine > 0) |
|
currentLine--; |
|
length += startPos - styler.LineStart(currentLine); |
|
// make sure lines overlap |
|
startPos = styler.LineStart(currentLine); |
|
|
|
WordList &keywords = *keywordlists[0]; |
|
WordList &keywords2 = *keywordlists[1]; |
|
WordList &keywords3 = *keywordlists[2]; |
|
WordList &keywords4 = *keywordlists[3]; |
|
WordList &keywords5 = *keywordlists[4]; |
|
WordList &keywords6 = *keywordlists[5]; |
|
WordList &keywords7 = *keywordlists[6]; |
|
WordList &keywords8 = *keywordlists[7]; |
|
WordList &keywords9 = *keywordlists[8]; |
|
|
|
if (currentLine > 0) { |
|
int ls = styler.GetLineState(currentLine - 1); |
|
lineState = tLineState(ls & LS_MASK_STATE); |
|
expected = LS_COMMAND_EXPECTED == tLineState(ls & LS_COMMAND_EXPECTED); |
|
subBrace = LS_BRACE_ONLY == tLineState(ls & LS_BRACE_ONLY); |
|
currentLevel = styler.LevelAt(currentLine - 1) >> 17; |
|
commentLevel = (styler.LevelAt(currentLine - 1) >> 16) & 1; |
|
} else |
|
styler.SetLevel(0, SC_FOLDLEVELBASE | SC_FOLDLEVELHEADERFLAG); |
|
bool visibleChars = false; |
|
|
|
int previousLevel = currentLevel; |
|
StyleContext sc(startPos, length, SCE_TCL_DEFAULT, styler); |
|
for (; ; sc.Forward()) { |
|
next: |
|
if (sc.ch=='\r' && sc.chNext == '\n') // only ignore \r on PC process on the mac |
|
continue; |
|
bool atEnd = !sc.More(); // make sure we coloured the last word |
|
if (lineState != LS_DEFAULT) { |
|
sc.SetState(SCE_TCL_DEFAULT); |
|
if (lineState == LS_OPEN_COMMENT) |
|
sc.SetState(SCE_TCL_COMMENTLINE); |
|
else if (lineState == LS_OPEN_DOUBLE_QUOTE) |
|
sc.SetState(SCE_TCL_IN_QUOTE); |
|
else if (lineState == LS_COMMENT_BOX && (sc.ch == '#' || (sc.ch == ' ' && sc.chNext=='#'))) |
|
sc.SetState(SCE_TCL_COMMENT_BOX); |
|
lineState = LS_DEFAULT; |
|
} |
|
if (subBrace) { // ${ overrides every thing even \ except } |
|
if (sc.ch == '}') { |
|
subBrace = false; |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
sc.ForwardSetState(SCE_TCL_DEFAULT); |
|
goto next; |
|
} else |
|
sc.SetState(SCE_TCL_SUB_BRACE); |
|
if (!sc.atLineEnd) |
|
continue; |
|
} else if (sc.state == SCE_TCL_DEFAULT || sc.state ==SCE_TCL_OPERATOR) { |
|
expected &= isspacechar(static_cast<unsigned char>(sc.ch)) || IsAWordStart(sc.ch) || sc.ch =='#'; |
|
} else if (sc.state == SCE_TCL_SUBSTITUTION) { |
|
switch (sc.ch) { |
|
case '(': |
|
subParen=true; |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
sc.ForwardSetState(SCE_TCL_SUBSTITUTION); |
|
continue; |
|
case ')': |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
subParen=false; |
|
continue; |
|
case '$': |
|
continue; |
|
case ',': |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
if (subParen) { |
|
sc.ForwardSetState(SCE_TCL_SUBSTITUTION); |
|
goto next; // Already forwarded so avoid loop's Forward() |
|
} |
|
continue; |
|
default : |
|
// maybe spaces should be allowed ??? |
|
if (!IsAWordChar(sc.ch)) { // probably the code is wrong |
|
sc.SetState(SCE_TCL_DEFAULT); |
|
subParen = 0; |
|
} |
|
break; |
|
} |
|
} else if (isComment(sc.state)) { |
|
} else if (!IsAWordChar(sc.ch)) { |
|
if ((sc.state == SCE_TCL_IDENTIFIER && expected) || sc.state == SCE_TCL_MODIFIER) { |
|
