lang: remove space inference via derive(Default)

[paul-schaaf]

• inefficient (it calls Default::default and then len)
• useless for dynamic types

[Henry-E]

Neither of these things seem like sufficient reason to remove space inference from derive(Default)

• it's inefficient but it's a quality of life improvement, assuming you're not using any dynamic types
• Dynamic types needing manual space definitions is fine, it's just a byte per u8 or character in a string

Is this being replaced by any way to manually call this? Or some other way to simplify the calculation of how much space is needed. Like should we start adding LEN functions to all of our accounts structs now.

(this comment is probably coming too late but thought it worth responding anyway)

[paul-schaaf]

Dynamic types needing manual space definitions is fine, it's just a byte per u8 or character in a string

agree that this is not a big deal

but

it's inefficient

this is.

we want anchor to be as much of a zero-cost abstraction framework as possible and serializing an entire type at runtime just to get its static length is definitely far from that.

Like should we start adding LEN functions to all of our accounts structs now.

const variables (see the tests in this PR)

One thing Ive thought about is a #[len] macro that reads the types and writes the const variable. I'll create an issue for a discussion on this.

[paul-schaaf]

#1647

[Henry-E]

Just how inefficient is it? There's quite a bit of mental overhead to calculating the size of different accounts and then keeping them up to date. It's sort of the equivalent of how comments and their underlying code sometimes stray from each other.

That said, I like that there's discussion of alternative solutions and I will comment there too. Thanks for listening!

[ceciEstErmat]

Also, this seems to me to be a tool. If you need the efficiency then use space, if you do not care and want the quality of life then use Default.
Nothing in the framework enforced its use.

[paul-schaaf]

pre txwide compute unit changes I wouldve agreed but now the efficiency of your instruction affects the efficiency of the entire transaction. Therefore, all instructions should be as efficient as possible.

But if you still want to use default to calculate the size, you can do it manually still