Minimum reserves check

[JanKuczma]

No description provided.

[deuszx]

Let's add some tests around this. Ideally we'd test that no matter the decimals of the pool tokens the minimum reserves constraint makes sense.

[JanKuczma]

Let's add some tests around this. Ideally we'd test that no matter the decimals of the pool tokens the minimum reserves constraint makes sense.

I added some tests. Ideally, these tests will be refactored appropriately in #17.

[deuszx]

Let's add some tests around this. Ideally we'd test that no matter the decimals of the pool tokens the minimum reserves constraint makes sense.

I added some tests. Ideally, these tests will be refactored appropriately in #17.

I hope not in #17 as it's already huge as hell

[deuszx]

As discussed on Slack – let's add tests for removing liquidity.

[deuszx]

I left comments under commit (f24facb#comments) – do you think you can take a look there? ;)

LGTM otherwise 💪

Alternative would be to write a proptest with proper generators but given the amount of tests now that's probably unnecessary.

[woocash2]

Potential problem

I see that MIN_RESERVE is set to TARGET_PRECISION which means that for any token, we require at least 1 full unit of that token to be present in the pool at any time, after the first liquidity deposit.

This doesn't look nice to me, considering tokens whose $ value is very high. Take for example that we or someone else wants to create a pair which contains BTC, for example (BTC, stakedBTC) (I don't know if such a pair makes sense, but it's the best example I could think of). MIN_RESERVE implies that:

• the first person making a deposit has to be pretty rich - they have to invest at least 120K$ into this pair
• at least 1 BTC and 1 stakedBTC have to exist in the pool forever, which means that effectively ~120K$ is "irrecoverable" by this pool.

ETH is a less dramatic example as it costs ~3000$, but still, (ETH, stakedETH) pool makes around 6000$ "irrecoverable".

Questions

• Are trades impossible in the case of reserves (X, 0) or (0, 0), i.e. will contract panic (abort), or will the user just get 0 profit, or giga profit (depending on the trade direction), but contract call will be successful? This means: is reserve[i] >= MIN_RESERVE required for the pool to work without error?
• Have you considered at least making the MIN_RESERVE smaller than 1 unit? I suppose that the only reserve values which prevent the contract from working are 0. So why not MIN_RESERVE = 1 or some other less significat value than 1 unit?