====Column Diff====
dev environment:
Fedora 17, gcc 4.7 with c++0x
files:
DataSource.h: Define interface for datasource class. This interface can be used for getting and writing data from/to file, database or network.
FileDataSource.h: Inherit DataSource interface and implment default behavior for file related operations.
InputFileDataSource.h: Specific implementation for input file
OutputFileDataSource.h: Specific implementation for output file
DataSourceManager.h: Manage all the datasource, include but not limit to create, get, update and delete datasource object. With manger class, we can easily use different datasource for input or output. For example, we can convenitently use file and database for input and network for output.
DataSourceFactory.h: Facotry is used to create class for each specific types.
Diff.h: Diff class define inteface for diff object, which implementas specific diff operation of 2 types.
IntDiffInt.h: Specific diff behavior between 2 int object
StringDiffString.h: Specific diff behavior between 2 string object
DiffManager.h: Manage all the diff object of 2 types. There is a map inside DiffManage to avoid creating duplicate diff objects.
DiffFactory.h: Similar to DataSourceFactory class.
Utils.h: Define serveral helper functions.
design logic:
1. DataSource class is an abstraction for data source. User need to inherit the datasource interface to create specific datasource class. The DataSource Manager class is used to manage all the available datasources. This will help to seperate specific implementation of each data source from all data sources' common interface. The Factory class is used to create specific data source object.
2. Diff class is used to define the diff operation between 2 types. User need to inherit Diff virtual class in order to define specific diff operation. Similar to datasource, the diff class also has its own manager factory class.
3. I extensively use shared pointer instead of nake pointer to avoid potential memory leaking issue.
4. I use C-style file functions (fgets, fprintf) to process the file, instead of cin and cout, since cin use inherit mechinism within implementation which make it very slow when processing large files. Using fgets instead of fscanf is to avoid potential buffer overflow bug.
5. I use char "_" to separate column type, each value should be no more that 255 characater in length. The empty line inside file will be skiped when comparing.
how to compile, run and test:
0. use make to compile
1. use ./colDiff input1 input2 output
2. testcase dir contain serveral testcases
possible improvement:
1. Refactor factory and manager to create an abstract base class.
2. Add exeception control related code to do more check about input
3. Add all the config related parameter into static Utils class
4. Add more test caes to test the corner and extreme cases.
5. Add XML config to define which diff function currently support in this program, load
each addon function without re-compile the program.