Retransmit

[aeyakovenko]

• adds a subscriber data structure for layer 0 and 1 of the network (layer 2 is the everyone else)
• retransmit service for resending the blobs we receive from the leader

[codecov-io]

Codecov Report

Merging #129 into master will increase coverage by 0.22%.
The diff coverage is 96.95%.

@@            Coverage Diff             @@
##           master     #129      +/-   ##
==========================================
+ Coverage   96.16%   96.39%   +0.22%     
==========================================
  Files          17       18       +1     
  Lines        1671     1886     +215     
==========================================
+ Hits         1607     1818     +211     
- Misses         64       68       +4

Impacted Files	Coverage Δ
src/packet.rs	95.41% <ø> (+1.25%)	⬆️
src/lib.rs	100% <ø> (ø)	⬆️
src/streamer.rs	96.71% <95.86%> (-1%)	⬇️
src/subscribers.rs	98.82% <98.82%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 83c5b3b...29f3230. Read the comment docs.

[garious]

Can you add a comment that explains why you can't use self.data.append(ns) here?

[aeyakovenko]

--> src/subscribers.rs:77:26
|
77 | self.data.append(ns);
| ^^ types differ in mutability

[aeyakovenko]

seems like it wants to pop from ns

[garious]

What's really strange here is to both see into_iter() and n.clone(). Ideally, you'd just remove that .clone(), but if there's something preventing that from working, then I'd expect to see &ns instead of ns.into_iter().

[aeyakovenko]

    self.data.extend_from_slice(ns);

!!!

[garious]

Can you add docs to this function?

[garious]

What are these variable names supposed to mean?

[garious]

What is a cast?

[garious]

What's doing retransmitting exactly? Seems more like a proxy.

[garious]

The rest of the codebase consistently uses this form:

Node { id, weight, addr }

[aeyakovenko]

fmt output

[garious]

I only pub the fields when I'm being lazy. Looks like you've done a good job of adding methods for this data. Can you remove the public access? It makes refactoring a little easier and gives us nice warnings if a field isn't used.

[garious]

If you make this &[&h.me, &h.leader] can you remove the clones above()?

[aeyakovenko]

its the most tiny amount of data

[aeyakovenko]

error[E0502]: cannot borrow `h.me` as immutable because `h` is also borrowed as mutable
  --> src/subscribers.rs:54:21
   |
54 |         h.insert(&[&h.me, &h.leader]);
   |         -           ^^^^            - mutable borrow ends here             
   |         |           |                                                                          
   |         |           immutable borrow occurs here                                                                                                         
   |         mutable borrow occur

it generally doesn't like borrowing two fields from a structure

[aeyakovenko]

this needs a copy, since we keep the node in the subscriber vector, and part of the structure

[garious]

Great edits, thanks a bunch. Merging.