Add functionality to the install_static library to tokenize strings and compare versions.

chrome/install_static/install_util.cc

 // Copyright 2014 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 "chrome/install_static/install_util.h"
 
 #include <windows.h>
 #include <assert.h>
 #include <algorithm>
+#include <iostream>
+#include <iterator>
 #include <memory>
+#include <sstream>
 
 #include "base/macros.h"
 #include "build/build_config.h"
 
 namespace install_static {
 
 ProcessType g_process_type = ProcessType::UNINITIALIZED;
 
 // TODO(ananta)
 // http://crbug.com/604923
@@ skipping 227 matching lines @@
 }
 
 // Returns the executable path for the current process.
 base::string16 GetCurrentProcessExePath() {
   wchar_t exe_path[MAX_PATH];
   if (GetModuleFileName(nullptr, exe_path, arraysize(exe_path)) == 0)
     return base::string16();
   return exe_path;
 }
 
-// UTF8 to UTF16 and vice versa conversion helpers. We cannot use the base
-// string conversion utilities here as they bring about a dependency on
-// user32.dll which is not allowed in this file.
-
-// Convert a UTF16 string to an UTF8 string.
-std::string utf16_to_utf8(const base::string16 &source) {
-  if (source.empty())
-    return std::string();
-  int size = ::WideCharToMultiByte(CP_UTF8, 0, &source[0],
-      static_cast<int>(source.size()), nullptr, 0, nullptr, nullptr);
-  std::string result(size, '\0');
-  if (::WideCharToMultiByte(CP_UTF8, 0, &source[0],
-          static_cast<int>(source.size()), &result[0], size, nullptr,
-          nullptr) != size) {
-    assert(false);
-    return std::string();
-  }
-  return result;
-}
-
-// Convert a UTF8 string to a UTF16 string.
-base::string16 utf8_to_string16(const std::string &source) {
-  if (source.empty())
-    return base::string16();
-  int size = ::MultiByteToWideChar(CP_UTF8, 0, &source[0],
-      static_cast<int>(source.size()), nullptr, 0);
-  base::string16 result(size, L'\0');
-  if (::MultiByteToWideChar(CP_UTF8, 0, &source[0],
-          static_cast<int>(source.size()), &result[0], size) != size) {
-    assert(false);
-    return base::string16();
-  }
-  return result;
-}
-
 bool RecursiveDirectoryCreate(const base::string16& full_path) {
   // If the path exists, we've succeeded if it's a directory, failed otherwise.
   const wchar_t* full_path_str = full_path.c_str();
   DWORD file_attributes = ::GetFileAttributes(full_path_str);
   if (file_attributes != INVALID_FILE_ATTRIBUTES) {
     if ((file_attributes & FILE_ATTRIBUTE_DIRECTORY) != 0) {
       Trace(L"%hs( %ls directory exists )\n", __FUNCTION__, full_path_str);
       return true;
     }
     Trace(L"%hs( %ls directory conflicts with an existing file. )\n",
@@ skipping 109 matching lines @@
   // 1. We consider current character of source.
   // 2. We ignore current character of source.
   if (pattern[pattern_index] == L'*') {
     return MatchPatternImpl(source, pattern, source_index + 1,
                             pattern_index) ||
            MatchPatternImpl(source, pattern, source_index, pattern_index + 1);
   }
   return false;
 }
 
+// Defines the type of whitespace characters typically found in strings.
+const char kWhiteSpaces[] = " \t\n\r\f\v";

[grt (UTC plus 2), 2016/05/27 14:13:20]

nit: constexpr

[ananta, 2016/05/27 20:05:25]

Done.

+
+// Trim whitespaces from left.
+void ltrim(std::string* str) {
+  str->erase(0, str->find_first_not_of(kWhiteSpaces));
+}
+
+// Trim whitespaces from right
+void rtrim(std::string* str) {
+  str->erase(str->find_last_not_of(kWhiteSpaces) + 1);
+}
+
+// Trim whitespaces from left & right
+void trim(std::string* str) {

[scottmg, 2016/05/27 17:27:22]

trim -> Trim

[ananta, 2016/05/27 20:05:25]

Done.

+  rtrim(str);

[scottmg, 2016/05/27 17:27:22]

Seems like over-functioning since ltrim() and rtrim() are never used. Just put
the two erase() calls into Trim().

[ananta, 2016/05/27 20:05:25]

Done.

