svx: import processed PDF text

Some PDFs don't include spaces in the text.
Instead, they rely on the explicit positioning
of each character to render visually separated words.
Latex seems to be prone to this approach, though not
exclusively.

Luckily, PDFium does process text and inserts
"generated" spaces where necessary, which is what
we retrieve and use as the text string while importing.

Change-Id: Ic21fe6c8416ecaba66f06b6260f1d6b040ff12af

2 files changed, 149 insertions, 0 deletions

diff --git a/external/pdfium/0012-svx-import-processed-PDF-text.patch.2 b/external/pdfium/0012-svx-import-processed-PDF-text.patch.2
new file mode 100644
index 000000000000..cae9ec808aba
--- /dev/null
+++ b/external/pdfium/0012-svx-import-processed-PDF-text.patch.2
@@ -0,0 +1,148 @@
+From 7e8ecec81f102993e3fe73256415dcf049c09e29 Mon Sep 17 00:00:00 2001
+From: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
+Date: Tue, 5 Jun 2018 11:35:39 +0200
+Subject: [PATCH 12/14] svx: import processed PDF text
+
+---
+ pdfium/core/fpdftext/cpdf_textpage.cpp | 29 ++++++++++++++++++++++++
+ pdfium/core/fpdftext/cpdf_textpage.h   |  2 ++
+ pdfium/fpdfsdk/fpdf_editpage.cpp       | 41 ++++++++++++++++++++++++++++++++++
+ pdfium/public/fpdf_edit.h              | 13 +++++++++++
+ 4 files changed, 85 insertions(+)
+
+diff --git a/pdfium/core/fpdftext/cpdf_textpage.cpp b/pdfium/core/fpdftext/cpdf_textpage.cpp
+index 5690698..4d7c48a 100644
+--- a/pdfium/core/fpdftext/cpdf_textpage.cpp
++++ b/pdfium/core/fpdftext/cpdf_textpage.cpp
+@@ -1464,3 +1464,32 @@ Optional<PAGECHAR_INFO> CPDF_TextPage::GenerateCharInfo(wchar_t unicode) {
+                                  info.m_Origin.x, info.m_Origin.y);
+   return info;
+ }
++
++WideString CPDF_TextPage::GetTextObjectText(CPDF_TextObject* pTextObj)
++{
++  if (!m_bIsParsed)
++    return WideString();
++
++  float posy = 0;
++  bool IsContainPreChar = false;
++  bool IsAddLineFeed = false;
++  WideString strText;
++  for (const auto& charinfo : m_CharList) {
++    if (charinfo.m_pTextObj == pTextObj) {
++      IsContainPreChar = true;
++      IsAddLineFeed = false;
++      if (charinfo.m_Unicode)
++        strText += charinfo.m_Unicode;
++    } else if (charinfo.m_Unicode == 32) {
++      if (IsContainPreChar && charinfo.m_Unicode) {
++        strText += charinfo.m_Unicode;
++        IsContainPreChar = false;
++        IsAddLineFeed = false;
++      }
++    } else {
++      IsContainPreChar = false;
++      IsAddLineFeed = true;
++    }
++  }
++  return strText;
++}
+diff --git a/pdfium/core/fpdftext/cpdf_textpage.h b/pdfium/core/fpdftext/cpdf_textpage.h
+index 43a0312..7d5d5ec 100644
+--- a/pdfium/core/fpdftext/cpdf_textpage.h
++++ b/pdfium/core/fpdftext/cpdf_textpage.h
+@@ -105,6 +105,8 @@ class CPDF_TextPage {
+   WideString GetPageText(int start, int count) const;
+   WideString GetAllPageText() const { return GetPageText(0, CountChars()); }
+ 
++  WideString GetTextObjectText(CPDF_TextObject* pTextObj);
++
+   int CountRects(int start, int nCount);
+   bool GetRect(int rectIndex, CFX_FloatRect* pRect) const;
+ 
+diff --git a/pdfium/fpdfsdk/fpdf_editpage.cpp b/pdfium/fpdfsdk/fpdf_editpage.cpp
+index f4a1688..f34d3b5 100644
+--- a/pdfium/fpdfsdk/fpdf_editpage.cpp
++++ b/pdfium/fpdfsdk/fpdf_editpage.cpp
+@@ -27,6 +27,7 @@
+ #include "core/fpdfapi/parser/cpdf_string.h"
+ #include "core/fpdfdoc/cpdf_annot.h"
+ #include "core/fpdfdoc/cpdf_annotlist.h"
++#include "core/fpdftext/cpdf_textpage.h"
+ #include "fpdfsdk/cpdfsdk_helpers.h"
+ #include "public/fpdf_formfill.h"
