[style] Rename trait type names from I$Name to $Name

[TennyZhuang]

Summary

E.g: rename ITable to Table. There is no point to add I prefix before a trait.

In rust, A trait type has two usages:

1. Template argument constrait: like Where T: Table, we can know Table is a trait as soon as we see it.
2. Trait objects: like Box<dyn Table>, dyn trait requires exiplit dyn keyword, there is also no confusion at all.

For the reason above, trait named with I$Name is meaningless.

[sundy-li]

This is from golang and cpp style, I vote for this change since it's meaningless to keep the prefix I for Trait in rust.

cc @drmingdrmer

[drmingdrmer]

For me, the greatest benefit is knowing at first sight what kind of symbol it is.

(Maybe) makes it easier to search for something through plain text, e.g., when you do not know the exact name of it.

(Maybe) the prefix "I" can reduce name conflicts.

Having one more letter does not bother me, although this is not very rust.

Vote for keeping it :DDD

[dantengsky]

I prefer not to use the I prefix on traits, otherwise, It will be unfair to Enum 💔

[BohuTANG]

Thank you very much for this suggestion, from rustacean view should not have I.
But I prefer to keep I:
For the people who reading the codes and not familiar with rust, at least this is more intuitive.

Vote for keeping I.

[leiysky]

I vote for the idiomatic style Table. 🚀

[tisonkun]

Vote for this change. I$name is from Hungarian notation, which is really redundant for modern development environment.

Type is explicit or can be easily identified by "find definition". We have advanced too much from text only ages.

[BohuTANG]

Wow, votes status:
rename: 4
keep: 3

But datafuse have some conflicts after rename:
trait IDataSource
struct Datasource

If we rename IDataSource to DataSource, how to deal the struct name?

[dantengsky]

Remove IDataSource, use struct DataSource directly? (replace Arc<dyn IDatasource> with Arc<DataSource>).

Till we have multiple impls for an "IDataSource", and at that time we will have different names for impls naturally.

[drmingdrmer]

IMHO here the discussion should not be about which style is better, but be about whether it's a good time to apply this change.

Changing it brings in extra refactoring work and name/type conflict.
Leaving as it is makes nothing better or worse.

[dantengsky]

If we do wanna follow the naming convention or "guideline":

The sooner the better.

With the supports of IDE (and some manual namespace tweaking), it's no big deal :

8c7888a

[BohuTANG]

If we do wanna follow the naming convention or "guideline":

The sooner the better.

With the supports of IDE (and some manual namespace tweaking), it's no big deal :

8c7888a

It's ok to me.
Just do it.

[dantengsky]

Roger that

[TennyZhuang]

Hi, any update here?

[BohuTANG]

I guess @dantengsky PR can be opened again at any time

[dantengsky]

@TennyZhuang @BohuTANG I'd like to re-open the PR (rebased with master branch) later today.

[BohuTANG]

Cool, Thanks.
Let's more rustacean!

[jyizheng]

Sound great to me to follow the guideline.