Hi @m-kus I would like to take this issue

Happy to work on that one!

Hey @baitcode! Assigned:)

@m-kus @dmirgaleev
Please check the PR.

I decided not to change ctx.add_contract as it saves data to db, which might confusing to some users who explicitly configured contract with padded address.

I added truncation on both addresses (one received from datasource and other from config) inside matching procedure (which might slow matching down, not cool I know), but that allows achieve the desired result without changing stored data. Let me know if that assumption was wrong.

Also while doing this I discovered couple of things, that might be important, but also might be not:

1. Startknet indexers are poorly tested (to put mildly)

2. Locally running indexer upon switching to real-time indexing produces weird logs:
   

3. [this on is defo not important ] Config package would much easier to work with if packages were grouped by their base classes, like:

   • datasource_evm_etherscan
   • handler_starknet_event

I understand that event are always in handlers, but it just easier to navigate especially because the config structure is
one of the first thing you learn.

Hi @baitcode !

I'll answer to your questions in this thread for consistency, but please feel free to open a separate issues even for small questions/suggestions.

1. Yup. The easiest way you can help with improving the situation is to find a complex real-world example, find a slice of blocks to cover the cases we're interested in, then drop a test config to tests/configs.

2. Logging for Subsquid+Node indexes is to be improved in 8.2 release.

3. We'd like to keep the package structure (both dipdup and user's one) as it is, because breaking changes bring a lot of pain in codegen, docs and migration between major versions.