heidemn / db.sh

Append-only Bash database

db.sh - Append-only Bash database

Roughly based on an example in Martin Kleppmann's book Designing Data-Intensive Applications.

Additional, unique feature: The database file itself is also a valid Bash script 🤪🥴.

How does it work?

Startup: The database file is "sourced" to evaluate all assignments from previous runs.

Write operations: Write operations are specified in the form of a Bash variable assigment instruction.
This instruction is 1) executed with eval and 2) appended to the database file.

Read operations: Read operations are specified simply as the variable name.
This variable is printed using eval "printf '%s\n' ...".

All variables are internally prefixed with dbsh_, to avoid overwriting predefined variables.

Examples:

# Write command:
myvar="\"That's a hell of a database\", he said."
# ...is executed as:
eval dbsh_myvar="\"That's a hell of a database\", he said."
echo dbsh_myvar="\"That's a hell of a database\", he said." >> "$db_file"

# Read command:
myvar
# ...is executed as:
printf eval "printf '%s\n' \"$dbsh_myvar\""

Is this safe?

This code is using eval. Can this be safe?

./db.sh -c "safe=Yes \"No, of course not...\""
./db.sh: line 62: No, of course not...: command not found

Since this seems too dangerous, the option -k was added.
The code then follows mostly the example from the book.
This avoids 99% of the security issues, but doesn't store the database as real a Bash script any more.

Should I use it in production?

If you currently use a Powershell script as your database, then it could be an improvement.

Otherwise, use it only if you like databases that allow arbitrary code execution.