Rewrite lower level details of extracting text from page

The current implementation of text extraction was difficult to
understand, duplicated logic that existed in other methods, and wasn't
clear about the units the inputs were in. It also didn't handle
control characters correctly.

The new implementation leans on the methods for converting indices
between the text buffer index and character list index spaces to avoid
duplication of code. It also makes it clear to the reader that inputs
are in the character list index space. Finally, it fixes issues being
seen in Chrome with respect of ranges being slightly off.

This CL also adds a test for extracting text that has control
characters.

BUG=pdfium:942,chromium:654578

Change-Id: Id9d1f360c2d7492c7b5a48d6c9ae29f530892742
Reviewed-on: https://pdfium-review.googlesource.com/20014
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: dsinclair <dsinclair@chromium.org>
Reviewed-by: Henrique Nakashima <hnakashima@chromium.org>

2 files changed, 38 insertions, 18 deletions

diff --git a/fpdfsdk/fpdftext.cpp b/fpdfsdk/fpdftext.cpp
index 5a2deb9a81..d9f7d572e9 100644
--- a/fpdfsdk/fpdftext.cpp
+++ b/fpdfsdk/fpdftext.cpp
@@ -179,25 +179,10 @@ FPDF_EXPORT int FPDF_CALLCONV FPDFText_GetText(FPDF_TEXTPAGE page,
     return 1;
   }
 
-  // char_* values are for a data structure that includes non-printing unicode
-  // characters, where the text_* values are from a data structure that doesn't
-  // include these characters, so translation is needed.
-  int text_start = textpage->TextIndexFromCharIndex(char_start);
-  if (text_start == -1)
-    return 0;
-
-  int char_last = char_start + char_count - 1;
-  int text_last = textpage->TextIndexFromCharIndex(char_last);
-  if (text_last == -1)
-    return 0;
-
-  int text_count = text_last - text_start + 1;
-  if (text_count < 1)
-    return 0;
+  WideString str = textpage->GetPageText(char_start, char_count);
 
-  WideString str = textpage->GetPageText(text_start, text_count);
-  if (str.GetLength() > static_cast<size_t>(text_count))
-    str = str.Left(static_cast<size_t>(text_count));
+  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.
diff --git a/fpdfsdk/fpdftext_embeddertest.cpp b/fpdfsdk/fpdftext_embeddertest.cpp
index 51216b9818..60654057b9 100644
--- a/fpdfsdk/fpdftext_embeddertest.cpp
+++ b/fpdfsdk/fpdftext_embeddertest.cpp
@@ -590,3 +590,38 @@ TEST_F(FPDFTextEmbeddertest, bug_782596) {
   FPDFText_ClosePage(textpage);
   UnloadPage(page);
 }
+
+TEST_F(FPDFTextEmbeddertest, ControlCharacters) {
+  EXPECT_TRUE(OpenDocument("control_characters.pdf"));
+  FPDF_PAGE page = LoadPage(0);
+  EXPECT_TRUE(page);
+
+  FPDF_TEXTPAGE textpage = FPDFText_LoadPage(page);
+  EXPECT_TRUE(textpage);
+
+  // Should not include the control characters in the output
+  static const char expected[] = "Hello, world!\r\nGoodbye, world!";
+  unsigned short fixed_buffer[128];
+  memset(fixed_buffer, 0xbd, sizeof(fixed_buffer));
+  int num_chars = FPDFText_GetText(textpage, 0, 128, fixed_buffer);
+
+  ASSERT_GE(num_chars, 0);
+  EXPECT_EQ(sizeof(expected), static_cast<size_t>(num_chars));
+  EXPECT_TRUE(check_unsigned_shorts(expected, fixed_buffer, sizeof(expected)));
+
+  // Attempting to get a chunk of text after the control characters
+  static const char expected_substring[] = "Goodbye, world!";
+  // Offset is the length of 'Hello, world!\r\n' + 2 control characters in the
+  // original stream
+  static const int offset = 17;
+  memset(fixed_buffer, 0xbd, sizeof(fixed_buffer));
+  num_chars = FPDFText_GetText(textpage, offset, 128, fixed_buffer);
+
+  ASSERT_GE(num_chars, 0);
+  EXPECT_EQ(sizeof(expected_substring), static_cast<size_t>(num_chars));
+  EXPECT_TRUE(check_unsigned_shorts(expected_substring, fixed_buffer,
+                                    sizeof(expected_substring)));
+
+  FPDFText_ClosePage(textpage);
+  UnloadPage(page);
+}