1 // Copyright 2014 PDFium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4 
5 // Original code copyright 2014 Foxit Software Inc. http://www.foxitsoftware.com
6 
7 #include "../../include/fpdfapi/fpdf_pageobj.h"
8 #include "../../include/fpdftext/fpdf_text.h"
9 #include "../../include/fpdfapi/fpdf_page.h"
10 class CPDF_TextStream
11 {
12 public:
13     CPDF_TextStream(CFX_WideTextBuf& buffer, FX_BOOL bUseLF, CFX_PtrArray* pObjArray);
~CPDF_TextStream()14     ~CPDF_TextStream() {}
15     FX_BOOL ProcessObject(const CPDF_TextObject* pObj, FX_BOOL bFirstLine);
16     CFX_WideTextBuf&	m_Buffer;
17     FX_BOOL				m_bUseLF;
18     CFX_PtrArray*		m_pObjArray;
19     const CPDF_TextObject*	m_pLastObj;
20 };
CPDF_TextStream(CFX_WideTextBuf & buffer,FX_BOOL bUseLF,CFX_PtrArray * pObjArray)21 CPDF_TextStream::CPDF_TextStream(CFX_WideTextBuf& buffer, FX_BOOL bUseLF, CFX_PtrArray* pObjArray) : m_Buffer(buffer)
22 {
23     m_pLastObj = NULL;
24     m_bUseLF = bUseLF;
25     m_pObjArray = pObjArray;
26 }
FPDFText_IsSameTextObject(const CPDF_TextObject * pTextObj1,const CPDF_TextObject * pTextObj2)27 FX_BOOL FPDFText_IsSameTextObject(const CPDF_TextObject* pTextObj1, const CPDF_TextObject* pTextObj2)
28 {
29     if (!pTextObj1 || !pTextObj2) {
30         return FALSE;
31     }
32     CFX_FloatRect rcPreObj(pTextObj2->m_Left, pTextObj2->m_Bottom, pTextObj2->m_Right, pTextObj2->m_Top);
33     CFX_FloatRect rcCurObj(pTextObj1->m_Left, pTextObj1->m_Bottom, pTextObj1->m_Right, pTextObj1->m_Top);
34     if (rcPreObj.IsEmpty() && rcCurObj.IsEmpty()) {
35         return TRUE;
36     }
37     if (!rcPreObj.IsEmpty() || !rcCurObj.IsEmpty()) {
38         rcPreObj.Intersect(rcCurObj);
39         if (rcPreObj.IsEmpty()) {
40             return FALSE;
41         }
42         if (FXSYS_fabs(rcPreObj.Width() - rcCurObj.Width()) > rcCurObj.Width() / 2) {
43             return FALSE;
44         }
45         if (pTextObj2->GetFontSize() != pTextObj1->GetFontSize()) {
46             return FALSE;
47         }
48     }
49     int nPreCount = pTextObj2->CountItems();
50     int nCurCount = pTextObj1->CountItems();
51     if (nPreCount != nCurCount) {
52         return FALSE;
53     }
54     for (int i = 0; i < nPreCount; i++) {
55         CPDF_TextObjectItem itemPer, itemCur;
56         pTextObj2->GetItemInfo(i, &itemPer);
57         pTextObj1->GetItemInfo(i, &itemCur);
58         if (itemCur.m_CharCode != itemPer.m_CharCode) {
59             return FALSE;
60         }
61     }
62     return TRUE;
63 }
GetCharWidth(FX_DWORD charCode,CPDF_Font * pFont)64 int GetCharWidth(FX_DWORD charCode, CPDF_Font* pFont)
65 {
66     if(charCode == -1) {
67         return 0;
68     }
69     int w = pFont->GetCharWidthF(charCode);
70     if(w == 0) {
71         CFX_ByteString str;
72         pFont->AppendChar(str, charCode);
73         w = pFont->GetStringWidth(str, 1);
74         if(w == 0) {
75             FX_RECT BBox;
76             pFont->GetCharBBox(charCode, BBox);
77             w = BBox.right - BBox.left;
78         }
79     }
80     return w;
81 }
FPDFText_ProcessInterObj(const CPDF_TextObject * pPrevObj,const CPDF_TextObject * pObj)82 int FPDFText_ProcessInterObj(const CPDF_TextObject* pPrevObj, const CPDF_TextObject* pObj)
83 {
84     if(FPDFText_IsSameTextObject(pPrevObj, pObj)) {
85         return -1;
86     }
87     CPDF_TextObjectItem item;
