Summary

This is a bugfix as inner transactions weren't actually being saved to the database after #5478 . The main problem was that only ledger.AddBlock() was used to persist blocks which were missing ApplyData. This function doesn't modify the block which it is adding (it can evaluate a different block for validation purposes though). The fix works by using logic similar to the transaction pool and the recent simulator:

1. use eval.StartEvaluator() with generate = true to instantiate a block evaluator
2. call eval.TransactionGroup() on each group that the generator provides
3. call eval.GenerateBlock() at the end to create a validated block which contains the evaluated ApplyData with inner transactions
4. call ledger.AddValidatedBlock() to persist.

Additional changes have been hilighted in my comments below.

TODO's Carried Over

The ones that haven't yet been done should be carried over to Part 3: swaps

• remove all explicit ApplyData settings/params as Ledger.Eval() should be in charge of this now and all existing references are empty (generator.go)
• the values of the effects map should be slices of a type different from TxTypeID (DONE: now they're plain strings)
• the generator only uses effects map to update txnCounter (or intra) NOPE: this is now being updated from the block's evaluator directly. However, there is still an assertion that the number of expected transactions is what the evaluator provided.
• there is a unit test that asserts the total number of transactions for each TxTypeID with effects is exactly as expected (DONE: see TestEffectsMap )
• the effects are only used at report time in run.go to split out the various effects
• introduce generate_apps.go for application specific generator code
• introduce generator_ledger.go for ledger specific generator code
• return the number of transactions generated instead of updating intra (generator.go)
• README improvements:
  • incorporate spirit of the comment below explaining the complexity of getActualAppCall()
  • an in-depth explanation of how the block generator works
• need to update explanatory comment for the effects map in the next PR for because of the new teal files and updated names
• "these may not make sense for the swap optin" (make_transactions.go)
• add sanity checks similar to the one for the case appTxTypeCreate for the other cases (generateAppCallInternal())
• figure out if there is some simple code organization to help with handling the dependencies between states of applications. Maybe we can plan for some larger architecture changes in the future (i.e. using the ledger, it wasn't available in earlier versions of this tool).
• remove type appData struct field kind if still unused after Part 3 (generator_types.go)
• lines 319-334 of generate.go should become: vBlock, ledgerTxnCount, err := g.generateBlock()
• OR: generating the tx groups is part of generating the block, so could push 319-340 into g.evaluateBlock, and rename evaluateBlock to generateBlock
• rand.Seed has been deprecated since Go 1.20 and an alternative has been available since Go 1.0: Programs that call Seed and then expect a specific sequence of results from the global random source (using functions such as Int) can be broken when a dependency changes how much it consumes from the global random source. To avoid such breakages, programs that need a specific result sequence should use NewRand(NewSource(seed)) to obtain a random generator that other packages cannot access. (SA1019)

Test Plan

In addition to some new unit tests, I've tested it locally using the same method described on #5478

FAQ

Q: How to calculate the number of inner transactions in the database?

A:

select count(*) from txn where extra->'root-txid' is not null;