Querying addresses that have millions of transactions is supported however
takes hundreds of seconds to fully calculate the balance. Creating a cache of
previous results wasn't currently working because the `isSpent` query is always
based on the current bitcoind tip. Thus the balance of the outputs would be included
however wouldn't be removed when spent as the output wouldn't be checked again
when querying for blocks past the last checkpoint. Including the satoshis in the
inputs address index would make it possible to subtract the spent amount,
however this degrades optimizations elsewhere. The syncing times or querying
for addresses with 10,000 transactions per address.
It may preferrable to have an additional address service that handles high-volume
addresses be on an opt-in basis so that a custom running client could select
high volume addresses to create optimizations for querying balances and history.
The strategies for creating indexes differs on these use cases.
There was a bug when getting unspent outputs that would include an output
that was spent in the mempool in addition to the new output with the change
address. This lead to a balance having an output counted twice towards the
end balance. The solution is to have the isSpent method for the address service
to also include if the output was spent in the mempool, as the isSpent
method exposed from bitcoind only includes if the output was spent in a block.
- To be able to query an inputTxId and inputIndex that spends an outputTxId and outputIndex
- Extends the mempoolSpentIndex to include the inputTxId and inputIndex
- Sort and paginate before getting full transaction details.
- Only get detailed transaction information for items within the current page.
- Improves the performance with large sets of transactions.
- Reindexing the bitcore-node database is required with this change.
- Address indexes are updated to include address and height in spent index so that both inputs and outputs can be queried by address and height using "start" and "stop" to limit the range of the query.
- Address history also now supports paginated results using "from" and "to" values that indicate an index in the array.
