之前看到一篇關於 MySQL index / key 的文章，以中文 MySQL 文章來說，larry 覺得寫得滿好的，分享給大家
https://blog.csdn.net/nanaMasuda/article/details/52543177
(文章內的「索引」就是 index)

“索引被用來快速找出在一個列上用一特定值的行。沒有索引，MySQL不得不首先以第一條記錄開始並然後讀完整個表直到它找出相關的行。”

文章第一句就點出 index 的主要功能：快速檢索。為啥可以快速檢索？因為 MySQL index 會建一物理結構 BTree 加快檢索，也可以想成是用空間換取速度。

Key 與 Index 的差異？

Key 包含了兩個功能

1. 約束 (constraint)，包含本文下方提到的 Unique constraint.
2. 索引 (index)，也就是 key 功能上包含了 index.

Primary 與 Unique key 的差異？

Primary key 的特性是

1. Unique constraint.
2. Each table can only have one Primary key.
3. Cannot be null.

Unique key 的特性是

1. Unique constraint.
2. Each table can have multiple Unique keys.
3. Can be null.

當然，請記得所有的 key 都包含 index 的功能。

Unique key 能否為 null？

larry 有做一些實驗跟查找文章
http://www.mysqltutorial.org/mysql-unique/

“Unlike other database systems, MySQL considers NULL values as distinct values. Therefore, you can have multiple NULL values in the UNIQUE index.

This is how MySQL was designed. It is not a bug even though it was reported as a bug.”

所以以 MySQL 來說，Unique key 是可以為 null 的。

相關文章：MySQL server 在 Mac 上關不掉的問題

這兩天一直在處理一個很詭異的問題，如題：MySQL server 在 Mac 上關不掉。

試了直接砍 process，不行，系統會產生新的 mysql process。搜尋了不少文章後，才發現這篇文章

由上面連結得知，Mac 開機/登入時的 services 有分：

• Launch Daemon
• Launch Agent

Launch Daemon 在開機時動作，Launch Agent 在使用者登入時動作。

到底什麼是 Launch Daemon 呢？其實就是以下這些路徑下的 plist (property list) files:
/System/Library/LaunchDaemons
/Library/LaunchDaemons

系統在開機時會 load 上述路徑的 plist files, 然後要特別注意你的 plist file 是否有 KeepAlive, RunAtLoad 兩個 tag, 如果你不希望該 service keep alive 或 run at load 的話，請做適當的設定。

上面連結也有提到 Launch Agent plist file 的路徑：
/System/Library/LaunchAgents
/Library/LaunchAgents
~/Library/LaunchAgents

總之可以看一下上述幾個路徑底下的 plist file 和其中的設定，像是 mysql 的 plist file 就是在 /Library/LaunchDaemons, 如底下連結所敘：
https://dev.mysql.com/doc/refman/5.1/en/osx-installation-launchd.html

因為 Larry 將 mysql plist file 刪掉後，再重開機，MySQL server 關不掉的問題就不會再犯了，所以就沒有繼續實驗下去。不過照上文推斷，MySQL server 關不掉的問題應該在裝 MySQL 客戶端工具時就一直存在了…

controller.navigationController?.pushViewController(myViewController, animated: true)

如果 controller.navigationController 不為 nil, 執行 function. 如果 controller.navigationController 為 nil, 整行程式碼不執行。他可以替換成：

if controller.navigationController != nil {
    controller.navigationController.pushViewController(myViewController, animated: true)
}

The “if let” Syntax

if let nav = controller.navigationController {
    nav.pushViewController(myViewController, animated: true)
} else {
    //show an alert ot something else
}

“if let” 的語法隱含了 “and if nav != nil”. 當你有多行程式碼需要 unwrap, use “if let”.

The “as?” Syntax

let myVC = controller.presentedViewController as? MyViewController
myVC?.lastUpdated = NSDate()

“as?” 有點像 C++ 的 dynamic cast, 以上面例子來說，myVC 有可能是 nil. 正因如此，”as?” 底下通常要接 “class?property”. 另外如果用 “as” 而不是 “as?” (沒有問號)，則是 C++ 的 static cast，上面文章連結也有提到你要 101% 確定可以 cast 才用 “as” 去 cast.

pic1

拖拉一個 “Bar Button Item” 到 navigation bar, 在右邊 attributes inspector 可調整 “Bar Button Item” 為 Add, Edit Done, Cancel 等一般我們在 App 上方看到的。Control-drag 剛剛新增的 bar button 到你要到的 view controller, 兩個 view controller 會用 pic2 (或是 pic3) 做連結，切換過去的 view controller 左上角會自然出現 Back.

pic2

pic3

pic1, 2, 3 在蘋果官方文件稱之為 “segue” (發音類似 Segway, 順道一提，Segway 被中國的 Ninebot 買了下來，雷軍是 Ninebot 的股東，所以 10/19 在小米電視3的發佈會上雷軍也發佈了 Ninebot 的產品，這樣應該有助於記憶 segue

Segue 有幾種類型：push, modal, 等 (Xcode7是叫 show, present modally), 目前的理解是一般在 App 中切換頁面是用 push, modal 比較像之前 Win32 modal dialog 的概念，像是 add new item 就可以用 modal.
上面提到 push 切換過去的 view controller 左上角自然出現 back, 但是 modal 切換過去的 view controller 是完全空白的 (所以要 dismiss 自己)

About the Unwind Segue.

