feat(osmomath): mutative and efficient BigDec truncations with arbitrary decimals

[p0mvn]

Background

Currently, our SDKDec() method of BigDec does a lot of reallocations. Additionally, this function only truncates to sdk.Dec with precision of 18.

Suggested Design

We should have a way to truncate an arbitrary number of digits. The number of reallocations should be minimized. Mutative and non-mutative versions should be implemented.

See chopPrecisionAndTruncate function for reference. Also, see how chopPrecisionAndRoundUpNonMutative and chopPrecisionAndRoundUp are implemented. Clean up and enhance abstractions where applicable.

Acceptance Criteria

• Implement arbitrary decimal truncation
• Implement mutative and non-mutative versions
• Reallocations minimized
• Refactor SDKDec() to utilize these new functions to minimize reallocations while preserving state-compatibility.
• Test coverage maximized
• Optimize this to be the first client of efficient variable truncation: refactor(CL): convert CalculatePriceToTickDec to operate on BigDec conversions in a state-compatible way #6256 (comment)

[pysel]

hey @p0mvn. I have a question on this: wdym mutative and non-mutative versions? do you mean SDKDec should mutate it's receiver parameter?

[p0mvn]

Hey. For reference, please see Mul and MulMut methods defined on the sdk.Decimal

The benefit of the mutative version is that it doesn't reallocate and mutates the receiver directly. However, it cannot be used always. As a result, we need two versions for most math methods

[pysel]

in cases of mutative Mul, Quo, Add, and etc, both receiver and the return value have same types (sdk.Dec in case of Mul, for example). in the case of SDKDec function, the receiver is of type BigDec and the output is sdk.Dec, hence, we cannot set the latter to the former.

we could set the underlying *big.Int of the latter to the former, but I suppose that it breaks the point of osmomath.BigDec -> sdk.Dec conversion. if this is the desired approach, please, lmk! thanks

cc: @p0mvn

[p0mvn]

osmomath.BigDec is 36 decimals. sdk.Dec is 18 decimals.

To make the mutative SDKDec version defined on the osmomath.BigDec, we need to make it reuse the first 18 decimals of the original buffer and truncate the rest.

There should be a warning made that it makes the original osmomath.BigDec unusable