Working with BigNumber and different decimals

[austinbv]

Hey All,

I am trying to figure out the best way to do math on numbers that have different precision.

There's a contract I interact with that returns

• Price of a token in a BigNumber with 18 decimals price
• Balance of the token as a BigNumber with different decimals balance
• Decimals as a number used by the notate the number decimals the balance uses

In python or an app that has big number support I would do something like

balance * price / 10 ** decimals

With BigNumber doing this I lose significant figures if I do something similar. Also in JS I am doing floating point math which is bad...

The example I have is a balance of 50905607240 tokens in a decimal of 10 at a price of 49656985639362028180831 in a Decimal of 18

How would you get the most precise value of the pool using ethers BigNumber.

I have tried using parseUnits but because the balance comes is as a string it's something like this

parseUnits(balance.toString(), decimals).mul(price)

But that number is wrong... I feel like I am just throwing things against the wall, including using toTwos and the works...

Any ideas would be helpful

[ricmoo]

If you only care about the truncate whole value, the math should work fine, balance.mul(price).div(ethers.constants.WeiPerEther). The order is important. Are you sure you want to divide though? If you do unit analysis, are you targeting your price in ether or in wei?

You can also use the ethers.FixedNumber class to work things through and see where the fractional bits are.

[austinbv]

First thanks for your help!

Interesting. I am not sure I understand what you're getting at.

The contract is the tricrypto pool on curve. It returns 3 assets, and 3 pieces of info about each asset. Price (USD per token of the asset), Balance and Decimals I want to get the total value (USD) of the pool.

Asset	Price	Balance	Decimals
Dollar	1	23980063263820051878265489	18
WBTC	46080446917336126955087	51509502100	10
WETH	3400831264281950775263	6958168492774986022522	18

the formula I would use in pseudo code is

Sum of each assets (price * balance * 10 ^ decimal)

That would end up looking something like this

dollar = 1 * (23980063263820051878265489 / 10 ^ 18)
wbtc = 46080446917336126955087 * (51509502100 / 10 ^ 10)
weth = 3400831264281950775263 * (6958168492774986022522 / 10 ^ 18)

total = dollar + wbtc + weth

But that uses floating-point math... I would prefer to never work with decimals and only work with integers converting to decimals only when displaying to a user... if possible

[zemse]

When doing arthimetic for FixedPoint decimals you have to first convert them to same decimals. Since you are normalizing everything in dollars, it makes sense to use the dollar's decimals.

const btc_price = ethers.BigNumber.from('46080446917336126955087') // decimals 18
const btc_bal = ethers.BigNumber.from('51509502100') // decimals 8
const weth_price = ethers.BigNumber.from('3400831264281950775263') // decimals 18
const weth_bal = ethers.BigNumber.from('6958168492774986022522') // decimals 18
const ten = ethers.BigNumber.from('10');

const btc_in_dollar = btc_price.mul(btc_bal).div(ten.pow(btc_price_decimals + btc_bal_decimals - dollar_decimals)) // 18 + 8 - 18
const eth_in_dollar = eth_price.mul(eth_bal).div(ten.pow(eth_price_decimals + eth_bal_decimals - dollar_decimals)) // 18 + 18 - 18

ethers.utils.formatUnits(btc_in_dollar.add(eth_in_dollar), dollar_decimals); // should give you desired amount

In this example dollar_decimals is 18, which is >= both the decimals of btc and eth, that's why there was no underflow/precision loss. If you had to use 6 (as USDC, USDT have that), then it would cause the precision loss. For preventing underflow, you can select max of all decimals.

[austinbv]

You're a saint. This works great. It's essentially what I was doing in my comment below but I like normalizing the dollar decimals out.

For preventing underflow, you can select max of all decimals.

Out of curiosity - how would you go about doing this?

[zemse]

For preventing underflow, you can select max of all decimals.

Out of curiosity - how would you go about doing this?

When you are working with BTC (8 decimals), ETH (18 decimals), SHIB (27 decimals), then to keep all the precision you can use the resultant decimals as 27. This may only be useful on the front end if you want to display a very accurate answer.

However in the smart contract you cannot do that, for example if you are working with USDC (decimals as 6), then you cannot transfer a value smaller than 10^-6 and greater than 0. E.g. if the calculation with the above max decimals amounts to 1.123456789 dollars, then you have to choose to transfer either 1.123456 USDC or 1.123457 USDC.

[austinbv]

I want to add my partial solution but I was hoping there is a more elegant way with ethers

Which is simpler in utils.BigNumbers

formatUnits(balance.mul(price), (18+decimal))

EDIT: If you use darkmode you cannot see the formula... Bummer

[austinbv]

The issue with the above solution is I cannot do any other math on it