char w[100]; |
|
sc.GetCurrent(w, sizeof(w)); |
|
char *s=w; |
|
if (w[strlen(w)-1]=='\r') |
|
w[strlen(w)-1]=0; |
|
while (*s == ':') // ignore leading : like in ::set a 10 |
|
++s; |
|
bool quote = sc.state == SCE_TCL_IN_QUOTE; |
|
if (commentLevel || expected) { |
|
if (keywords.InList(s)) { |
|
sc.ChangeState(quote ? SCE_TCL_WORD_IN_QUOTE : SCE_TCL_WORD); |
|
} else if (keywords2.InList(s)) { |
|
sc.ChangeState(quote ? SCE_TCL_WORD_IN_QUOTE : SCE_TCL_WORD2); |
|
} else if (keywords3.InList(s)) { |
|
sc.ChangeState(quote ? SCE_TCL_WORD_IN_QUOTE : SCE_TCL_WORD3); |
|
} else if (keywords4.InList(s)) { |
|
sc.ChangeState(quote ? SCE_TCL_WORD_IN_QUOTE : SCE_TCL_WORD4); |
|
} else if (sc.GetRelative(-static_cast<Sci_Position>(strlen(s))-1) == '{' && |
|
keywords5.InList(s) && sc.ch == '}') { // {keyword} exactly no spaces |
|
sc.ChangeState(SCE_TCL_EXPAND); |
|
} |
|
if (keywords6.InList(s)) { |
|
sc.ChangeState(SCE_TCL_WORD5); |
|
} else if (keywords7.InList(s)) { |
|
sc.ChangeState(SCE_TCL_WORD6); |
|
} else if (keywords8.InList(s)) { |
|
sc.ChangeState(SCE_TCL_WORD7); |
|
} else if (keywords9.InList(s)) { |
|
sc.ChangeState(SCE_TCL_WORD8); |
|
} |
|
} |
|
expected = false; |
|
sc.SetState(quote ? SCE_TCL_IN_QUOTE : SCE_TCL_DEFAULT); |
|
} else if (sc.state == SCE_TCL_MODIFIER || sc.state == SCE_TCL_IDENTIFIER) { |
|
sc.SetState(SCE_TCL_DEFAULT); |
|
} |
|
} |
|
if (atEnd) |
|
break; |
|
if (sc.atLineEnd) { |
|
lineState = LS_DEFAULT; |
|
currentLine = styler.GetLine(sc.currentPos); |
|
if (foldComment && sc.state!=SCE_TCL_COMMENT && isComment(sc.state)) { |
|
if (currentLevel == 0) { |
|
++currentLevel; |
|
commentLevel = true; |
|
} |
|
} else { |
|
if (visibleChars && commentLevel) { |
|
--currentLevel; |
|
--previousLevel; |
|
commentLevel = false; |
|
} |
|
} |
|
int flag = 0; |
|
if (!visibleChars && foldCompact) |
|
flag = SC_FOLDLEVELWHITEFLAG; |
|
if (currentLevel > previousLevel) |
|
flag = SC_FOLDLEVELHEADERFLAG; |
|
styler.SetLevel(currentLine, flag + previousLevel + SC_FOLDLEVELBASE + (currentLevel << 17) + (commentLevel << 16)); |
|
|
|
// Update the line state, so it can be seen by next line |
|
if (sc.state == SCE_TCL_IN_QUOTE) { |
|
lineState = LS_OPEN_DOUBLE_QUOTE; |
|
} else { |
|
if (prevSlash) { |
|
if (isComment(sc.state)) |
|
lineState = LS_OPEN_COMMENT; |
|
} else if (sc.state == SCE_TCL_COMMENT_BOX) |
|
lineState = LS_COMMENT_BOX; |
|
} |
|
styler.SetLineState(currentLine, |
|
(subBrace ? LS_BRACE_ONLY : 0) | |
|
(expected ? LS_COMMAND_EXPECTED : 0) | lineState); |
|
if (lineState == LS_COMMENT_BOX) |
|
sc.ForwardSetState(SCE_TCL_COMMENT_BOX); |
|
else if (lineState == LS_OPEN_DOUBLE_QUOTE) |
|
sc.ForwardSetState(SCE_TCL_IN_QUOTE); |
|
else |
|
sc.ForwardSetState(SCE_TCL_DEFAULT); |
|
prevSlash = false; |
|
previousLevel = currentLevel; |
|
visibleChars = false; |
|
goto next; |
|
} |
|
|
|
if (prevSlash) { |
|
prevSlash = false; |
|
if (sc.ch == '#' && IsANumberChar(sc.chNext)) |
|
sc.ForwardSetState(SCE_TCL_NUMBER); |
|
continue; |
|
} |
|
prevSlash = sc.ch == '\\'; |
|
if (isComment(sc.state)) |
|
continue; |
|
if (sc.atLineStart) { |
|
visibleChars = false; |
|
if (sc.state!=SCE_TCL_IN_QUOTE && !isComment(sc.state)) |
