Rust and mmap
The library is published in crates.io and can help with counting and filtering json files with symmetric records as the format below.
[{..}, {..}, ..]
The idea of using memory mapped i/o is check the performance while filtering and creating another file etc from rust while putting serde in harmsway :grin:. Although, here is the best read on the topic from Linux.
Functionality
- Count the number of records in JSON file.
- Count the number of records with filter
- Filter the JSON file with a condition ( provided by the caller ) and save it to a file specified.
Test
The airports JSON has been taken from the location. It needs to be downloaded and put in the data/ directory.
Thanks to the original mmap lib.
General performance seems to be of the order below.
Debug
- counttestsimplenestedjson: 256.37µs seconds for counting 1 records
- counttestsimplewithfilter_json: 322.471µs seconds.
- counttestsimplenestedwithfilterjson: 396.664µs seconds
- counttestsimple_json: 258.594µs seconds for counting 1 records
- filteroutjsonnoresults: 675.752µs seconds.
- filteroutjsonbyvalue: 785.313µs seconds.
- count_airports: 690.802302ms seconds.
- countwithfilter_airports: 3.913697422s seconds for filtering 57265 records
- filteroutairportsincountry 3.91415908s: seconds for filtering 57265 records
- filteroutairportsnoresults: 3.922528546s seconds for filtering 57265 records
Release
- counttestsimplenestedjson: 147.219µs seconds for counting 1 records
- counttestsimplewithfilter_json: 164.697µs seconds.
- counttestsimple_json: 151.941µs seconds for counting 1 records
- counttestsimplenestedwithfilterjson: 339.415µs seconds
- filteroutjsonnoresults: 631.994µs seconds.
- filteroutjsonbyvalue: 888.325µs seconds.
- count_airports: 20.457554ms seconds.
- filteroutairportsnoresults: 347.208947ms seconds.
- countwithfilter_airports: 349.70095ms seconds.
- filteroutairportsincountry 354.040548ms: seconds for filtering 57265 records