88     int nItem = pPrevObj->CountItems();
89     pPrevObj->GetItemInfo(nItem - 1, &item);
90     FX_WCHAR preChar = 0, curChar = 0;
91     CFX_WideString wstr = pPrevObj->GetFont()->UnicodeFromCharCode(item.m_CharCode);
92     if(wstr.GetLength()) {
93         preChar = wstr.GetAt(0);
94     }
95     FX_FLOAT last_pos = item.m_OriginX;
96     int nLastWidth = GetCharWidth(item.m_CharCode, pPrevObj->GetFont());
97     FX_FLOAT last_width = nLastWidth * pPrevObj->GetFontSize() / 1000;
98     last_width = FXSYS_fabs(last_width);
99     pObj->GetItemInfo(0, &item);
100     wstr = pObj->GetFont()->UnicodeFromCharCode(item.m_CharCode);
101     if(wstr.GetLength()) {
102         curChar = wstr.GetAt(0);
103     }
104     int nThisWidth = GetCharWidth(item.m_CharCode, pObj->GetFont());
105     FX_FLOAT this_width = nThisWidth * pObj->GetFontSize() / 1000;
106     this_width = FXSYS_fabs(this_width);
107     FX_FLOAT threshold = last_width > this_width ? last_width / 4 : this_width / 4;
108     CFX_AffineMatrix prev_matrix, prev_reverse;
109     pPrevObj->GetTextMatrix(&prev_matrix);
110     prev_reverse.SetReverse(prev_matrix);
111     FX_FLOAT x = pObj->GetPosX(), y = pObj->GetPosY();
112     prev_reverse.Transform(x, y);
113     if (FXSYS_fabs(y) > threshold * 2) {
114         return 2;
115     }
116     threshold = (FX_FLOAT)(nLastWidth > nThisWidth ? nLastWidth : nThisWidth);
117     threshold = threshold > 400 ? (threshold < 700 ? threshold / 4 :  threshold / 5) : (threshold / 2);
118     threshold *= nLastWidth > nThisWidth ? FXSYS_fabs(pPrevObj->GetFontSize()) : FXSYS_fabs(pObj->GetFontSize());
119     threshold /= 1000;
120     if (FXSYS_fabs(last_pos + last_width - x) > threshold && curChar != L' ' && preChar != L' ')
121         if(curChar != L' ' && preChar != L' ') {
122             if((x - last_pos - last_width) > threshold || (last_pos - x - last_width) > threshold) {
123                 return 1;
124             }
125             if(x < 0 && (last_pos - x - last_width) > threshold) {
126                 return 1;
127             }
128             if((x - last_pos - last_width) > this_width || (x - last_pos - this_width) > last_width ) {
129                 return 1;
130             }
131         }
132     if(last_pos + last_width > x + this_width && curChar == L' ') {
133         return 3;
134     }
135     return 0;
136 }
ProcessObject(const CPDF_TextObject * pObj,FX_BOOL bFirstLine)137 FX_BOOL CPDF_TextStream::ProcessObject(const CPDF_TextObject* pObj, FX_BOOL bFirstLine)
138 {
139     CPDF_Font* pFont = pObj->GetFont();
140     CFX_AffineMatrix matrix;
141     pObj->GetTextMatrix(&matrix);
142     int item_index = 0;
143     if (m_pLastObj) {
144         int result = FPDFText_ProcessInterObj(m_pLastObj, pObj);
145         if (result == 2) {
146             int len = m_Buffer.GetLength();
147             if (len && m_bUseLF && m_Buffer.GetBuffer()[len - 1] == L'-') {
148                 m_Buffer.Delete(len - 1, 1);
149                 if (m_pObjArray) {
150                     m_pObjArray->RemoveAt((len - 1) * 2, 2);
151                 }
152             } else {
153                 if (bFirstLine) {
154                     return TRUE;
155                 }
156                 if (m_bUseLF) {
157                     m_Buffer.AppendChar(L'\r');
158                     m_Buffer.AppendChar(L'\n');
