Allow strings and bigints instead of numbers in decoder input fields #6131
Conversation
haltman-at
changed the title
packages/decoder/lib/types.ts
Outdated
| /** | ||
| * Index of the log within the block. | ||
| */ | ||
| logIndex?: number; | ||
| /** | ||
| * Index within the block of the emitting transaction; null if | ||
| * block is pending. | ||
| */ | ||
| transactionIndex?: number | null; |
There was a problem hiding this comment.
I think these optional fields need explanation on why they might be undefined. Probably warrants a note in the DecodedLog docstring, to say that @truffle/decoder re-emits everything given, and unspecified optional input fields will thus not appear in output
There was a problem hiding this comment.
Actually, that's not how this works! Which raises a good point: These fields shouldn't be optional. I guess I should just not make this an extension of Log, huh? Because it doesn't save me any work if everything is mandatory. :-/
To explain: If you use decodeLog, passing in your own logs, you just get back the decodings, you don't get DecodedLog. You only get DecodedLog if you call events(), to which one does not pass in their own logs. So actually these fields will always be present in this type.
...so do you agree that's what I should do?
There was a problem hiding this comment.
Can this emit the same info in both situations? It's weird that passing in a Log will remove fields, while events() will return everything log-related
There was a problem hiding this comment.
What? Passing in a Log doesn't remove fields. What I'm saying is that decodeLog doesn't use the DecodedLog type at all. It returns the decodings, they're not inside a DecodedLog container. Only events uses that.
There was a problem hiding this comment.
Ohh, sorry to misunderstand. Yeah, let's not extend Log
PR description
Addresses #6080. This makes it so that fields in
LogorTransactionthat previously werenumberare nownumber | string | bigint, and fields that werenumber | nullare nownumber | string | bigint | null.Most of these fields are ignored, so the type is changed and that's all.
The one changed field that's actually used is
blockNumber, so I just changed things so that if we get that and it's not null, we convert it to a number.I do have to also discuss what happened to the
DecodedLogtype. For simplicity,DecodedLogextendsLog. However,Loghas just had its type broadened. But while that's fine forLog, because it's an input type,DecodedLogis an output type. So to broaden its type would be a breaking change. So I re-narrowed the types of the appropriate fields to prevent that.Anyway that's it (aside from tests); this is pretty simple.
Testing instructions
I altered some decoder tests so that they pass in strings for the block number (the only part that matters), I think that should be sufficient.