lbcwallet/txstore/serialization.go

1221 lines
30 KiB
Go
Raw Normal View History

/*
* Copyright (c) 2013, 2014 Conformal Systems LLC <info@conformal.com>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
package txstore
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
import (
"bytes"
"encoding/binary"
"errors"
"fmt"
"io"
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
"io/ioutil"
"os"
"path/filepath"
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
"time"
"github.com/btcsuite/btcd/wire"
"github.com/btcsuite/btcutil"
"github.com/btcsuite/btcwallet/rename"
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
)
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
// filename is the name of the file typically used to save a transaction
// store on disk.
const filename = "tx.bin"
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
// All Store versions (both old and current).
const (
versFirst uint32 = iota
// versRecvTxIndex is the version where the txout index
// was added to the RecvTx struct.
versRecvTxIndex
// versMarkSentChange is the version where serialized SentTx
// added a flags field, used for marking a sent transaction
// as change.
versMarkSentChange
// versCombined is the version where the old utxo and tx stores
// were combined into a single data structure.
versCombined
// versFastRewrite is the version where the combined store was
// rewritten with a focus on insertion and lookup speed.
versFastRewrite
// versCurrent is the current tx file version.
versCurrent = versFastRewrite
)
// byteOrder is the byte order used to read and write txstore binary data.
var byteOrder = binary.LittleEndian
// ReadFrom satisifies the io.ReaderFrom interface by deserializing a
// transaction store from an io.Reader.
func (s *Store) ReadFrom(r io.Reader) (int64, error) {
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
// Don't bother locking this. The mutex gets overwritten anyways.
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
var buf [4]byte
uint32Bytes := buf[:4]
// Read current file version.
n, err := io.ReadFull(r, uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
vers := byteOrder.Uint32(uint32Bytes)
// Reading files with versions before versFastRewrite is unsupported.
if vers < versFastRewrite {
return n64, ErrUnsupportedVersion
}
// Reset store.
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
*s = *New(s.path)
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
// Read block structures. Begin by reading the total number of block
// structures to be read, and then iterate that many times to read
// each block.
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
blockCount := byteOrder.Uint32(uint32Bytes)
// The blocks slice is *not* preallocated to blockCount size to prevent
// accidentally allocating so much memory that the process dies.
for i := uint32(0); i < blockCount; i++ {
b := &blockTxCollection{
txIndexes: map[int]uint32{},
}
tmpn64, err := b.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
s.blocks = append(s.blocks, b)
s.blockIndexes[b.Height] = i
// Recreate store unspent map.
for blockIndex, i := range b.txIndexes {
tx := b.txs[i]
for outputIdx, cred := range tx.credits {
if cred == nil {
continue
}
if cred.spentBy == nil {
op := wire.OutPoint{
Hash: *tx.tx.Sha(),
Index: uint32(outputIdx),
}
s.unspent[op] = BlockTxKey{
BlockIndex: blockIndex,
BlockHeight: b.Height,
}
}
}
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
}
}
// Read unconfirmed transactions and their spend tracking.
tmpn64, err := s.unconfirmed.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
return n64, nil
}
// WriteTo satisifies the io.WriterTo interface by serializing a transaction
// store to an io.Writer.
func (s *Store) WriteTo(w io.Writer) (int64, error) {
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
s.mtx.RLock()
defer s.mtx.RUnlock()
return s.writeTo(w)
}
func (s *Store) writeTo(w io.Writer) (int64, error) {
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
var buf [4]byte
uint32Bytes := buf[:4]
// Write current file version.
byteOrder.PutUint32(uint32Bytes, versCurrent)
n, err := w.Write(uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
// Write block structures. This begins with a uint32 specifying that
// some N blocks have been written, followed by N serialized transaction
// store blocks.
//
// The store's blockIndexes map is intentionally not written. Instead,
// it is recreated on reads after reading each block.
byteOrder.PutUint32(uint32Bytes, uint32(len(s.blocks)))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
for _, b := range s.blocks {
n, err := b.WriteTo(w)
n64 += n
if err != nil {
return n64, err
}
}
// Write unconfirmed transactions and their spend tracking.
tmpn64, err := s.unconfirmed.WriteTo(w)
n64 += tmpn64
if err != nil {
return n64, err
}
// The store's unspent map is intentionally not written. Instead, it
// is recreated on reads after each block transaction collection has
// been read. This makes reads more expensive, but writing faster, and
// as writes are far more common in application use, this was deemed to
// be an acceptable tradeoff.
return n64, nil
}
func (b *blockTxCollection) ReadFrom(r io.Reader) (int64, error) {
var buf [8]byte
uint64Bytes := buf[:8]
uint32Bytes := buf[:4]
// Read block hash, unix time (int64), and height (int32).
n, err := io.ReadFull(r, b.Hash[:])
n64 := int64(n)
if err != nil {
return n64, err
}
n, err = io.ReadFull(r, uint64Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
b.Time = time.Unix(int64(byteOrder.Uint64(uint64Bytes)), 0)
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
b.Height = int32(byteOrder.Uint32(uint32Bytes))
// Read amount deltas as a result of transactions in this block. This
// is the net total spendable balance as a result of transaction debits
// and credits, and the block reward (not immediately spendable) for
// coinbase outputs. Both are int64s.
n, err = io.ReadFull(r, uint64Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
b.amountDeltas.Spendable = btcutil.Amount(byteOrder.Uint64(uint64Bytes))
n, err = io.ReadFull(r, uint64Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
b.amountDeltas.Reward = btcutil.Amount(byteOrder.Uint64(uint64Bytes))
// Read number of transaction records (as a uint32) followed by a read
// for each expected record.