+  ltrim(str);
+}
+
+bool is_valid_number(const std::string &str) {

[scottmg, 2016/05/27 17:27:22]

IsValidNumber

[ananta, 2016/05/27 20:05:25]

Done.

+  return std::all_of(str.begin(), str.end(), ::isdigit);

[grt (UTC plus 2), 2016/05/27 14:13:20]

as implemented, this will return true if std.empty(). should it not return false
instead?

[ananta, 2016/05/27 20:05:25]

Done.

+}
+
 }  // namespace
 
 bool IsSxSChrome(const wchar_t* exe_path) {
   return wcsstr(exe_path, L"Chrome SxS\\Application") != NULL;
 }
 
 bool IsSystemInstall(const wchar_t* exe_path) {
   wchar_t program_dir[MAX_PATH] = {};
   DWORD ret = ::GetEnvironmentVariable(L"PROGRAMFILES", program_dir,
                                        arraysize(program_dir));
@@ skipping 137 matching lines @@
   // http://crbug.com/591504.
   if (!RecursiveDirectoryCreate(crash_dir->c_str()))
     return false;
   crash_dir->append(L"\\Crashpad");
   return true;
 }
 
 
 std::string GetEnvironmentString(const std::string& variable_name) {
   DWORD value_length = ::GetEnvironmentVariable(
-      utf8_to_string16(variable_name).c_str(), NULL, 0);
+      utf8_to_utf16(variable_name).c_str(), NULL, 0);
   if (value_length == 0)
     return std::string();
   std::unique_ptr<wchar_t[]> value(new wchar_t[value_length]);
-  ::GetEnvironmentVariable(utf8_to_string16(variable_name).c_str(),
+  ::GetEnvironmentVariable(utf8_to_utf16(variable_name).c_str(),
       value.get(), value_length);
   return utf16_to_utf8(value.get());
 }
 
 bool SetEnvironmentString(const std::string& variable_name,
                           const std::string& new_value) {
-  return !!SetEnvironmentVariable(utf8_to_string16(variable_name).c_str(),
-      utf8_to_string16(new_value).c_str());
+  return !!SetEnvironmentVariable(utf8_to_utf16(variable_name).c_str(),
+      utf8_to_utf16(new_value).c_str());
 }
 
 bool HasEnvironmentVariable(const std::string& variable_name) {
-  return !!::GetEnvironmentVariable(utf8_to_string16(variable_name).c_str(),
+  return !!::GetEnvironmentVariable(utf8_to_utf16(variable_name).c_str(),
       NULL, 0);
 }
 
 bool GetExecutableVersionDetails(const base::string16& exe_path,
                                  base::string16* product_name,
                                  base::string16* version,
                                  base::string16* special_build,
                                  base::string16* channel_name) {
   assert(product_name);
   assert(version);
@@ skipping 130 matching lines @@
   registry_path += kBrowserCrashDumpMetricsSubKey;
   return registry_path;
 }
 
 bool MatchPattern(const base::string16& source,
                   const base::string16& pattern) {
   assert(pattern.find(L"**") == base::string16::npos);
   return MatchPatternImpl(source, pattern, 0, 0);
 }
 
+std::string utf16_to_utf8(const base::string16 &source) {

[grt (UTC plus 2), 2016/05/27 14:13:20]

nit: & before space.  please use "git cl format" to catch these sorts of things.

[ananta, 2016/05/27 20:05:25]

Yeah. I had done that. It seems every time I copy paste blocks in VS it
rearranges these things. Need to look for the option to turn it off.

+  if (source.empty())

[grt (UTC plus 2), 2016/05/27 14:13:20]

|| source.size() > std::numeric_limits<int>::max()

[ananta, 2016/05/27 20:05:25]

Done.

+    return std::string();
+  int size = ::WideCharToMultiByte(CP_UTF8, 0, &source[0],
+    static_cast<int>(source.size()), nullptr, 0, nullptr, nullptr);
+  std::string result(size, '\0');
+  if (::WideCharToMultiByte(CP_UTF8, 0, &source[0],
+    static_cast<int>(source.size()), &result[0], size, nullptr,
+    nullptr) != size) {
+    assert(false);
+    return std::string();
+  }
+  return result;
+}
+
+base::string16 utf8_to_utf16(const std::string &source) {
+  if (source.empty())

[grt (UTC plus 2), 2016/05/27 14:13:20]

|| source.size() > std::numeric_limits<int>::max()

[ananta, 2016/05/27 20:05:25]

Done.

