Simplify CPDF_TextPage::FindTextlineFlowOrientation().

core/fpdftext/fpdf_text_int.cpp

 // Copyright 2014 PDFium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 // Original code copyright 2014 Foxit Software Inc. http://www.foxitsoftware.com
 
 #include <algorithm>
 #include <cctype>
 #include <cwctype>
 #include <memory>
@@ skipping 125 matching lines @@
   }
   if (pDst) {
     FX_WCHAR n = wFind;
     while (n--) {
       *pDst++ = *pMap++;
     }
   }
   return (FX_STRSIZE)wFind;
 }
 
+float MaskPercentFilled(const std::vector<bool>& mask,
+                        int32_t start,
+                        int32_t end) {
+  if (start >= end)
+    return 0;
+  float count = std::count_if(mask.begin() + start, mask.begin() + end,
+                              [](bool r) { return r; });
+  return count / (end - start);
+}
+
 }  // namespace
 
 CPDF_TextPage::CPDF_TextPage(const CPDF_Page* pPage, FPDFText_Direction flags)
     : m_pPage(pPage),
       m_parserflag(flags),
       m_pPreTextObj(nullptr),
       m_bIsParsed(false),
       m_TextlineDir(TextOrientation::Unknown),
       m_CurlineRect(0, 0, 0, 0) {
   m_TextBuf.EstimateSize(0, 10240);
@@ skipping 514 matching lines @@
     segmentStatus = 2;
     m_Segments.Add(segment);
     segment.m_Start = 0;
     segment.m_nCount = 0;
   }
   return m_Segments.GetSize();
 }
 
 CPDF_TextPage::TextOrientation CPDF_TextPage::FindTextlineFlowOrientation()
     const {
-  const int32_t nPageWidth = static_cast<int32_t>(m_pPage->GetPageWidth());
-  const int32_t nPageHeight = static_cast<int32_t>(m_pPage->GetPageHeight());
-  std::vector<uint8_t> nHorizontalMask(nPageWidth);
-  std::vector<uint8_t> nVerticalMask(nPageHeight);
-  uint8_t* pDataH = nHorizontalMask.data();
-  uint8_t* pDataV = nVerticalMask.data();
-  int32_t index = 0;
-  FX_FLOAT fLineHeight = 0.0f;
   if (m_pPage->GetPageObjectList()->empty())
     return TextOrientation::Unknown;
 
-  for (auto& pPageObj : *m_pPage->GetPageObjectList()) {
-    if (!pPageObj || !pPageObj->IsText())
+  const int32_t nPageWidth = static_cast<int32_t>(m_pPage->GetPageWidth());
+  const int32_t nPageHeight = static_cast<int32_t>(m_pPage->GetPageHeight());
+  std::vector<bool> nHorizontalMask(nPageWidth);
+  std::vector<bool> nVerticalMask(nPageHeight);
+  FX_FLOAT fLineHeight = 0.0f;
+  for (const auto& pPageObj : *m_pPage->GetPageObjectList()) {
+    if (!pPageObj->IsText())
       continue;
 
-    int32_t minH =
-        (int32_t)pPageObj->m_Left < 0 ? 0 : (int32_t)pPageObj->m_Left;
-    int32_t maxH = (int32_t)pPageObj->m_Right > nPageWidth
-                       ? nPageWidth
-                       : (int32_t)pPageObj->m_Right;
-    int32_t minV =
-        (int32_t)pPageObj->m_Bottom < 0 ? 0 : (int32_t)pPageObj->m_Bottom;
-    int32_t maxV = (int32_t)pPageObj->m_Top > nPageHeight
-                       ? nPageHeight
-                       : (int32_t)pPageObj->m_Top;
+    int32_t minH = std::max(static_cast<int32_t>(pPageObj->m_Left), 0);
+    int32_t maxH =
+        std::min(static_cast<int32_t>(pPageObj->m_Right), nPageWidth);
+    int32_t minV = std::max(static_cast<int32_t>(pPageObj->m_Bottom), 0);
+    int32_t maxV = std::min(static_cast<int32_t>(pPageObj->m_Top), nPageHeight);
     if (minH >= maxH || minV >= maxV)
       continue;
 
-    FXSYS_memset(pDataH + minH, 1, maxH - minH);
-    FXSYS_memset(pDataV + minV, 1, maxV - minV);
+    for (int32_t i = minH; i < maxH; ++i)
+      nHorizontalMask[i] = true;
+    for (int32_t i = minV; i < maxV; ++i)
+      nVerticalMask[i] = true;
     if (fLineHeight <= 0.0f)
       fLineHeight = pPageObj->m_Top - pPageObj->m_Bottom;

[dsinclair, 2016/06/15 13:37:50]

What happens if we have different line heights on the page? This just happens to
pick the first one?

If we have an extra large header, and then really small following text won't
this value be wrong?

[Lei Zhang, 2016/06/15 23:46:33]

Everything here is heuristics based, so this could plausibly work, or one can
argue we need to take more samples and average them, or take the max. My goal
with this CL is just to refactor. Not fix up the heuristics.

