Don't cut bytes from non-address topics + account for empty topics

[catalyst17]

TL;DR

Improved handling of topic values in the getTopicValueFormat function.

What changed?

• Modified the getTopicValueFormat function to handle empty topic strings.
• Updated the logic for processing non-empty topics:
  • Now uses ethereum.FromHex to convert the topic to bytes
    • ⚠️ prev. used ethereum.HexToAddress was wrong in this context cause topics are not necessarily addresses)
  • Ensures the byte slice is exactly 32 bytes by left-padding with zeros.
  • Converts the padded bytes to a hash and returns the hexadecimal representation.

How to test?

1. Test the function with an empty string input to ensure it returns an empty string.
2. Test with various non-empty topic strings, including:
   • Short strings (less than 32 bytes)
   • Exactly 32-byte strings
   • Strings longer than 32 bytes
3. Verify that the output is always a 32-byte hexadecimal string (or empty for empty input).

Why make this change?

This change improves the robustness of topic value handling:

• It correctly handles empty topics, which are stored as empty strings by the indexer.
• It ensures consistent 32-byte padding for all non-empty topics, regardless of their original length.
• This approach aligns better with Ethereum's topic encoding standards, potentially fixing issues related to topic matching or filtering.

[catalyst17]

• Enable aggregates and group_by #109
• Don't cut bytes from non-address topics + account for empty topics #108 👈
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @catalyst17 and the rest of your teammates on Graphite

[iuwqyir]

Can we change this to nil already since we're already working on this or will it blow it up too much?

[catalyst17]

that's a separate thing I wanted to bring up, actually

saw this and thought that maybe we can actually leave it as is, but change the columns to not allow NULLs?

[iuwqyir]

yeah, sounds good. Doesn't really matter '' or null for our use case. We can change the schema and then apply it manually

[catalyst17]

since we anyway seem to have no benefit of NULL-ability in these columns and bloom filters are behaving the same way with negation - we can at least optimise some space and performance by completely turning it of

(and, consequently, continuing to use empty strings for topics that don't exist)