+ #include "third_party/base/logging.h"
+@@ -732,6 +733,46 @@ FPDF_EXPORT int FPDF_CALLCONV FPDFTextObj_GetText(FPDF_PAGEOBJECT text_object,
+   return ret_count;
+ }
+ 
++FPDF_EXPORT int FPDF_CALLCONV
++FPDFTextObj_GetTextProcessed(FPDF_PAGEOBJECT text_object,
++                             FPDF_TEXTPAGE page,
++                             int char_start,
++                             int char_count,
++                             unsigned short* result)
++{
++  if (!page || !text_object || char_start < 0 || char_count < 0 || !result)
++    return 0;
++
++  CPDF_TextObject* pTxtObj = CPDFTextObjectFromFPDFPageObject(text_object);
++  CPDF_TextPage* textpage = CPDFTextPageFromFPDFTextPage(page);
++  int char_available = textpage->CountChars() - char_start;
++  if (char_available <= 0)
++    return 0;
++
++  char_count = std::min(char_count, char_available);
++  if (char_count == 0) {
++    // Writing out "", which has a character count of 1 due to the NUL.
++    *result = '\0';
++    return 1;
++  }
++
++  WideString str = textpage->GetTextObjectText(pTxtObj);
++
++  if (str.GetLength() > static_cast<size_t>(char_count))
++    str = str.Left(static_cast<size_t>(char_count));
++
++  // UFT16LE_Encode doesn't handle surrogate pairs properly, so it is expected
++  // the number of items to stay the same.
++  ByteString byte_str = str.UTF16LE_Encode();
++  size_t byte_str_len = byte_str.GetLength();
++  constexpr size_t kBytesPerCharacter = sizeof(unsigned short);
++  int ret_count = byte_str_len / kBytesPerCharacter;
++
++  ASSERT(ret_count <= char_count + 1);  // +1 to account for the NUL terminator.
++  memcpy(result, byte_str.GetBuffer(byte_str_len), byte_str_len);
++  return ret_count;
++}
++
+ FPDF_EXPORT FPDF_BOOL FPDF_CALLCONV
+ FPDFTextObj_GetColor(FPDF_PAGEOBJECT text_object,
+                      unsigned int* R,
+diff --git a/pdfium/public/fpdf_edit.h b/pdfium/public/fpdf_edit.h
+index f249e64..e14b2a5 100644
+--- a/pdfium/public/fpdf_edit.h
++++ b/pdfium/public/fpdf_edit.h
+@@ -1065,6 +1065,19 @@ FPDFTextObj_GetText(FPDF_PAGEOBJECT text_object,
+                     int char_count,
+                     unsigned short* result);
+ 
++// Get the processed text of a text object.
++//
++// text_object - Handle of text object returned by FPDFPageObj_NewTextObj
++//               or FPDFPageObj_NewTextObjEx.
++// Return Value:
++// The number of characters (not bytes) written in result.
++FPDF_EXPORT int FPDF_CALLCONV
++FPDFTextObj_GetTextProcessed(FPDF_PAGEOBJECT text_object,
++                             FPDF_TEXTPAGE page,
++                             int char_start,
++                             int char_count,
++                             unsigned short* result);
++
+ // Get the stroke RGBA of a text. Range of values: 0 - 255.
+ //
+ // path   - the handle to the path object.
+-- 
+2.16.3
+
diff --git a/external/pdfium/UnpackedTarball_pdfium.mk b/external/pdfium/UnpackedTarball_pdfium.mk
index 6880ac0b670b..5525e9ddf65d 100644
--- a/external/pdfium/UnpackedTarball_pdfium.mk
+++ b/external/pdfium/UnpackedTarball_pdfium.mk
@@ -25,6 +25,7 @@ pdfium_patches += 0008-svx-correct-the-positioning-of-PDF-Paths-and-the-str.patc
 pdfium_patches += 0009-svx-support-color-text-for-imported-PDFs.patch.2
 pdfium_patches += 0010-svx-support-importing-forms-from-PDFs.patch.2
 pdfium_patches += 0011-svx-correctly-possition-form-objects-from-PDF.patch.2
+pdfium_patches += 0012-svx-import-processed-PDF-text.patch.2
 
 $(eval $(call gb_UnpackedTarball_UnpackedTarball,pdfium))