Fix for issue 4205 (optimize eth_getLogs)

[leonardchinonso]

I used rawdb.ReadCanonicalBodyWithTransactions to get the list of Transactions associated with the block hash and number. Then used the txLogIndex for the particular transaction. Apart from the overhead of using the api compared to rawdb, I do not see any other optimization. Did I do it wrong?

[AskAlexSharov]

@leonardchinonso rawdb read from db only, but when --snapshots=true historical blocks are in snapshots only. So, need use api._blockReader (it's BlockReaderWithSnapshots class) - it will detect by block number: read from snapshots or from db.

[AskAlexSharov]

you can use another method of api._blockReader or add new method there.

[leonardchinonso]

I checked out the _blockReader field and it already had the BodyWithTransactions service that rawdb had. Like you said, it searches the snapshots files too. However, is there any more optimization we can get?
I don't know if we need the whole body of transactions in that block since we will only take the ones that match our criteria

[AskAlexSharov]

By _blockReader can iterate over transactions (like in db).
Is it possible to know which transactions need to read without reading them?

[leonardchinonso]

Fair enough. In that case, the changes should work for this. Any tests I need to add or rerun?

[AskAlexSharov]

Or return error message or log it