Recently FOMO got to me and I jumped on crypto speculation game. You gots to do bunch of accounting/counting/reconciling/tracking in it. Sounds boring, but clojure made it fun. And that’s what I want to share here.

There’s plenty of portfolio tracking tools on the nets. If it wasn’t for clojure I would go for one of those and would reluctantly give-out info about my trades to a third party. Rather that though than teaching C# or java compiler about structure of my data with classes/structs so that I can start deriving info from it. Even with dynamic Python it would be a hassle.

Wrap some code around data

Instead of waxing poetic about simplicity, I would best describe hacking on clojure as “wrap some code around data”. Let me illustrate: in my crypto game, I have a bunch of transactions - facts about buying/selling cryptos and deposits I sent into the exchange. When I make a transaction I append it to an EDN file “transactions.edn”:

Transaction log - “transactions.edn”

{
 :deposit 
  [ 
   ^{:date #inst "2020-04-02"} [2340] 
   ^{:date #inst "2020-04-07"} [3200]
  ]
  :buy ;[symbol amount unit-price]
  [ 
     ^{:date #inst "2020-04-02"} [:FIL 8 146.02473684210526]
     ^{:date #inst "2020-04-08"} [:DOGE 30000 0.05223200000000001]
     ^{:date #inst "2020-04-09"} [:LTC 3 192.82999999999998]
     ^{:date #inst "2020-04-11"} [:XRP 370 1.1966222222222223]
     ^{:date #inst "2020-05-10" 
       :fee 15.31 :txfe0 27.58}  [:ETH 0.3 579.0333333]
     ^{:date #inst "2020-05-20"} [:XRP 150 0.9904]
     ^{:date #inst "2020-05-24"} [:XRP 120 0.71428571428]
  ]
 :sell ;[symbol amount at-unit-price for-unit-price]
  [
    ^{:date #inst "2020-04-17"} [:DOGE 20000 0.05223200000000001 0.2255935]
  ] 
}

It’s a map with 3 kinds of transactions: :deposit :buy :sell. Transaction is an array (vector) that has a tag of a crypto, amount and price. Dates are metadata about my transactions and clojure has a notion of metadata built-in, you can attach it to the peace of data (like an array), read it when you want it, but keep it out of the way of calculations.

Tags are important to designate data, it’s built-in into clojure too. Data that starts with a colon is a tag. Clojure calls those keywords. Plus, in clojure, keywords can be namespaced like :sidestep.crypto.game/BTC.

Time to wrap some code around my transactions to start deriving info about how am I doing in the crypto game. First, I want to know my position - the aggregated price of each of my purchased cryptos, compare it to the current market price, calculate the diffs, sum-up the diffs to get the totals. So I create a portfolio.clj file, load my transactions with one liner, and wrap some code on them:

(let [{:keys [deposit buy sell]} (load-file "transactions.edn")
   ;pseudo code to derive info from my transaction log:
  (print-table
    (position 
       (market-data (tags-from buy))
       (aggregated buy sell))))

This line: (let [{:keys [deposit buy sell]} (load-file "transactions.edn") makes all my different kinds of transactions available for analysis in datastructures I need them in (arrays, maps, sets) populated with typed data. Numbers are numbers (floats, ints doubles etc.). Strings, dates, tags (keywords) and metadata are automagically parsed, typed and ready for me. This is not your JSONs.

The beauty is that I get all this pretty much out of the box. Clojure is ready to work with your data. In it’s core it has multitude of functions ready to be wrapped around it.

Data sits in the center

Further, I need accounting info for reconciliation with an exchange, I need alerts for lucrative market changes, I need market info on cryptos I keep my eye on etc… All of those modules I coded gradually in dedicated files that orbit the transaction log. Data in transactions.edn is what drives them in metaprogramming fashion.

Data all the things!!!

First thing that looks weird about clojure to newcomers is the parenthesis tangle. It’s because you’re looking at it wrong. Look at it as data - code is data.

This is clearly data, right?How about this then?
<position>
  <market-data>
     <tags-from buy />
  </market-data>
  <aggregated buy sell />
</position>
(position 
     (market-data (tags-from buy))
     (aggregated buy sell))

Instead of

<position> <market-data>...</market-data>...</position>

you write

(position (market-data ...)...)

see what I mean? You are literally writing data. Evaluatable data. Data you can wrap with more evaluatable data (code) and process.

This is not your Javas.

What else you can do with data? Manipulate it with code of course! Emm… code is data… manipulate data with code… 🤯

You can also store it for later manipulation/evaluation. Or perhaps send it to another data processor that wraps it, evaluates it and spits-out more evaluatable data. You can compose those data evaluators in pipes-and-filters fashion, perhaps with channels. Or hub-and-spoke them as in my example where data evaluators orbit the central data store.

Start from data

You see how data-centric clojure is. To get the fun of clojure try this data-first mindset on:

This is not your TDDs.

[Add comment]