Populate memo in blockstore signatures-for-address

[CriesofCarrots]

Problem

cc #19512 Payment providers would like to use memos to annotate transactions, but wallets can't easily display these memos in the tx-history list. ConfirmedTransactionStatusWithSignature and RpcConfirmedTransactionStatusWithSignature contain memo fields for this purpose, but these are not being populated.

Summary of Changes

Store memos in new blockstore TransactionMemos column via the TransactionStatusService, and use it to populate the memo field for getSignaturesForAddress response objects that are queried from blockstore.

This will require a stubbed TransactionMemos column in v1.6, or backporting both memo-population PRs to v1.6. Preference?

I also considered the alternate solution of a new AddressSignatures column to store the memo alongside each address, similar to the schema in bigtable, but ultimately decided that the very small perf savings in Blockstore::get_confirmed_signatures_for_address2() are not worth the additional disk to duplicate memos for every address. (However, it may still be worth plumbing a new AddressSignatures column sometime in the future to fix the column key to return signatures in proper execution order.)

[codecov]

Codecov Report

Merging #19515 (e27024b) into master (c550b32) will decrease coverage by 0.2%.
The diff coverage is 39.9%.

@@            Coverage Diff            @@
##           master   #19515     +/-   ##
=========================================
- Coverage    82.7%    82.4%   -0.3%     
=========================================
  Files         461      468      +7     
  Lines      131215   131675    +460     
=========================================
+ Hits       108519   108628    +109     
- Misses      22696    23047    +351     

[mvines]

This will require a stubbed TransactionMemos column in v1.6, or backporting both memo-population PRs to v1.6. Preference?

A full backport seems worthwhile as this looks low risk and gets us nice memos in RPC sooner

However, it may still be worth plumbing a new AddressSignatures column sometime in the future to fix the column key to return signatures in proper execution order

Can you please remind me of the limitation here? Signatures from the same block aren't returned in chronological entry order?

[CriesofCarrots]

Can you please remind me of the limitation here? Signatures from the same block aren't returned in chronological entry order?

@mvines , yes, that's right. The key on the AddressSignatures column is (unused_u8, Pubkey, Slot, Signature, which means that within each slot, the signatures are stored and return in order of Signature bytes. A fixed AddressSignatures column would be keyed: (Pubkey, Slot, SlotIndex, Signature)

Pull request has been modified.

[mvines]

Can you please remind me of the limitation here? Signatures from the same block aren't returned in chronological entry order?

@mvines , yes, that's right. The key on the AddressSignatures column is (unused_u8, Pubkey, Slot, Signature, which means that within each slot, the signatures are stored and return in order of Signature bytes. A fixed AddressSignatures column would be keyed: (Pubkey, Slot, SlotIndex, Signature)

Thanks. This may become an issue with the confidential token transfer work that we're looking at, I'll make a note and we can revisit it soon-ish if so!