|
{ |
|
sc.SetState(SCE_TCL_DEFAULT); |
|
expected = IsAWordStart(sc.ch)|| isspacechar(static_cast<unsigned char>(sc.ch)); |
|
} |
|
} |
|
|
|
switch (sc.state) { |
|
case SCE_TCL_NUMBER: |
|
if (!IsANumberChar(sc.ch)) |
|
sc.SetState(SCE_TCL_DEFAULT); |
|
break; |
|
case SCE_TCL_IN_QUOTE: |
|
if (sc.ch == '"') { |
|
sc.ForwardSetState(SCE_TCL_DEFAULT); |
|
visibleChars = true; // necessary if a " is the first and only character on a line |
|
goto next; |
|
} else if (sc.ch == '[' || sc.ch == ']' || sc.ch == '$') { |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
expected = sc.ch == '['; |
|
sc.ForwardSetState(SCE_TCL_IN_QUOTE); |
|
goto next; |
|
} |
|
continue; |
|
case SCE_TCL_OPERATOR: |
|
sc.SetState(SCE_TCL_DEFAULT); |
|
break; |
|
} |
|
|
|
if (sc.ch == '#') { |
|
if (visibleChars) { |
|
if (sc.state != SCE_TCL_IN_QUOTE && expected) |
|
sc.SetState(SCE_TCL_COMMENT); |
|
} else { |
|
sc.SetState(SCE_TCL_COMMENTLINE); |
|
if (sc.chNext == '~') |
|
sc.SetState(SCE_TCL_BLOCK_COMMENT); |
|
if (sc.atLineStart && (sc.chNext == '#' || sc.chNext == '-')) |
|
sc.SetState(SCE_TCL_COMMENT_BOX); |
|
} |
|
} |
|
|
|
if (!isspacechar(static_cast<unsigned char>(sc.ch))) { |
|
visibleChars = true; |
|
} |
|
|
|
if (sc.ch == '\\') { |
|
prevSlash = true; |
|
continue; |
|
} |
|
|
|
// Determine if a new state should be entered. |
|
if (sc.state == SCE_TCL_DEFAULT) { |
|
if (IsAWordStart(sc.ch)) { |
|
sc.SetState(SCE_TCL_IDENTIFIER); |
|
} else if (IsADigit(sc.ch) && !IsAWordChar(sc.chPrev)) { |
|
sc.SetState(SCE_TCL_NUMBER); |
|
} else { |
|
switch (sc.ch) { |
|
case '\"': |
|
sc.SetState(SCE_TCL_IN_QUOTE); |
|
break; |
|
case '{': |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
expected = true; |
|
++currentLevel; |
|
break; |
|
case '}': |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
expected = true; |
|
--currentLevel; |
|
break; |
|
case '[': |
|
expected = true; |
|
[[fallthrough]]; |
|
case ']': |
|
case '(': |
|
case ')': |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
break; |
|
case ';': |
|
expected = true; |
|
break; |
|
case '$': |
|
subParen = 0; |
|
if (sc.chNext != '{') { |
|
sc.SetState(SCE_TCL_SUBSTITUTION); |
|
} else { |
|
sc.SetState(SCE_TCL_OPERATOR); // $ |
|
sc.Forward(); // { |
|
sc.ForwardSetState(SCE_TCL_SUB_BRACE); |
|
subBrace = true; |
|
} |
|
break; |
|
case '#': |
|
if ((isspacechar(static_cast<unsigned char>(sc.chPrev))|| |
|
isoperator(static_cast<char>(sc.chPrev))) && IsADigit(sc.chNext,0x10)) |
|
sc.SetState(SCE_TCL_NUMBER); |
|
break; |
|
case '-': |
|
sc.SetState(IsADigit(sc.chNext)? SCE_TCL_NUMBER: SCE_TCL_MODIFIER); |
|
break; |
|
default: |
|
if (isoperator(static_cast<char>(sc.ch))) { |
|
sc.SetState(SCE_TCL_OPERATOR); |
|
} |
|
} |
|
} |
|
} |
|
} |
|
sc.Complete(); |
|
} |
|
|
|
static const char *const tclWordListDesc[] = { |
|
"TCL Keywords", |
|
"TK Keywords", |
|
"iTCL Keywords", |
|
"tkCommands", |
|
"expand", |
|
"user1", |
|
"user2", |
|
"user3", |
|
"user4", |
|
0 |
|
}; |
|
|
|
// this code supports folding in the colourizer |
|
extern const LexerModule lmTCL(SCLEX_TCL, ColouriseTCLDoc, "tcl", 0, tclWordListDesc);
|
|
|