For loop is running un-nacessrily after match.

pdf/pdfium/pdfium_engine.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "pdf/pdfium/pdfium_engine.h"
 
 #include <math.h>
 
 #include "base/json/json_writer.h"
 #include "base/logging.h"
@@ skipping 2703 matching lines @@
 PDFiumEngine::SelectionChangeInvalidator::~SelectionChangeInvalidator() {
   // Offset the old selections if the document scrolled since we recorded them.
   pp::Point offset = previous_origin_ - engine_->GetVisibleRect().point();
   for (size_t i = 0; i < old_selections_.size(); ++i)
     old_selections_[i].Offset(offset);
 
   std::vector<pp::Rect> new_selections;
   GetVisibleSelectionsScreenRects(&new_selections);
   for (size_t i = 0; i < new_selections.size(); ++i) {
     for (size_t j = 0; j < old_selections_.size(); ++j) {
-      if (new_selections[i] == old_selections_[j]) {
+      if (!old_selections_[j].IsEmpty() &&
+          new_selections[i] == old_selections_[j]) {
         // Rectangle was selected before and after, so no need to invalidate it.
         // Mark the rectangles by setting them to empty.
         new_selections[i] = old_selections_[j] = pp::Rect();
+        break;

[raymes, 2014/09/29 01:09:51]

I don't think we can early return from this loop (though I could be wrong). To
me it appears that this loop is trying to find which selections haven't changed
so that we don't need to invalidate those regions on the screen. AFAICT this
requires searching through all of them.

[Deepak, 2014/09/29 02:41:28]

@raymes, When below condition 
new_selections[i] == old_selections_[j]
matchs , Then we are setting both new_selection[i] and old+selection[j] as empty
rect, After that if we run inner loop then 
if old_selection is empty rect then only condition will match as new_selection
is already set empty rect.
and what is the point of checking empty rect's match in condition and then
assigning them empty rect again?

As their is no point of checking condition if old_selection is empty.
so this changes will avoid un-neccessary loop running.

[raymes, 2014/09/29 03:08:16]

Oh I see! Is it ever possible for old_selections_ to contain duplicates? I guess
not but I haven't checked.

[Deepak, 2014/09/29 03:19:52]

I think old selection does not contain duplicates.
And even it contains duplicates it will naver match as we have alreay set
new_selection[i] as empty rect :).

Little example:

new_selection: [A,C,D]
old_selection: [A,C,Z,A]

so in the above loop when new_selection[0] matches with old_selection[0]
then it is like
new_selection: [" ",C,D]
old_selection: [" ",C,Z,A]
then even old_selection have duplicate or not , new_seletion[0] will naver match
to any of the item in old_selction.So no need to run the loop.

Now when outerloop runs again then their is no need to match new_selection[1]
with old_selection[0] as old_selection[0] is already empty, then again
new_selection[1] will matchs with old_selection[1] and it will be 
new_selection: [" "," ",D]
old_selection: [" "," ",Z,A]
Then no point of running inner loop as new_selection[1] is empty and it will
only matchs with old_selection that have empty, and even after match their is no
point of setting them empty again as they are already empty.

       }
     }
   }
 
   for (size_t i = 0; i < old_selections_.size(); ++i) {
     if (!old_selections_[i].IsEmpty())
       engine_->client_->Invalidate(old_selections_[i]);
   }
   for (size_t i = 0; i < new_selections.size(); ++i) {
     if (!new_selections[i].IsEmpty())
@@ skipping 780 matching lines @@
     double* height) {
   FPDF_DOCUMENT doc = FPDF_LoadMemDocument(pdf_buffer, pdf_buffer_size, NULL);
   if (!doc)
     return false;
   bool success = FPDF_GetPageSizeByIndex(doc, page_number, width, height) != 0;
   FPDF_CloseDocument(doc);
   return success;
 }
 
 }  // namespace chrome_pdf