starovoid / kvrs

About

Console binary key-value storage with output highlighting, searchability with regular expressions and aliasing of bash commands.

Usage

Basic operations

Set a key-value pair:

kvrs set "key" "value" --file="storage.kv"

Get the value by key

kvrs get "key" --file="storage.kv"

Update the value by key

kvrs update "key" "new value" --file="storage.kv"

Delete a key-value pair

kvrs remove "key" --file="storage.kv"

Find a key-value pair by regexp

kvrs find "regexp" --file="storage.kv"

This command tries to treat the keys as text and match them to the sample, returning all matches.

Sort

kvrs sort --file="storage.kv"

Returns a list of keys in lexicographic order.

Run the value as a bash command

kvrs set "list" "ls -la"
kvrs cmd "list" --file="storage.kv"

Common parameters

• --file - the database file that the command will work with. By default, this parameter will be equal to the path to the file "storage.kv" in the home directory.

Roadmap

• get, set, update, remove
• cmd
• sort
• find

kvrs's Issues

Develop a CLI interface that supports 4 basic operations

Primitive CLI interface without pretty output, which panics when something get wrong (it's just a prototype).

Creating new data file

When should we create a new data file?
Now I suggest doing this only with a specific "new" command:

kvrs new "path"

Development of the data file format

Previously, it will look like this.

But there are still a few things to solve:

1. What is the best way to represent the Index in the program: IndexMap<String, u64> or IndexMap<Vec<u8>, u64>. IndexMap is represented by the crate indexmap and allows to track the order in which keys are added.
2. In what format should the Index be serialized? Now we are focusing on postcard.

UPDATED:
Now we are leaning towards using IndexMap<Vec<u8>, u64> in the prototype and potentially replacing Vec<u8> with SmallVec<u8> by the first stable release.

Data format

The data file contains binary values and by default has the .kv extension.

Section	Type	Size (in bytes)
identifier	u64	8
version	u8	1
index position	u64	8
vacant blocks list position	u64	8
record 1	[u8]	arbitary
...	...	...
record n	[u8]	arbitary
vacant blocks list	[u8]	arbitary

---

• Identifier

  The identifier is a "magic" number so that using it the program can make sure that it is actually dealing with a file of the correct format.

• Version

  Just the file format version.

• Index position

  The byte number from which begins record containing the index.

• Vacant blocks list position

  The byte number from which the list of vacant blocks begins.

Record format

Section	Type	Size (in bytes)
data len (size)	u64	8
data	[u8]	size

Vacant blocks list format

Sorted Vec<VacantBlock> serialized in postcard format.

struct VacantBlock {
    pos: u64,   // It's position in data file
    size: u64,  // Size in bytes
}