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
txCount := byteOrder.Uint32(uint32Bytes)
// The txs slice is *not* preallocated to txcount size to prevent
// accidentally allocating so much memory that the process dies.
for i := uint32(0); i < txCount; i++ {
t := &txRecord{}
tmpn64, err := t.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
b.txs = append(b.txs, t)
// Recreate txIndexes map. For each transaction record, map the
// block index of the underlying transaction to the slice index
// of the record.
b.txIndexes[t.tx.Index()] = i
}
return n64, nil
}
func (b *blockTxCollection) WriteTo(w io.Writer) (int64, error) {
var buf [8]byte
uint64Bytes := buf[:8]
uint32Bytes := buf[:4]
// Write block hash, unix time (int64), and height (int32).
n, err := w.Write(b.Hash[:])
n64 := int64(n)
if err != nil {
return n64, err
}
byteOrder.PutUint64(uint64Bytes, uint64(b.Time.Unix()))
n, err = w.Write(uint64Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
byteOrder.PutUint32(uint32Bytes, uint32(b.Height))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
// Write amount deltas as a result of transactions in this block.
// This is the net total spendable balance as a result of transaction
// debits and credits, and the block reward (not immediately spendable)
// for coinbase outputs. Both are int64s.
byteOrder.PutUint64(uint64Bytes, uint64(b.amountDeltas.Spendable))
n, err = w.Write(uint64Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
byteOrder.PutUint64(uint64Bytes, uint64(b.amountDeltas.Reward))
n, err = w.Write(uint64Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
// Write number of transaction records (as a uint32) followed by each
// transaction record.
byteOrder.PutUint32(uint32Bytes, uint32(len(b.txs)))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
for _, t := range b.txs {
n, err := t.WriteTo(w)
n64 += n
if err != nil {
return n64, err
}
}
// The block's txIndexes and unspent bookkeeping maps are intentionally
// not written. They are instead recreated on reads. This makes reads
// more expensive, but writing faster, and as writes are far more common
// in application use, this was deemed to be an acceptable tradeoff.
return n64, nil
}
const (
nilPointer byte = iota
validPointer
)
func byteMarksValidPointer(b byte) (bool, error) {
switch b {
case nilPointer:
return false, nil
case validPointer:
return true, nil
default:
s := "invalid byte representation of valid pointer"
return false, errors.New(s)
}
}
const (
falseByte byte = iota
trueByte
)
func byteAsBool(b byte) (bool, error) {
switch b {
case falseByte:
return false, nil
case trueByte:
return true, nil
default:
return false, errors.New("invalid byte representation of bool")
}
}
func (t *txRecord) ReadFrom(r io.Reader) (int64, error) {
var buf [8]byte
uint64Bytes := buf[:8]
uint32Bytes := buf[:4]
singleByte := buf[:1]
// Read transaction index (as a uint32).
n, err := io.ReadFull(r, uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
txIndex := int(byteOrder.Uint32(uint32Bytes))
// Deserialize transaction.
msgTx := new(msgTx)
tmpn64, err := msgTx.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
// Create and save the btcutil.Tx of the read MsgTx and set its index.
tx := btcutil.NewTx((*wire.MsgTx)(msgTx))
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
tx.SetIndex(txIndex)
t.tx = tx
// Read identifier for existance of debits.
n, err = io.ReadFull(r, singleByte)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
hasDebits, err := byteMarksValidPointer(singleByte[0])
if err != nil {
return n64, err
}
// If debits have been set, read them. Otherwise, set to nil.
if hasDebits {
// Read debited amount (int64).
n, err := io.ReadFull(r, uint64Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
amount := btcutil.Amount(byteOrder.Uint64(uint64Bytes))
// Read number of written outputs (as a uint32) this record
// debits from.
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
spendsCount := byteOrder.Uint32(uint32Bytes)
// For each expected output key, allocate and read the key,
// appending the result to the spends slice. This slice is
// originally set empty (*not* preallocated to spendsCount
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
// size) to prevent accidentally allocating so much memory that
// the process dies.
var spends []BlockOutputKey
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
for i := uint32(0); i < spendsCount; i++ {
k := BlockOutputKey{}
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
tmpn64, err := k.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
spends = append(spends, k)
}
t.debits = &debits{amount, spends}
} else {
t.debits = nil
}
// Read number of pointers (as a uint32) written to be read into the
// credits slice (although some may be nil).
