| /* | |
| * RichEdit - string operations | |
| * | |
| * Copyright 2004 by Krzysztof Foltman | |
| * | |
| * This library is free software; you can redistribute it and/or | |
| * modify it under the terms of the GNU Lesser General Public | |
| * License as published by the Free Software Foundation; either | |
| * version 2.1 of the License, or (at your option) any later version. | |
| * | |
| * This library is distributed in the hope that it will be useful, | |
| * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| * Lesser General Public License for more details. | |
| * | |
| * You should have received a copy of the GNU Lesser General Public | |
| * License along with this library; if not, write to the Free Software | |
| * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
| */ | |
| #include "editor.h" | |
| WINE_DEFAULT_DEBUG_CHANNEL(richedit); | |
| int ME_GetOptimalBuffer(int nLen) | |
| { | |
| return ((2*nLen+1)+128)&~63; | |
| } | |
| ME_String *ME_MakeString(LPCWSTR szText) | |
| { | |
| ME_String *s = ALLOC_OBJ(ME_String); | |
| s->nLen = lstrlenW(szText); | |
| s->nBuffer = ME_GetOptimalBuffer(s->nLen+1); | |
| s->szData = ALLOC_N_OBJ(WCHAR, s->nBuffer); | |
| lstrcpyW(s->szData, szText); | |
| return s; | |
| } | |
| ME_String *ME_MakeStringN(LPCWSTR szText, int nMaxChars) | |
| { | |
| ME_String *s = ALLOC_OBJ(ME_String); | |
| int i; | |
| for (i=0; i<nMaxChars && szText[i]; i++) | |
| ; | |
| s->nLen = i; | |
| s->nBuffer = ME_GetOptimalBuffer(s->nLen+1); | |
| s->szData = ALLOC_N_OBJ(WCHAR, s->nBuffer); | |
| lstrcpynW(s->szData, szText, s->nLen+1); | |
| return s; | |
| } | |
| ME_String *ME_StrDup(ME_String *s) | |
| { | |
| return ME_MakeStringN(s->szData, s->nLen); | |
| } | |
| void ME_DestroyString(ME_String *s) | |
| { | |
| FREE_OBJ(s->szData); | |
| FREE_OBJ(s); | |
| } | |
| void ME_AppendString(ME_String *s1, ME_String *s2) | |
| { | |
| if (s1->nLen+s2->nLen+1 <= s1->nBuffer) { | |
| lstrcpyW(s1->szData+s1->nLen, s2->szData); | |
| s1->nLen += s2->nLen; | |
| } | |
| else | |
| { | |
| WCHAR *buf; | |
| s1->nBuffer = ME_GetOptimalBuffer(s1->nLen+s2->nLen+1); | |
| buf = ALLOC_N_OBJ(WCHAR, s1->nBuffer); | |
| lstrcpyW(buf, s1->szData); | |
| lstrcpyW(buf+s1->nLen, s2->szData); | |
| FREE_OBJ(s1->szData); | |
| s1->szData = buf; | |
| s1->nLen += s2->nLen; | |
| } | |
| } | |
| ME_String *ME_ConcatString(ME_String *s1, ME_String *s2) | |
| { | |
| ME_String *s = ALLOC_OBJ(ME_String); | |
| s->nLen = s1->nLen+s2->nLen; | |
| s->nBuffer = ME_GetOptimalBuffer(s1->nLen+s2->nLen+1); | |
| s->szData = ALLOC_N_OBJ(WCHAR, s->nBuffer); | |
| lstrcpyW(s->szData, s1->szData); | |
| lstrcpyW(s->szData+s1->nLen, s2->szData); | |
| return s; | |
| } | |
| ME_String *ME_VSplitString(ME_String *orig, int charidx) | |
| { | |
| ME_String *s; | |
| /*if (charidx<0) charidx = 0; | |
| if (charidx>orig->nLen) charidx = orig->nLen; | |
| */ | |
| assert(charidx>=0); | |
| assert(charidx<=orig->nLen); | |
| s = ME_MakeString(orig->szData+charidx); | |
| orig->nLen = charidx; | |
| orig->szData[charidx] = L'\0'; | |
| return s; | |
| } | |
| int ME_IsWhitespaces(ME_String *s) | |
| { | |
| /* FIXME multibyte */ | |
| WCHAR *pos = s->szData; | |
| while(ME_IsWSpace(*pos++)) | |
| ; | |
| pos--; | |
| if (*pos) | |
| return 0; | |
| else | |
| return 1; | |
| } | |
| int ME_IsSplitable(ME_String *s) | |
| { | |
| /* FIXME multibyte */ | |
| WCHAR *pos = s->szData; | |
| WCHAR ch; | |
| while(ME_IsWSpace(*pos++)) | |
| ; | |
| pos--; | |
| while((ch = *pos++) != 0) | |
| { | |
| if (ME_IsWSpace(*pos++)) | |
| return 1; | |
| } | |
| return 0; | |
| } | |
| /* FIXME multibyte */ | |
| /* | |
| int ME_CalcSkipChars(ME_String *s) | |
| { | |
| int cnt = 0; | |
| while(cnt < s->nLen && s->szData[s->nLen-1-cnt]==' ') | |
| cnt++; | |
| return cnt; | |
| } | |
| */ | |
| int ME_StrLen(ME_String *s) { | |
