Build a Tiny Shell‑Based Database in Two Commands – How It Works

Simple shell‑based key‑value store

Two shell functions implement a persistent key‑value database stored in a plain‑text file data.db.

db_set appends a line key,value to the file:

db_set() {
    echo "$1,$2" >> data.db
}

This log‑append write is O(1) and guarantees durability because the data is flushed to disk.

db_get retrieves the most recent value for a given key by scanning the file, selecting matching lines, taking the last occurrence, and stripping the key:

db_get() {
    grep "^$1" data.db | tail -n 1 | sed -e "s/^$1,//"
}

grep "^$1"

matches lines whose prefix equals the key argument. tail -n 1 keeps the newest entry, and sed -e "s/^$1,//" removes the key and the separating comma, leaving only the value.

Shell primitives used

grep

– pattern matching at the beginning of each line. tail – selects the last line of the filtered output. sed -e – applies a single substitution; the omission of the g flag ensures only the first match on each line is replaced, which is sufficient because each line contains a single key, prefix.

Performance characteristics

Writes are constant‑time (O(1)) because they are simple appends, similar to a write‑ahead log used by many production databases. Reads require a full scan of data.db, giving linear time O(n) where n is the number of stored entries.

Possible extensions – indexing

To improve read performance an auxiliary index can be maintained. The index stores metadata (e.g., byte offset or line number) for each key, allowing db_get to jump directly to the latest entry. Maintaining the index adds write overhead because each db_set must also update the index structure.

Illustrative images

References

Zhihu article: https://zhuanlan.zhihu.com/p/351897096

Book: Designing Data‑Intensive Applications