Crate cleanups

[str4d]

No description provided.

[Eirik0]

Everything looks great! The only thing I have not reviewed is the addition of log = "0.4".

[Eirik0]

Is there a benefit to dereferencing here? I am wondering if this is not necessary, as of Rust 1.26: https://rust-lang.github.io/rfcs/2005-match-ergonomics.html

[str4d]

The clippy warning that this change addressed is:

warning: you don't need to add `&` to all patterns
   --> bellman/src/gadgets/boolean.rs:415:9
    |
415 | /         match self {
416 | |             &Boolean::Constant(c) => Some(c),
417 | |             &Boolean::Is(ref v) => v.get_value(),
418 | |             &Boolean::Not(ref v) => v.get_value().map(|b| !b),
419 | |         }
    | |_________^
    |
    = note: #[warn(clippy::match_ref_pats)] on by default
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#match_ref_pats
help: instead of prefixing all patterns with `&`, you can dereference the expression
    |
415 |         match *self {
416 |             Boolean::Constant(c) => Some(c),
417 |             Boolean::Is(ref v) => v.get_value(),
418 |             Boolean::Not(ref v) => v.get_value().map(|b| !b),
    |

Dereferencing is still necessary here:

$ git diff
diff --git a/bellman/src/gadgets/boolean.rs b/bellman/src/gadgets/boolean.rs
index b26bb19..dafa82f 100644
--- a/bellman/src/gadgets/boolean.rs
+++ b/bellman/src/gadgets/boolean.rs
@@ -412,7 +412,7 @@ impl Boolean {
     }

     pub fn get_value(&self) -> Option<bool> {
-        match *self {
+        match self {
             Boolean::Constant(c) => Some(c),
             Boolean::Is(ref v) => v.get_value(),
             Boolean::Not(ref v) => v.get_value().map(|b| !b),
$ cargo build
   Compiling bellman v0.1.0 (/.../librustzcash/bellman)
error[E0308]: mismatched types
   --> bellman/src/gadgets/boolean.rs:416:42
    |
416 |             Boolean::Constant(c) => Some(c),
    |                                          ^
    |                                          |
    |                                          expected bool, found &bool
    |                                          help: consider dereferencing the borrow: `*c`
    |
    = note: expected type `bool`
               found type `&bool`

error: aborting due to previous error

For more information about this error, try `rustc --explain E0308`.
error: Could not compile `bellman`.

To learn more, run the command again with --verbose.
$ rustc --version
rustc 1.36.0 (a53f9df32 2019-07-03)

[Eirik0]

Is there a distinction between to_bits_ and into_bits_? I am comparing this to the name change in bellman/src/gadgets/num.rs.

[str4d]

The clippy warning that this change addresses is:

warning: methods called `into_*` usually take self by value; consider choosing a less ambiguous name
  --> bellman/src/gadgets/uint32.rs:78:25
   |
78 |     pub fn into_bits_be(&self) -> Vec<Boolean> {
   |                         ^^^^^
   |
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#wrong_self_convention

This follows the corresponding entry in the Rust style guide.

I altered the various methods to be consistent with how they used self in context. I first tried having the into_* methods consume self, and where that was not possible due to the caller relying on passing an immutable reference, I altered the method name to to_*.

[Eirik0]

This change is also made when running cargo fix --edition-idioms.

[Eirik0]

This could be replaced with

// Mask away the unused most-significant bits.
let tmp = Fs(FsRepr([
    rng.next_u64(),
    rng.next_u64(),
    rng.next_u64(),
    rng.next_u64() & (0xffffffffffffffff >> REPR_SHAVE_BITS),
]));

And then we wouldn't need the loop below, but that may be beyond the scope of this PR.

[str4d]

I added log to the dependecy tree in order to provide distinct branches inside equihash::validate_subtrees, otherwise clippy complained that the three branches of the Equihash conditional were identical. We could alternatively (or additionally) set up an Error type instead of using bool. I think log is an independently beneficial crate to depend on, particularly if we want to move more zcashd logic into Rust - we should be able to log from the Rust side into zcashd's debug.log. (Implementing the connecting logic in librustzcash would require a separate PR.)

[ebfull]

Looking good, will re-review when #113 is merged since it will conflict with this I think.

[str4d]

Rebased on master to fix merge conflicts with #113.

[Eirik0]

utACK. Rebase looks good.

[defuse]

utACK. I left some non-blocking comments.

[defuse]

Rust noob question: Is it correct that I can be sure this isn't modifying any part of params because it isn't &mut in the function signature? (I've been making this assumption in a couple cases above where an explicit .clone() was removed, too.).

[str4d]

JubjubParams::montgomery_a takes &self, so you are guarateed that params is not modified.

The reason for removing the clone() is that JubjubParams::montgomery_a returns JubjubEngine::Fr (which resolves to ff::ScalarEngine::Fr), which via ff::Field is required to implement Copy, and thus makes the explicit clone unnecessary.

[defuse]

Previously this loop's body would execute 1 << log_new_n times. With this change it should execute num_cpus times according to the length of tmp when it's created above, right?. Won't that change the behavior?

Oh, I see tmp, is shadowed by the outermost loop so here it will have length 1 << log_new_n and the behavior is unchanged. It might be worth renaming some variables to avoid similar confusion.