[dsinclair, 2016/06/16 15:06:10]

Acknowledged.

   }
-  int32_t nStartH = 0;
-  int32_t nEndH = 0;
-  FX_FLOAT nSumH = 0.0f;
+  int32_t index = 0;
   for (index = 0; index < nPageWidth; index++) {
-    if (1 == nHorizontalMask[index])
+    if (nHorizontalMask[index])
       break;

[dsinclair, 2016/06/15 13:37:50]

Am I misreading this or, isn't index after this for() always going to be minH
since that's the first bit we set and after the next for() always nPageWidth -
maxH  as that's the max bit we set?

What do we need these for loops for?

[Lei Zhang, 2016/06/15 23:46:33]

Which |minH|? It's the smallest |minH| value that reaches line 711 in that
for-loop. I suppose we could have "minminH" to track that in the for loop above
and remove this one?

[dsinclair, 2016/06/16 15:06:10]

Crap, there is another level of for() loop on 698 that I missed. So, why not
move nStartH and nEndH into that for and just set it to the min/max seen instead
of doing these for() loops?

[Lei Zhang, 2016/06/16 22:47:03]

Done.

   }
-  nStartH = index;
+  const int32_t nStartH = index;
   for (index = nPageWidth; index > 0; index--) {
-    if (1 == nHorizontalMask[index - 1])
+    if (nHorizontalMask[index - 1])
       break;
   }
-  nEndH = index;
-  for (index = nStartH; index < nEndH; index++) {
-    nSumH += nHorizontalMask[index];
-  }
-  nSumH /= nEndH - nStartH;
-  int32_t nStartV = 0;
-  int32_t nEndV = 0;
-  FX_FLOAT nSumV = 0.0f;
+  const int32_t nEndH = index;

[dsinclair, 2016/06/15 13:37:50]

Shoudn't this be index - 1? Or, should the mask vectors be one element bigger?
Do we go from 0 - nPageWidth or from 0 - (nPageWidth - 1)?

[Lei Zhang, 2016/06/15 23:46:33]

Hmm, this probably helps prevent a divide by 0? It is off by one, but maybe that
does not matter that much for the heuristics.

   for (index = 0; index < nPageHeight; index++) {
-    if (1 == nVerticalMask[index])
+    if (nVerticalMask[index])
       break;
   }
-  nStartV = index;
+  const int32_t nStartV = index;
   for (index = nPageHeight; index > 0; index--) {
-    if (1 == nVerticalMask[index - 1])
+    if (nVerticalMask[index - 1])
       break;
   }
-  nEndV = index;
-  for (index = nStartV; index < nEndV; index++)
-    nSumV += nVerticalMask[index];
-  nSumV /= nEndV - nStartV;
-
-  if ((nEndV - nStartV) < (int32_t)(2 * fLineHeight))
+  const int32_t nEndV = index;
+  const int32_t nDoubleLineHeight = 2 * fLineHeight;
+  if ((nEndV - nStartV) < nDoubleLineHeight)
     return TextOrientation::Horizontal;
-  if ((nEndH - nStartH) < (int32_t)(2 * fLineHeight))
+  if ((nEndH - nStartH) < nDoubleLineHeight)
     return TextOrientation::Vertical;
 
+  const FX_FLOAT nSumH = MaskPercentFilled(nHorizontalMask, nStartH, nEndH);

[dsinclair, 2016/06/15 13:37:50]

Doesn't this always come out to 1? Didn't we set every bit between startH and
endH to true in the mask?

[Lei Zhang, 2016/06/15 23:46:33]

No, we didn't. There's gaps in between because there are multiple text objects.

[dsinclair, 2016/06/16 15:06:10]

Right, missed the outer for() above, though we always set the same bits in the
vector.

[Lei Zhang, 2016/06/16 22:47:03]

Not sure what you mean exactly. Are you still trying to say you think this
always comes out as 1? If so, I assure you that's not the case.

[dsinclair, 2016/06/16 23:39:05]

Sorry, was trying to say my comment was wrong, and you were right. I missed the
outer loop above which is what confused me.

   if (nSumH > 0.8f)
     return TextOrientation::Horizontal;
 
+  const FX_FLOAT nSumV = MaskPercentFilled(nVerticalMask, nStartV, nEndV);
   if (nSumH > nSumV)
     return TextOrientation::Horizontal;
   if (nSumH < nSumV)
     return TextOrientation::Vertical;
   return TextOrientation::Unknown;
 }
 
 void CPDF_TextPage::AppendGeneratedCharacter(FX_WCHAR unicode,
                                              const CFX_Matrix& formMatrix) {
   PAGECHAR_INFO generateChar;
@@ skipping 1568 matching lines @@
   return index < m_LinkArray.size() ? m_LinkArray[index].m_strUrl : L"";
 }
 
 std::vector<CFX_FloatRect> CPDF_LinkExtract::GetRects(size_t index) const {
   if (index >= m_LinkArray.size())
     return std::vector<CFX_FloatRect>();
 
   return m_pTextPage->GetRectArray(m_LinkArray[index].m_Start,
                                    m_LinkArray[index].m_Count);
 }