// In your origin view controller
override func prepareForSegue(segue: UIStoryboardSegue, sender: AnyObject?) {
} 

// In your destination view controller
@IBAction func unwindToYourDestinationView(sender: UIStoryboardSegue) {
}

將 storyboard origin view 的 Save button 拉到 origin view 上方的 Exit item, 選擇 unwindToYourDestinationView. 這樣按 Save 時就會觸發上方兩個 function (將 modal view 設定的資料傳回)。Cancel 的部分簡單藉由 Action handler, 在其中調用 dismissViewControllerAnimated 即可。

var someInts = [Int]()
var threeDoubles = [Double](count: 3, repeatedValue: 0.0)

array literal (shorthand):

var shoppingList: [String] = ["Eggs", "Milk"]
var shoppingList = ["Eggs", "Milk"]

用 [0], [1], … subscript syntax (蘋果官方文件取的名詞) 來存取 member.

Set
initializer syntax:

var letters = Set<Character>()

array literal (對，他不是叫 set literal):

var favoriteGenres: Set<String> = ["Rock", "Classical", "Hip hop"]
var favoriteGenres: Set = ["Rock", "Classical", "Hip hop"]

(沒有指定 type Set 就會被當作 Arrary)

Set 不能用 subscript syntax 來存取 member，合理，因為 Set 沒有順序性。用 mySet.contains(item) 來判斷 item 是否存在於 mySet.

Dictionary
initializer syntax:

var namesOfIntegers = [Int: String]()

dictionary literal:

var airports: [String: String] = ["YYZ": "Toronto Pearson", "DUB": "Dublin"]
var airports = ["YYZ": "Toronto Pearson", "DUB": "Dublin"]

存取用 subscript syntax。

跟 Python 比起來，Swift collections 都是中括號。Swift 也有 Tuple (小括號)，用於 loop 的參數值，function 的 input/output 值等等，但不屬於 collection.

3. OGRE Android building tutorial:
http://www.ogre3d.org/tikiwiki/tiki-index.php?page=CMake+Quick+Start+Guide&tikiversion=Android
The sample was successfully built and run, but the details of the building steps are still unknown.

Here is a nice article regarding using C++ to develop metro style app:
http://blogs.msdn.com/b/ericsk/archive/2012/05/02/metro-style-app-dev-using-cpp.aspx

The C++ in metro style app development is actually called C++/CX. From wiki:
http://en.wikipedia.org/wiki/C%2B%2B/CX
“The language extensions borrow syntax from C++/CLI but target the Windows Runtime and native code instead of the Common Language Runtimeand managed code.”

For me, C++/CX is so like C++/CLI. Since I have some experience in developing C++/CLI, and personally I don't prefer it. Therefore, I am conservative about C++/CX. Here are the pros and cons when using C++ in metro style app development.
Pros:
1. It's C++. It's somehow a general language for all programmers.
2. It can be easily integrated with Boost library, multimedia codecs, and other C++ libraries.

Cons:
1. C++/CX is not perfectly the same as C++ and it's not popular, while, say, C++, JAVA, and C# are popular.
2. (At least for me) It's ambiguous. Ambiguity comes from things are alike, but one has slight differences from another.

Then, how about the huge number of MFC softwares which have been existing for many many years? They still can run on Windows 8 desktop mode.
Since now we have WinRT and Win32 even in the same computer, is there any cross-platform issue? Yes, there is. Here is a Microsoft nice article:
http://msdn.microsoft.com/zh-tw/magazine/jj651574.aspx

At this time I won't say it's a cross-platform issue. In practice, I think the metro project and the desktop project are going to be different two, with largely different UI. At most, they use shared modules, but personally I don't think it's a cross-platform concept when we are using C++ to develop. On the other hand, if you really plan to do a cross-platform project (including Android and IOS), html/javascript is the best choice. But again, what to do depends on the project requirements and what is already there.

TEST(test_case_name, test_name) 
{
  //... test body ...
}

Test Fixture

TEST_F(test_case_name, test_name) 
{
  //... test body ...
}

The above test_case_name must be your test fixture class, which must inherit from testing::Test. You can optionally override the functions: SetUp and TearDown. In a TEST_F, the flow is like:
1. the test fixture class constructs.
2. its SetUp function is called (if there is).
3. do your arrangement and assertion works.
4. the TearDown function is called (if there is).
5. the test fixture class destructs.

The above a, b, d, e are automatically done, which means you don't see it in a TEST_F. However, step c is the tester's job.


Typed Test
1. Implement a template fixture class (a template class inherits from testing::Test).
2. Use TYPED_TEST_CASE to define the data types that will run as the template argument of the template fixture class.
3. Use TYPED_TEST to start a test case. Note the 1st argument should be the template fixture class. Besides, the test case will run N times depending on how many data types you define in the above step 2.


Type-Parameterized Test
1. Implement a template fixture class (a template class inherits from testing::Test).
2. Use TYPED_TEST_CASE_P to inform the framework that you will have a test case.
3. Use TYPED_TEST_P to start a test case. Note until now the framework has no idea what data types to run in the test case.
4. An extra step: use REGISTER_TYPED_TEST_CASE_P to register the test case and all the tests under it.
5. Finally, tell the framework what data types to run in the test case by using INSTANTIATE_TYPED_TEST_CASE_P (see v1_6_AdvancedGuide for the usage).

By the test method, the same test case can have different sets of testing data types.


Value-Parameterized Test
1. Implement a fixture class (inherits from testing::TestWithParam<T>, and T must be assigned).

TestEventListeners& listeners = unit_test.listeners();
delete listeners.Release(listeners.default_result_printer());
listeners.Append(new TersePrinter);

Then you can have your custom event listener when, say, a test starts/ends, assertion fails/succeeds, etc.

Note that sample10_unittest.cc demonstrates a nice way for memory leak check, where each class in your system should keep the creating count in such class, and the client side also keeps a count. Also note this is just one way to prevent from memory leak.