Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/ashwanthkumar/structure_fi_coding_challenge

Structure Financial Coding Challenge
https://github.com/ashwanthkumar/structure_fi_coding_challenge

Last synced: about 6 hours ago
JSON representation

Structure Financial Coding Challenge

Awesome Lists containing this project

README 

        
# structure_fi_coding_challenge



See [PROBLEM.md](PROBLEM.md) for the description of the problem statement.



## Usage

This requires Go 1.17 toolchain to be installed on the machine.



```

$ make && ./structure_fi_coding_challenge

```



In case you have docker and not have Go toolchain installed, you can also use Docker to build and run the project using the following commands



```

# Build the docker image

$ docker build . -t structure_fi_by_ashwanth_kumar

$ docker run -p 8080:8080 -it structure_fi_by_ashwanth_kumar

```



This should start the application on port `8080` as default and we should be consuming the streams on the background.



Once the service starts, you can visit [http://localhost:8080/swagger/index.html](http://localhost:8080/swagger/index.html) for a swagger interface to test out the API implementation.



I've also exposed an interface of `pprof` so we can visualize the profiling of the application while it is running. Once the server is started on `8080`, you can use the following tool to view the profile information on the browser:



```

$ go tool pprof -http=:8081 http://localhost:8080/debug/pprof/heap

```



This should automatically open the browser to see the profile information.



## Solution Notes

> First of all, thank you for the problem statement I had so much fun learning about t-digest and Perfect Hash Functions along the way.



The solution is written in Go and I've written tests pragmatically.



### Median Calculation

Median calculation is backed by [t-digest](https://github.com/tdunning/t-digest/) data structure implementation. I'm using a [go](https://github.com/spenczar/tdigest) implementation in my solution. You can find on [store/store.go](store/store.go).



### Hash Map



> If you follow through the commit history (see `3d12171`, `107122d`, `f344776`), you will notice that I got a working solution (a thin slice) using golang's in-built map data structure with a [Store abstraction](store/store.go) and then eventually replaced it with the custom hash map implementation. This is usually my development style, where I would like to focus on small slices of work and incrementally change / improve the design / code / performance as needed.



Custom Hash Map implementation can be found at [store/custom_map.go](store/custom_map.go). I spent good time going through various [lectures](https://www.youtube.com/watch?v=0M_kIqhwbFo&list=PLUl4u3cNGP61Oq3tWYp6V_F-5jb5L2iHb&index=9), to find an general purpose hash function which will not have any collision and later stumbled upon a blog post that described about Minimal Perfect Hash Function against a known set of keys. Given our keys are the total list of symbols, we can indeed build a MPH using that list. I ended up using the implementation of ["Hash, displace, and compress"](http://cmph.sourceforge.net/papers/esa09.pdf) algorithm by the blog author at https://github.com/dgryski/go-mph. Now given a perfect hash function, I wrote a custom wrapper on top of this to store the values for each key in an array. The maximum memory requirement of this map is `O(n)` where `n` is the total number of symbols. Given the list is finite and known upfront I went with this solution.



Thank you for reading so far, irrespective of the outcome if you've any comments please do send them across.



### Demo



[![Structure.fi Backend Engineer Coding Challenge Demo](https://img.youtube.com/vi/KOoqQsezjRc/hqdefault.jpg)](https://www.youtube.com/watch?v=KOoqQsezjRc)