benchmarks

Benchmarks

Process

Every new benchmark instance generates its own set of unique fixture data! This data is passed & cloned to each function being benchmarked. In this way, every snippet is working with the same data and is not able to introduce side effects to other codes' execution. It is also important to note that cloning the data-array is not tracked as execution time.

Each fixture entry is completely randomized.

• Strings (--type string) can consist of any 8-16 character string
• Numbers (--type number) can be positive or negative, and are usually 9 digits
• Arrays (--type array) are nested every 3rd index; values are Strings (see above) when not another Array
• Objects (--type object) are nested every 3rd index; keys are 8 random alpha-chars; and values are Strings (see above) when not another object

Any number of fixtures can be generated. By default, arrays with type x and length of 100, 500, 1000, 5000, and 10000 are created. You may use the --size flag to pass any number you'd like.

Together, the combination of --types & --sizes should serve as a good basis for testing & benchmarking these functions (and their competitors) in a fair and wide variety of cases.

Please let me know if there is a way in which I can improve the fairness of these benchmarks. Thanks!

Setup

$ npm install

Usage

Commands should be run from the root arr repo; aka cd ..

# run ALL benchmarks -- CAUTION!
$ npm run bench

# run 'map' benchmarks -- string, number, object
$ npm run bench -- map

# specify 'type' and 'size'
$ npm run bench -- map --type string --size 100

# special case: 'flatten'
$ npm run bench -- flatten --type array

Note: The --type flag defaults to string, number, object. The --size flag defaults to 100, 500, 1000, 5000, 10000.

Results:

• Node 4
• Node 6
• Node 7
• Node 8
• Node 10
• Node 12