| return s->nLen; | |
| } | |
| int ME_StrVLen(ME_String *s) { | |
| return s->nLen; | |
| } | |
| /* FIXME we use widechars, not multibytes, inside, no need for complex logic anymore */ | |
| int ME_StrRelPos(ME_String *s, int nVChar, int *pRelChars) | |
| { | |
| TRACE("%s,%d,&%d\n", debugstr_w(s->szData), nVChar, *pRelChars); | |
| assert(*pRelChars); | |
| if (!*pRelChars) return nVChar; | |
| if (*pRelChars>0) | |
| { | |
| while(nVChar<s->nLen && *pRelChars>0) | |
| { | |
| nVChar++; | |
| (*pRelChars)--; | |
| } | |
| return nVChar; | |
| } | |
| while(nVChar>0 && *pRelChars<0) | |
| { | |
| nVChar--; | |
| (*pRelChars)++; | |
| } | |
| return nVChar; | |
| } | |
| int ME_StrRelPos2(ME_String *s, int nVChar, int nRelChars) | |
| { | |
| return ME_StrRelPos(s, nVChar, &nRelChars); | |
| } | |
| int ME_VPosToPos(ME_String *s, int nVPos) | |
| { | |
| return nVPos; | |
| /* | |
| int i = 0, len = 0; | |
| if (!nVPos) | |
| return 0; | |
| while (i < s->nLen) | |
| { | |
| if (i == nVPos) | |
| return len; | |
| if (s->szData[i]=='\\') i++; | |
| i++; | |
| len++; | |
| } | |
| return len; | |
| */ | |
| } | |
| int ME_PosToVPos(ME_String *s, int nPos) | |
| { | |
| if (!nPos) | |
| return 0; | |
| return ME_StrRelPos2(s, 0, nPos); | |
| } | |
| void ME_StrDeleteV(ME_String *s, int nVChar, int nChars) | |
| { | |
| int end_ofs; | |
| assert(nVChar >=0 && nVChar <= s->nLen); | |
| assert(nChars >= 0); | |
| assert(nVChar+nChars <= s->nLen); | |
| end_ofs = ME_StrRelPos2(s, nVChar, nChars); | |
| assert(end_ofs <= s->nLen); | |
| memmove(s->szData+nVChar, s->szData+end_ofs, 2*(s->nLen+1-end_ofs)); | |
| s->nLen -= (end_ofs - nVChar); | |
| } | |
| int ME_GetCharFwd(ME_String *s, int nPos) | |
| { | |
| int nVPos = 0; | |
| assert(nPos < ME_StrLen(s)); | |
| if (nPos) | |
| nVPos = ME_StrRelPos2(s, nVPos, nPos); | |
| if (nVPos < s->nLen) | |
| return s->szData[nVPos]; | |
| return -1; | |
| } | |
| int ME_GetCharBack(ME_String *s, int nPos) | |
| { | |
| int nVPos = ME_StrVLen(s); | |
| assert(nPos < ME_StrLen(s)); | |
| if (nPos) | |
| nVPos = ME_StrRelPos2(s, nVPos, -nPos); | |
| if (nVPos < s->nLen) | |
| return s->szData[nVPos]; | |
| return -1; | |
| } | |
| int ME_FindNonWhitespaceV(ME_String *s, int nVChar) { | |
| int i; | |
| for (i = nVChar; isspace(s->szData[i]) && i<s->nLen; i++) | |
| ; | |
| return i; | |
| } | |
| /* note: returns offset of the first trailing whitespace */ | |
| int ME_ReverseFindNonWhitespaceV(ME_String *s, int nVChar) { | |
| int i; | |
| for (i = nVChar; i>0 && isspace(s->szData[i-1]); i--) | |
| ; | |
| return i; | |
| } | |
| /* note: returns offset of the first trailing nonwhitespace */ | |
| int ME_ReverseFindWhitespaceV(ME_String *s, int nVChar) { | |
| int i; | |
| for (i = nVChar; i>0 && !isspace(s->szData[i-1]); i--) | |
| ; | |
| return i; | |
| } | |
| LPWSTR ME_ToUnicode(HWND hWnd, LPVOID psz) | |
| { | |
| if (IsWindowUnicode(hWnd)) | |
| return (LPWSTR)psz; | |
| else { | |
| WCHAR *tmp; | |
| int nChars = MultiByteToWideChar(CP_ACP, 0, (char *)psz, -1, NULL, 0); | |
| if((tmp = ALLOC_N_OBJ(WCHAR, nChars)) != NULL) | |
| MultiByteToWideChar(CP_ACP, 0, (char *)psz, -1, tmp, nChars); | |
| return tmp; | |
| } | |
| } | |
| void ME_EndToUnicode(HWND hWnd, LPVOID psz) | |
| { | |
| if (IsWindowUnicode(hWnd)) | |
| FREE_OBJ(psz); | |
| } | |
| LPSTR ME_ToAnsi(HWND hWnd, LPVOID psz) | |
| { | |
| if (!IsWindowUnicode(hWnd)) | |
| return (LPSTR)psz; | |
| else { | |
| char *tmp; | |
| int nChars = WideCharToMultiByte(CP_ACP, 0, (WCHAR *)psz, -1, NULL, 0, NULL, NULL); | |
| if((tmp = ALLOC_N_OBJ(char, nChars)) != NULL) | |
| WideCharToMultiByte(CP_ACP, 0, (WCHAR *)psz, -1, tmp, nChars, NULL, NULL); | |
| return tmp; | |
| } | |
| } | |
| void ME_EndToAnsi(HWND hWnd, LPVOID psz) | |
| { | |
| if (!IsWindowUnicode(hWnd)) | |
| FREE_OBJ(psz); | |
| } |