guaqueta / in-memory-db-takehome

Takehome Assessment

In-memory Database

I provide two implementations

Keeping count with Python

The db.py file can be run with python3. With no arguments it provides an interactive prompt to run database commands. On a shell you can also stream a file of commands to it (one per line) by running python3 db.py < filename.

In the python implementation we maintain two hash-maps: one for the database values, and one for the counts. Transactions are implemented as a stack of databases, with more recent transactions shadowing older ones. SET, GET, DELETE, and COUNT are O(m) (where m is the number of open transactions). BEGIN and ROLLBACK are O(1), and COMMIT is O(m k), where m is the number of open transactions and k is the number of affected keys per transaction.

Being lazy with Clojure

To run the clojure implementation you can:

1. Install the clojure command-line tools. With db.clj and deps.edn in your directory you can then run clojure -M db.clj.
2. Use this replit. I uploaded my solution to a repl.it so that you can run it in the browser. It's slow, but doesn't require any installation.

The clojure implementation uses a library that implements a hash-map sorted by values. This priority-map data structure provides O(log n)-or-better GET, SET, DELETE, and COUNT. For transactions we rely on clojure's immutable data structures. Opening a new transaction is as simple as hanging on to a reference to the current database value. BEGIN, ROLLBACK, and COMMIT are all O(1), since they are just changing the current reference to the database. I couldn't resist implementing the clojure solution since all the hard work is already done for me.