159                     if (m_pObjArray) {
160                         for (int i = 0; i < 4; i ++) {
161                             m_pObjArray->Add(NULL);
162                         }
163                     }
164                 } else {
165                     m_Buffer.AppendChar(' ');
166                     if (m_pObjArray) {
167                         m_pObjArray->Add(NULL);
168                         m_pObjArray->Add(NULL);
169                     }
170                 }
171             }
172         } else if (result == 1) {
173             m_Buffer.AppendChar(L' ');
174             if (m_pObjArray) {
175                 m_pObjArray->Add(NULL);
176                 m_pObjArray->Add(NULL);
177             }
178         } else if (result == -1) {
179             m_pLastObj = pObj;
180             return FALSE;
181         } else if (result == 3) {
182             item_index = 1;
183         }
184     }
185     m_pLastObj = pObj;
186     int nItems = pObj->CountItems();
187     FX_FLOAT Ignorekerning = 0;
188     for(int i = 1; i < nItems - 1; i += 2) {
189         CPDF_TextObjectItem item;
190         pObj->GetItemInfo(i, &item);
191         if (item.m_CharCode == (FX_DWORD) - 1) {
192             if(i == 1) {
193                 Ignorekerning = item.m_OriginX;
194             } else if(Ignorekerning > item.m_OriginX) {
195                 Ignorekerning = item.m_OriginX;
196             }
197         } else {
198             Ignorekerning = 0;
199             break;
200         }
201     }
202     FX_FLOAT spacing = 0;
203     for (; item_index < nItems; item_index ++) {
204         CPDF_TextObjectItem item;
205         pObj->GetItemInfo(item_index, &item);
206         if (item.m_CharCode == (FX_DWORD) - 1) {
207             CFX_WideString wstr = m_Buffer.GetWideString();
208             if (wstr.IsEmpty() || wstr.GetAt(wstr.GetLength() - 1) == L' ') {
209                 continue;
210             }
211             FX_FLOAT fontsize_h = pObj->m_TextState.GetFontSizeH();
212             spacing = -fontsize_h * (item.m_OriginX - Ignorekerning) / 1000;
213             continue;
214         }
215         FX_FLOAT charSpace = pObj->m_TextState.GetObject()->m_CharSpace;
216         if(nItems > 3 && !spacing) {
217             charSpace = 0;
218         }
219         if((spacing || charSpace) && item_index > 0) {
220             int last_width = 0;
221             FX_FLOAT fontsize_h = pObj->m_TextState.GetFontSizeH();
222             FX_DWORD space_charcode = pFont->CharCodeFromUnicode(' ');
223             FX_FLOAT threshold = 0;
224             if (space_charcode != -1) {
225                 threshold = fontsize_h * pFont->GetCharWidthF(space_charcode) / 1000 ;
226             }
227             if(threshold > fontsize_h / 3) {
228                 threshold = 0;
229             } else {
230                 threshold /= 2;
231             }
232             if (threshold == 0) {
233                 threshold = fontsize_h;
234                 int this_width = FXSYS_abs(GetCharWidth(item.m_CharCode, pFont));
235                 threshold = this_width > last_width ? (FX_FLOAT)this_width : (FX_FLOAT)last_width;
236                 int nDivide = 6;
237                 if (threshold < 300) {
238                     nDivide = 2;
239                 } else if (threshold < 500) {
240                     nDivide = 4;
