STM retry/orElse semantics

[durban]

This issue is a case of "behaves differently from Haskell", so I'm not even sure this is a bug. (Although the Haskell behavior seems more useful to me.)

Apparently, Haskell's STM retries a transaction if any of the variables read in any branch of an orElse change. (See section 3.4 here.) It seems ZSTM doesn't do this: it only retries if the TRef read by the RHS/second transaction changes. For example, this program hangs:

    for {
      ref1 <- TRef.makeCommit(0)
      ref2 <- TRef.makeCommit(0)
      txn1 = ref1.get.flatMap {
        case 0 => STM.retry
        case n => STM.succeed(n)
      } orElse ref2.get.flatMap {
        case 0 => STM.retry
        case n => STM.succeed(n)
      }
      txn2 = ref1.set(1) // if this is `ref2.set(1)`, it doesn't hang
      fib <- txn1.commit.forkDaemon
      _ <- ZIO.sleep(zio.durationInt(1).second)
      _ <- txn2.commit
      r <- fib.join
    } yield ()

[durban]

Based on this part of the documentation, it probably should work like haskell. And note, that for orTry it does. So this seems like a bug in orElse.

[jdegoes]

/bounty $250 for fix and test to accompany it.

[algora-pbc]

💎 $250 bounty • ZIO

Steps to solve:

1. Start working: Comment /attempt #8696 with your implementation plan
2. Submit work: Create a pull request including /claim #8696 in the PR body to claim the bounty
3. Receive payment: 100% of the bounty is received 2-5 days post-reward. Make sure you are eligible for payouts

Additional opportunities:

• 🔴 Livestream on Algora TV while solving this bounty & earn $200 upon merge! Comment /livestream once live

Thank you for contributing to zio/zio!

Add a bounty • Share on socials

Attempt	Started (GMT+0)	Solution
🟢 @BijenderKumar1	May 11, 2024, 6:17:07 PM	#8845

[BijenderKumar1]

/attempt #8696
I reproduced the issue and found that the changes in the LHS (first branch) for orElse do not trigger a retry of the transaction, as pointed out. However, it works correctly with orTry instead of orElse. I dived deep into the ZSTM.scala file where the orElse function is defined. It seems that the issue is with the prepareResetJournal function, which resets the journal to empty each time it is invoked, and therefore, any changes in the first branch are not triggering a retry. I will be fixing the issue and raising the PR soon. Thanks!

Options

• Cancel my attempt

[algora-pbc]

💡 @BijenderKumar1 submitted a pull request that claims the bounty. You can visit your bounty board to reward.

[algora-pbc]

🎉🎈 @BijenderKumar1 has been awarded $250! 🎈🎊

• Share on socials
• Give feedback