[doc] fix documents

- ready for release
- fix testbench README command error.
- fix some statements in CodeGen.md
- add serialized binary data description in GenCode.md

docs/GenCode.md

@@ -2,13 +2,13 @@
 
 ## Before Reading
 
-Before starting your reading, please allow me introduce the aim of each generated languages. And, a small, arrogant and realistic warning for C++ code user.
+Before starting your reading, please allow me introduce the aim of each generated languages.
 
-Compiler supports 4 languages now, Python, C\#, C++ and Flatbuffers. Python code is generated for quick test, small demo and etc. Python code is not ready for production environment.   
+Compiler supports 3 languages officially now, Python, C\#, C++ and Flatbuffers. Python code is generated for quick test, small demo and etc. Python code is not ready for production environment.   
 Oppositely, C\# and C++ code is more suit for your productions. C\# code can be used in some game engine easily, such as Unity. C++ code can be applied to game server and etc.  
 Flatbuffers code generation is a migration way. As I said in readme, if you want to migrate to more stable binary protocol, use this output can let you feel more fluent about migration.
 
-You may notice the requirement of some laguages are extremely high. For example, we order C++17 for C++ and .Net Core 2.1 for C\#. That's because the feature we needed only provided in the version higher that our requested version. If you think these version are too high, you can create a fork and change the code on your requirement freely.
+You may notice the requirement of some laguages are extremely high. For example, we order C++17 for C++ and .Net Core 2.1 / .Net Framework 4.6 for C\#. That's because the features we needed only provided in the version higher that our requested version. If you think these version are too high, you can create a fork and change the code on your requirement freely.
 
 ## Data Type Conversion
 
@@ -30,21 +30,37 @@ Following table introduce the data type conversion used by Bp Compiler.
 |uint64	|int	|ulong	|uint64_t		|ulong		|
 |string	|str	|string	|std::string	|string		|
 
-## Container
+### Container
 
 |Bp		|Python		|C\#									|C++				|Flatbuffers	|
 |:---	|:---		|:---									|:---				|:---			|
-|tuple	|list[T]	|T[]									|CStyleArray\<T, N\>|[T:N] (Array)	|
-|list	|list[T]	|System.Collections.Generic.List\<T\>	|std::vector\<T\>	|[T] (Vector)	|
+|tuple	|list[T]	|T[]									|CStyleArray\<T, N\>|[T:N]			|
+|list	|list[T]	|System.Collections.Generic.List\<T\>	|std::vector\<T\>	|[T]			|
 
 Tips:
 
 * Python do not have compulsory type system, so I use Python type hints system instead.
 * Python use `list[T]` for the static and dynamic array of Bp file just due to `tuple[T]` do not support assigning value one by one.
+* In Flatbuffers, `[T:N]` is called Array, and `[T]` is called Vector.
 * String type need special treatment in almost languages although it is grouped as basic type. For convenience, we name the collection of all basic types without string as a new name, primitive type.
 * At the opposite, we call string type and struct type as non-primitive type.
 * Enum also can be seen as primitive type because its inheriting type is primitive type.
 
+## Serialized Binary Data
+
+This section will briefly introduce some features of Serialized Binary Data.
+
+* Serialized Binary Data always is written in Little Endian, no matter the OS endian it is.
+* Align and padding is like C compiler does.
+* String is stored with following steps.
+  1. Encode this string into UTF8 encoding.
+  1. Get the length of this UTF8 string and store it as an `uint32`.
+  1. Store the whole UTF8 string without null terminator.
+* When storing tuple, no space between each members in tuple.
+* List is stored with following steps.  
+  1. Get the length of this list and store it as an `uint32`.
+  1. Store the members of list like tuple.
+
 ## Name Conflict Principle
 
 The templates of generated code may occupy some names to define something. Obviously the generated code can not work if you name your msg or class member to these names. There is the principle instructing how we occupy these names.
@@ -182,6 +198,10 @@ ms.SetLength(0);
 * Bp use `nullptr` as the symbol of null pointer, not `NULL`. If you want to check null pointer for some Bp managed variables (however you might not need to check in almost situation.), please compare its pointer with `nullptr`, not `NULL`.
 * Bp use naked pointer if it needed. But for the most cases, pointer is not used.
 
+### Encoding
+
+The principle order Bp string must be UTF8 encoding. However, our generated C++ code do not ensure this. You should create your encoding convertion code to make sure that all you stored in std::string is writtin in UTF8 encoding.
+
 ### Ownerships
 
 All generated Bp struct and msg follows RAII. And each of them have implemented copy and move semantics.

testbenches/README.md

@@ -1,7 +1,7 @@
 # Testbench Readme
 
 This testbench is served for checking whether the generated code can work.  
-Before using this testbench, you should execute `./bmmo_protocol_compiler -i examples/codegen.bp -p CodeGenTest.py -c CodeGenTest.cs -d CodeGenTest.hpp -s CodeGenTest.cpp -t CodeGenTest.pb` first to get all necessary files.
+Before using this testbench, you should execute `./bmmo_protocol_compiler -i examples/codegen.bp -p CodeGenTest.py -c CodeGenTest.cs -d CodeGenTest.hpp -s CodeGenTest.cpp -b CodeGenTest.fbs` first to get all necessary files.
 
 Next step, put `testbench.py` with `CodeGenTest.py`. Then executing `python3 testbench.py` (I develop it under Python 3.9 and I use a frequently changed module `typing`. Please choose a proper python version.).  
 At first running, script will create a folder called `TestbenchData`. And benchmark the performance of generated ython code.
@@ -15,4 +15,5 @@ Technical infos:
 * C\# need 2 different tests.
   - .Net Framework 4.6 Test
   - .Net Core 2.1 Test
+* Flatbuffers is not included in test.