jsonschema-key-compression

Compress json-data based on its json-schema while still having valid json. It works by compressing long attribute-names into smaller ones and backwards.

For example this:

{
    "firstName": "Corrine",
    "lastName": "Ziemann",
    "title": "Ms.",
    "gender": "f",
    "zipCode": 75963,
    "countryCode": "en",
    "birthYear": 1960,
    "active": false,
    "shoppingCartItems": [
        {
            "productNumber": 29857,
            "amount": 1
        },
        {
            "productNumber": 53409,
            "amount": 6
        }
    ]
}

becomes this:

{
    "|e": "Corrine",
    "|g": "Ziemann",
    "|j": "Ms.",
    "|f": "f",
    "|k": 75963,
    "|d": "en",
    "|c": 1960,
    "|a": false,
    "|i": [
        {
            "|h": 29857,
            "|b": 1
        },
        {
            "|h": 53409,
            "|b": 6
        }
    ]
}

Efficiency

The efficiency depends on the amount and length of the attribute names.

• The uncompressed json-object from above has about 230 chars as string
• With the key-compression, this can be reduced to 140 chars which saves about 40%
• Just using gzip on the json would result in 180 chars
• Using gzip+key-compression ends in a string with only 127 chars

You can reproduce these results by running npm run test:efficiency.

Performance

The compression works pretty fast. Here are some time measurements on a single intel i7 CPU.

• Creating a compression-table from the schema of the object above takes about 0.02ms
• Compressing the example-object from above takes about 0.021ms
• Decompressing takes about 0.027ms per object

You can reproduce these results by running npm run test:performance.

You should use this when

• you want to save storage space in an NoSQL-database but still want to have valid queryable json-data
• you transmit many objects in many small requests over the network so that gzip cannot be efficient
• you want to store json-data inside of the browser-storage (indexedDB or localstorage) and you reach the storage limit

You should NOT use this when

• you send many objects in a single request, you should rely on gzip instead
• you do not want to still have valid json-data, you should use protobuf instead
• you have schema-less data

Comparison with gzip

Gzip generates its compression-flags from the input. This makes it more efficient, the more data is compressed at once. But gzip is less efficient the smaller the dataset is. The key-compression creates the compression-table from the jsonschema up front with has advantages when small pieces of data are compressed.

Usage

Install

npm install jsonschema-key-compression --save

createCompressionTable

Creates a compression-table from the json-schema.

import {
    createCompressionTable
} from 'jsonschema-key-compression';
const compressionTable = createCompressionTable(jsonSchema);

compressObject

Compress a json-object based on its schema.

import {
    compressObject
} from 'jsonschema-key-compression';
const compressedObject = compressObject(
    compressionTable,
    jsonObject
);

decompressObject

Decompress a compressed object.

import {
    decompressObject
} from 'jsonschema-key-compression';
const jsonObject = decompressObject(
    compressionTable,
    compressedObject
);

compressedPath

Transform a chain of json-attributes into its compressed format.

import {
    compressedPath
} from 'jsonschema-key-compression';
const compressed = compressedPath(
    compressionTable,
    'whateverNested.firstName'
); // > '|a.|b'

decompressedPath

Decompress a compressed path.

import {
    decompressedPath
} from 'jsonschema-key-compression';
const decompressed = decompressedPath(
    compressionTable,
    '|a.|b' // from compressedPath
); // > 'whateverNested.firstName'

compressQuery

Compress a mango-query so that it can run over a NoSQL-database that has stored compressed documents.

import {
    compressQuery
} from 'jsonschema-key-compression';
const compressed = compressQuery(
    compressionTable,
    {
        selector: {
            active: {
                $eq: true
            }
        },
        skip: 1,
        limit: 1,
        fields: [
            'id',
            'name'
        ],
        sort: [
            'name'
        ]
    }
);