+    return base::string16();
+  int size = ::MultiByteToWideChar(CP_UTF8, 0, &source[0],
+    static_cast<int>(source.size()), nullptr, 0);
+  base::string16 result(size, L'\0');
+  if (::MultiByteToWideChar(CP_UTF8, 0, &source[0],
+    static_cast<int>(source.size()), &result[0], size) != size) {
+    assert(false);
+    return base::string16();
+  }
+  return result;
+}
+
+std::vector<std::string> TokenizeString(const std::string& str,
+                                        const char delimiter,

[grt (UTC plus 2), 2016/05/27 14:13:20]

no const

[ananta, 2016/05/27 20:05:25]

Done.

+                                        bool trim_spaces) {

[scottmg, 2016/05/27 17:27:22]

I only see trim_spaces=false calls other than in test code. Can we just get rid
of this argument and the related trim functions?

[ananta, 2016/05/27 20:05:25]

Prefer to leave it as an argument. Just in case we need that functionality

+  std::vector<std::string> tokens;
+  std::istringstream buffer(str);
+  for (std::string token; std::getline(buffer, token, delimiter); ) {
+    if (trim_spaces)
+      trim(&token);
+    tokens.push_back(token);
+  }
+  return tokens;
+}
+
+bool CompareVersionStrings(const std::string& version1,
+                           const std::string& version2,
+                           int* result) {
+  if (version1.empty() || version2.empty())
+    return false;
+
+  // Tokenize both version strings with "." as the separator. If either of
+  // the returned token lists are empty then bail.
+  std::vector<std::string> version1_components =
+    install_static::TokenizeString(version1, '.', false);

[grt (UTC plus 2), 2016/05/27 14:13:20]

omit "install_static::"

[ananta, 2016/05/27 20:05:25]

Done.

+  if (version1_components.empty())
+    return false;
+
+  std::vector<std::string> version2_components =
+    install_static::TokenizeString(version2, '.', false);
+  if (version2_components.empty())
+    return false;
+
+  // You may have less tokens in either string. Use the minimum of the number
+  // of tokens as the initial count.
+  const size_t count = std::min(version1_components.size(),

[scottmg, 2016/05/27 17:27:22]

nit; weird line break

[ananta, 2016/05/27 20:05:25]

Done.

+      version2_components.size());
+  for (size_t i = 0; i < count; ++i) {
+    // If either of the version components don't contain valid numeric digits
+    // bail.
+    if (!is_valid_number(version1_components[i]) ||
+        !is_valid_number(version2_components[i])) {
+      return false;
+    }
+
+    int version1_component = std::stoi(version1_components[i]);

[scottmg, 2016/05/27 17:27:22]

Ooh, this defaults to base=10, not =0, woohoo!

[ananta, 2016/05/27 20:05:25]

Yeah

+    int version2_component = std::stoi(version2_components[i]);
+
+    if (version1_component > version2_component) {
+      *result = 1;
+      return true;
+    }
+
+    if (version1_component < version2_component) {
+      *result = -1;
+      return true;
+    }
+  }
+
+  // Handle remaining tokens. Here if we have non zero tokens remaining in the
+  // version 1 list then it means that the version1 string is larger. If the
+  // version 1 token list has tokens left, then if either of these tokens is
+  // greater than 0 then it means that the version1 string is smaller than the
+  // version2 string.
+  if (version1_components.size() > version2_components.size()) {
+    for (size_t i = count; i < version1_components.size(); ++i) {
+      // If the version components don't contain valid numeric digits bail.
+      if (!is_valid_number(version1_components[i]))
+        return false;
+
+      if (std::stoi(version1_components[i]) > 0) {
+        *result = 1;
+        return true;
+      }
+    }
+  } else if (version1_components.size() <  version2_components.size()) {
+    for (size_t i = count; i < version2_components.size(); ++i) {
+      // If the version components don't contain valid numeric digits bail.
+      if (!is_valid_number(version2_components[i]))
+        return false;
+
+      if (std::stoi(version2_components[i]) > 0) {
+        *result = -1;
+        return true;
+      }
+    }
+  }
+  // Here it means that both versions are equal.
+  *result = 0;
+  return true;
+}
+
 }  // namespace install_static