What started as me wanting to understand how databases actually work turned into a months-long project that I'm genuinely proud of.

I'll be completely honest - I relied on a book "Build Your Own Database from Scratch in Go" and studied a bunch of existing codebases to understand the concepts. But here's the thing: I implemented the code myself as a learning exercise, and ended up adding more features than the book covered.

What I built:

Complete B+ tree storage engine with memory-mapped I/O
Full ACID transactions with BEGIN/COMMIT/ABORT support
Concurrent reads using worker pools
SQL-like operations: CREATE, INSERT, UPDATE, DELETE, GET
Aggregate functions: COUNT, SUM, AVG, MIN, MAX
Range queries and composite indexing
Cross-platform single binary that works on Linux, macOS, Windows
Zero external dependencies except (golang.org/x/sys)

Performance numbers from my testing:

1,813 operations per second for inserts
1,848 operations per second for queries
Sub-millisecond latency
Only 0.88 KB storage overhead per record

This was purely educational - I wanted to understand database internals, memory management, and concurrent programming at a deep level. The fact that it actually works reliably and handles real workloads is honestly pretty exciting for me.

Sometimes the best way to learn something is to build it yourself, even if you need guidance along the way.

Check it Out:

Github: https://github.com/sharvitKashikar/FiloDB