241                 } else if (threshold < 700) {
242                     nDivide = 5;
243                 }
244                 threshold = threshold / nDivide;
245                 threshold = fontsize_h * threshold / 1000;
246             }
247             if(charSpace > 0.001) {
248                 spacing += matrix.TransformDistance(charSpace);
249             } else if(charSpace < -0.001) {
250                 spacing -= matrix.TransformDistance(FXSYS_fabs(charSpace));
251             }
252             if (threshold && (spacing && spacing >= threshold) ) {
253                 m_Buffer.AppendChar(L' ');
254                 if (m_pObjArray) {
255                     m_pObjArray->Add(NULL);
256                     m_pObjArray->Add(NULL);
257                 }
258             }
259             if (item.m_CharCode == (FX_DWORD) - 1) {
260                 continue;
261             }
262             spacing = 0;
263         }
264         CFX_WideString unicode_str = pFont->UnicodeFromCharCode(item.m_CharCode);
265         if (unicode_str.IsEmpty()) {
266             m_Buffer.AppendChar((FX_WCHAR)item.m_CharCode);
267             if (m_pObjArray) {
268                 m_pObjArray->Add((void*)pObj);
269                 m_pObjArray->Add((void*)(FX_INTPTR)item_index);
270             }
271         } else {
272             m_Buffer << unicode_str;
273             if (m_pObjArray) {
274                 for (int i = 0; i < unicode_str.GetLength(); i ++) {
275                     m_pObjArray->Add((void*)pObj);
276                     m_pObjArray->Add((void*)(FX_INTPTR)item_index);
277                 }
278             }
279         }
280     }
281     return FALSE;
282 }
_PDF_GetTextStream_Unicode(CFX_WideTextBuf & buffer,CPDF_PageObjects * pPage,FX_BOOL bUseLF,CFX_PtrArray * pObjArray)283 void _PDF_GetTextStream_Unicode(CFX_WideTextBuf& buffer, CPDF_PageObjects* pPage, FX_BOOL bUseLF,
284                                 CFX_PtrArray* pObjArray)
285 {
286     CPDF_TextStream textstream(buffer, bUseLF, pObjArray);
287     FX_POSITION pos = pPage->GetFirstObjectPosition();
288     while (pos) {
289         CPDF_PageObject* pObject = pPage->GetNextObject(pos);
290         if (pObject == NULL) {
291             continue;
292         }
293         if (pObject->m_Type != PDFPAGE_TEXT) {
294             continue;
295         }
296         textstream.ProcessObject((CPDF_TextObject*)pObject, FALSE);
297     }
298 }
PDF_GetFirstTextLine_Unicode(CPDF_Document * pDoc,CPDF_Dictionary * pPage)299 CFX_WideString PDF_GetFirstTextLine_Unicode(CPDF_Document* pDoc, CPDF_Dictionary* pPage)
300 {
301     CFX_WideTextBuf buffer;
302     buffer.EstimateSize(0, 1024);
303     CPDF_Page page;
304     page.Load(pDoc, pPage);
305     CPDF_ParseOptions options;
306     options.m_bTextOnly = TRUE;
307     options.m_bSeparateForm = FALSE;
308     page.ParseContent(&options);
309     CPDF_TextStream textstream(buffer, FALSE, NULL);
310     FX_POSITION pos = page.GetFirstObjectPosition();
311     while (pos) {
312         CPDF_PageObject* pObject = page.GetNextObject(pos);
313         if (pObject->m_Type != PDFPAGE_TEXT) {
314             continue;
315         }
316         if (textstream.ProcessObject((CPDF_TextObject*)pObject, TRUE)) {
317             break;
318         }
319     }
320     return buffer.GetWideString();
321 }
322