refactor: later requires in constructor, for readability

[adhusson]

Costs more gas on failed contract creations but much easier to read.

[QGarchery]

No strong feeling on this one

[peyha]

My initial implementation also had the require later in the constructor, no feeling on this one for me neither

[colin-morpho]

Same, I agree that names are shorter as we don't have to go through structs. On the other hand, it feels weird to compute (i.e. the constructor body) stuff that may fail just afterwards. What's your take on that @adhusson ?

[colin-morpho]

As a compromise I can only imagine using shorter names for the constructor parameters like _preLiquidationParams -> preLiqParams/pLP. But, I find these less readable actually.

[MathisGD]

I agree it's more readable, although I'm generally against using immutable values in the contructor, as it's not sure if they are initialised yet or not. Also I like to have the require at the top of functions.

An other possibility would be to remove the struct (would be curious to see the code without it)

[adhusson]

Tried here I'm not fond of it.

[colin-morpho]

I agree with you @adhusson.

[adhusson]

generally against using immutable values in the contructor, as it's not sure if they are initialised yet or not.

Are you referring to this part of the via_ir spec? It is a difficult area but I don't think it applies here.

[MathisGD]

My concern is that you could do something like

require(ABC ...);

ABC = abc;

without noticing

[adhusson]

On the other hand, it feels weird to compute (i.e. the constructor body) stuff that may fail just afterwards. What's your take on that @adhusson ?

There's always a simulation first, and the parameters do not depend on things that will evolve between the simulation and the actual tx inclusion. So I think there will be ~no failed onchain deployments of this contract, the wasted gas does not matter, and we can focus on readability.

[QGarchery]

Suggested change

	require(IMorpho(morpho).market(id).lastUpdate != 0, ErrorsLib.NonexistentMarket());
	MarketParams memory _marketParams = IMorpho(morpho).idToMarketParams(id);
	MORPHO = IMorpho(morpho);
	require(MORPHO.market(id).lastUpdate != 0, ErrorsLib.NonexistentMarket());
	MarketParams memory _marketParams = MORPHO.idToMarketParams(id);

Should we also do something like this then ?

[colin-morpho]

It would be more consistent, right?
Here's the take of ChatGPT on this matter
https://chatgpt.com/share/e/66fe5eb5-9364-800e-9d25-a2ff7dbc1bf6

[colin-morpho]

I'm amazed by it's ability to handle pseudo code.

[MathisGD]

Still not a big fan tbh. If there is no majority for it, i would say we close it.