test: update user-agents data, update dev deps

[antongolub]

Maintenance PR to make sure the library supports newly added user agents

[juliangruber]

I don't think we need to cache the UserAgent instances here

[juliangruber]

Also, I'm having trouble looking up the UserAgent API, do you know where its repo is located now?

[antongolub]

This just follows the guide: https://github.com/intoli/user-agents#useragentrandom. random() generates new instances too, but also reuses filtered data preset and so that's «over 100x faster than the initial construction».

[juliangruber]

ah sorry I thought we were using user-agent not user-agents

[juliangruber]

Are the tests too slow without this optimization?

[antongolub]

I can simplify this piece a bit to make it clearer. I'm a nerd. I do optimizations by default)

[juliangruber]

premature optimization is the root of a lot of evil :P

[juliangruber]

Thank you for submitting this PR!

[juliangruber]

this is still caching the user agent instance 🤔

[antongolub]

It caches the data preset which is attached to the instance. There's no other way to bind filtered values: setCumulativeWeightIndexPairs declared as module scope variable.
https://github.com/intoli/user-agents/blob/master/src/user-agent.js#L140

 random = () => {
    const userAgent = new UserAgent();
    setCumulativeWeightIndexPairs(userAgent, this.cumulativeWeightIndexPairs);
    userAgent.randomize();
    return userAgent;
  };

Authors literally suggest using this approach:
https://github.com/intoli/user-agents#useragentrandom

// Use the `random()` method to construct a second user agent.
const firstUserAgent = new UserAgent(filters);
const secondUserAgent = firstUserAgent.random();

[juliangruber]

I'm not against using .random(), I'm against caching UserAgent instances in ua[...]

[antongolub]

I've just noticed new Proxy inside the UserAgent constructor. Optimizations don't make any sense)

[juliangruber]

hey, thanks for removing the caching on the ua function, why did you also remove .random()? I thought that was good to add?

[antongolub]

random is just a constructor-copy-like factory exposed as a proto method. It requires a ua reference to bind its this.cumulativeWeightIndexPairs to new ones.

[antongolub]

Finally, I've replaced ua factory memoization with an in-iterator scope variable.

[juliangruber]

ok, I think just one last change, could you refactor this data structure to be an array of objects, like [{ family: 'mobile', tablet: false, ...}]? This way it's easier to read

[antongolub]

np

[juliangruber]

perfect, thank you so much!