Update go-graphsync v0.8.0

[hannahhoward]

Goals

Good news: The latest version of go-graphync and go-ipld-prime contains an important optimization to remove a bunch of unneccesary hashing when reading from local blockstores, which should cut down significantly on CPU usage
Bad news: The latest versions of these libraries contain a some important, but very breaking changes to syntax.
Good news: I'm updating them all the way to Lotus! And also, in most cases, from the point of view of the consuming libraries, they're simplifications.

Implementation

• Update to go-graphsync v0.8.0 with go-ipld-prime linksystem branch & trusted store.
• Fix syntax changes which are mostly to:
  • dagcbor Encode / Decode function name
  • use of linksystem over loader + storer.

Background

• More about the LinkSystem: introduce LinkSystem ipld/go-ipld-prime#143
• Trusted store branch: Add option to tell link system storage is trusted and we can skip hash on read ipld/go-ipld-prime#149
• PR to merge LinkSystem into go-graphsync: Update for LinkSystem ipfs/go-graphsync#161

[codecov-commenter]

Codecov Report

Merging #212 (e4f40bc) into master (db84ad0) will increase coverage by 0.11%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #212      +/-   ##
==========================================
+ Coverage   64.84%   64.96%   +0.11%     
==========================================
  Files          24       24              
  Lines        2600     2600              
==========================================
+ Hits         1686     1689       +3     
+ Misses        573      571       -2     
+ Partials      341      340       -1     

Impacted Files	Coverage Δ
message/message1_1/transfer_message_cbor_gen.go	34.52% <0.00%> (ø)
message/message1_1/transfer_request_cbor_gen.go	27.17% <0.00%> (ø)
message/message1_1/transfer_response_cbor_gen.go	29.68% <0.00%> (ø)
channels/channel_state.go	80.55% <100.00%> (ø)
encoding/encoding.go	60.00% <100.00%> (ø)
message/message1_1/transfer_request.go	76.36% <100.00%> (ø)
transport/graphsync/graphsync.go	77.48% <100.00%> (ø)
impl/receiver.go	67.10% <0.00%> (-1.32%)	⬇️
network/libp2p_impl.go	72.26% <0.00%> (+3.36%)	⬆️

---

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 db84ad0...e4f40bc. Read the comment docs.

[tchardin]

@dirkmc what is the timing for this? Unfortunately we cannot use the latest go-data-transfer release because it relies on changes made to an older version of graphsync that is incompatible with ipld-prime v0.7.0 used in our project. Cannot run it from the commit either since there are breaking changes with go-fil-markets. So sadly we can't benefit from the changes you made until this gets in. Thanks!

[dirkmc]

@tchardin we've been doing some performance testing of graphsync v0.8.0. You can follow the progress at #166

It seems like there may need to be further optimizations made before a newer version of graphsync can be integrated into go-data-transfer.

I'd suggest you align your code with the latest release of go-data-transfer v1.7.0 which uses graphsync v0.6.4

[tchardin]

Gotcha, thanks @dirkmc. So you confirm it would be faster to downgrade our project to ipld-prime v0.5.1 instead of waiting for this to get in?

[dirkmc]

I think that's probably your best bet 👍

[hannahhoward]

superseded by #246