Add readme about how to run and further improvements

README.md

@@ -0,0 +1,53 @@
+# TOP X numbers
+
+This program reads `a file` which contains one number each line and prints out the `top X numbers` as you wish.
+
+## How to build and run
+
+I use Docker to build and run this little project, so please make sure you have Docker installed.
+
+Then all you need to do are:
+
+```
+# Build
+docker build -t topx .
+
+# Then quick run with default parameters
+docker run -it --rm topx
+
+# Run tests
+docker run -it --rm topx sh -c "TOPX=4 go test ./... -v --cover"
+```
+
+## Available options for running the program
+
+We use environment variables to override program parameters just for simplicity. You can change the following two environment variables:
+
+- `INPUT ` The relative file path which contains the numbers
+- `TOPX` Indicates who many largest numbers you want to pick up.
+
+There are some sample commands below:
+
+```
+
+docker run -it --rm topx sh -c "INPUT=./test1.txt go run main.go"
+
+docker run -it --rm topx sh -c "INPUT=./test1.txt TOPX=1 go run main.go"
+
+docker run -it --rm topx sh -c "TOPX=1 go run main.go"
+
+```
+
+## Complexities
+
+- Space complexity: We only have a heap with size `x` to store the top numbers, and we don't read the entire file into memory , so the space complexity is `O(x)`.
+
+- Time complexity: As we use Priority Heap to store the top `x` numbers, the time complexity for both `Pop` and `Push` are `O(logx)`, and we need to do this for every number, given we have `n` numbers in total. So the complexity of finding top x numbers is `O(nlogx)`, and lastly we do another sorting for the top `x` numbers to make sure numbers are following the descending order, which the time complexity is `O(logx)` as well. So the total time complexity should still be `O(nlogx)`
+
+## Future improvements
+
+- Refactor the `calculator.CalculateLargestNumbers` function to extract file reading, so that this function doesn't need to reply on a actual file in unit and can be tested better, and even to be mocked out when unit testing.
+- Still in the `calculator.CalculateLargestNumbers` function, it would be better to just call `heap.Init()` once no matter how many top numbers we need to fetch.
+- Allow program to parse options for parameters, such as `./app -input x.txt -topx 5`
+- Add benchmark tests to reveal the actual performance of this program, and improve the program from there.
+- Better error handling. Such as custom error types, more specific error messages and so on.