josdejong
Question about performance compared to JSON.parse
Hi community
I have a question and unfortunately didn't find other place to ask.
So, could someone explain, why native JSON.parse or Response: json() method are working so slower than parse method of lossless-json?
I created example to compare. First of all you need to take big JSON. I used about 100Mb file. You can use this one if you want Then play here https://codepen.io/MaevskiyE/pen/vYwYPEq and compare parsing time
The lossless-json parser is actually slower than the native, built-in JSON.parse, and that is also what I would expect.
In your benchmark it will be easier to compare duration when you use console.time and console.timeEnd BTW :).
@josdejong , thank you for response. In most cases maybe lossless-json parser slower that native JSON.parse. But in my case with a lot of numbers - it's significantly faster. 2-3sec vs 21sec. As I understood it's because lossless-json parser even doesn't try to convert value to actual number and leave it as string. Can you confirm my assumption? Maybe you can suggest a way, how to make the same with JSON.parse using it's second argument reviver? Thank you
You have to use console.time and console.timeEnd like:
console.time('parse-my-thing')
// ... do the work
console.timeEnd('parse-my-thing')
I've run your example test.json file. In Firefox I get:
// Firefox, parsing 100 MB test.json file
JSON.parse: 443ms - timer ended
lossless-json: 1471ms - timer ended
Which is what I expect and have seen all the time, lossless-json being 2-5 fives slower than the built-in JSON.parse.
On Chrome however I see the issue you encounter:
// Chrome, parsing 100 MB test.json file
JSON.parse: 7042.577880859375 ms
lossless-json: 978.43701171875 ms
I've tested again with other JSON files (like example from here), I get the same, expected behavior of Chrome and Firefox both having the native implementation being faster than lossless-json (as expected). So it looks like your very JSON file has something specific that triggers a bad performance in Chrome. When I change the longitude and latitude properties from strings into numbers, the performance issue of Chrome is gone and it is faster than lossless-json (as expected).
@josdejong , I very appreciate your help. This really helped! You saved my life :)
p.s. now it's interesting why Chrome has problems with numbers like String and Firefox doesn't