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
creditsCount := byteOrder.Uint32(uint32Bytes)
// For each expected credits slice element, check whether the credit
// exists or the pointer is nil. If nil, append nil to credits and
// continue with the next. If non-nil, allocated and read the full
// credit structure. This slice is originally set to nil (*not*
// preallocated to creditsCount size) to prevent accidentally allocating
// so much memory that the process dies.
var credits []*credit
for i := uint32(0); i < creditsCount; i++ {
// Read identifer for a valid pointer.
n, err := io.ReadFull(r, singleByte)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
validCredit, err := byteMarksValidPointer(singleByte[0])
if err != nil {
return n64, err
}
if !validCredit {
credits = append(credits, nil)
} else {
// Read single byte that specifies whether this credit
// was added as change.
n, err = io.ReadFull(r, singleByte)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
change, err := byteAsBool(singleByte[0])
if err != nil {
return n64, err
}
// Read single byte. This was previously used to
// specify whether an unspent credit was locked or not,
// but this was removed as lockedness is volatile and
// should not be saved.
//
// This space can be used for additional flags in the
// future.
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
n, err = io.ReadFull(r, singleByte)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
// Read identifier for a valid pointer.
n, err = io.ReadFull(r, singleByte)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
validSpentBy, err := byteMarksValidPointer(singleByte[0])
if err != nil {
return n64, err
}
// If spentBy pointer is valid, allocate and read a
// transaction lookup key.
var spentBy *BlockTxKey
if validSpentBy {
spentBy = &BlockTxKey{}
tmpn64, err := spentBy.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
}
c := &credit{change, spentBy}
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
credits = append(credits, c)
}
}
t.credits = credits
// Read received unix time (int64).
n, err = io.ReadFull(r, uint64Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
received := int64(byteOrder.Uint64(uint64Bytes))
t.received = time.Unix(received, 0)
return n64, nil
}
func (t *txRecord) WriteTo(w io.Writer) (int64, error) {
var buf [8]byte
uint64Bytes := buf[:8]
uint32Bytes := buf[:4]
// Write transaction index (as a uint32).
byteOrder.PutUint32(uint32Bytes, uint32(t.tx.Index()))
n, err := w.Write(uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
// Serialize and write transaction.
tmpn64, err := (*msgTx)(t.tx.MsgTx()).WriteTo(w)
n64 += tmpn64
if err != nil {
return n64, err
}
// Write debit records, if any. This begins with a single byte to
// identify whether the record contains any debits or not.
if t.debits == nil {
// Write identifier for nil debits.
n, err = w.Write([]byte{nilPointer})
n64 += int64(n)
if err != nil {
return n64, err
}
} else {
// Write identifier for valid debits.
n, err = w.Write([]byte{validPointer})
n64 += int64(n)
if err != nil {
return n64, err
}
// Write debited amount (int64).
byteOrder.PutUint64(uint64Bytes, uint64(t.debits.amount))
n, err := w.Write(uint64Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
// Write number of outputs (as a uint32) this record debits
// from.
byteOrder.PutUint32(uint32Bytes, uint32(len(t.debits.spends)))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
// Write each lookup key for a spent transaction output.
for _, k := range t.debits.spends {
tmpn64, err := k.WriteTo(w)
n64 += tmpn64
if err != nil {
return n64, err
}
}
}
// Write number of pointers (as a uint32) in the credits slice (although
// some may be nil). Then, for each element in the credits slice, write
// an identifier whether the element is nil or valid, and if valid,
// write the credit structure.
byteOrder.PutUint32(uint32Bytes, uint32(len(t.credits)))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
for _, c := range t.credits {
if c == nil {
// Write identifier for nil credit.
n, err := w.Write([]byte{nilPointer})
n64 += int64(n)
if err != nil {
return n64, err
}
} else {
// Write identifier for valid credit.
n, err := w.Write([]byte{validPointer})
n64 += int64(n)
if err != nil {
return n64, err
}
// Write a single byte to specify whether this credit
// was added as change, plus an extra empty byte which
// used to specify whether the credit was locked. This
// extra byte is currently unused and may be used for
// other flags in the future.
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
changeByte := falseByte
if c.change {
changeByte = trueByte
}
n, err = w.Write([]byte{changeByte, 0})
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
n64 += int64(n)
if err != nil {
return n64, err
}
// If this credit is unspent, write an identifier for
// an invalid pointer. Otherwise, write the identifier
// for a valid pointer and write the spending tx key.
if c.spentBy == nil {
// Write identifier for an unspent credit.
n, err := w.Write([]byte{nilPointer})
n64 += int64(n)
if err != nil {
return n64, err
}
} else {
// Write identifier for an unspent credit.
n, err := w.Write([]byte{validPointer})
n64 += int64(n)
if err != nil {
return n64, err
}
// Write transaction lookup key.
tmpn64, err := c.spentBy.WriteTo(w)
n64 += tmpn64
if err != nil {
return n64, err
}
}
}
}
// Write received unix time (int64).
byteOrder.PutUint64(uint64Bytes, uint64(t.received.Unix()))
n, err = w.Write(uint64Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
return n64, nil
}
type msgTx wire.MsgTx
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
func (tx *msgTx) ReadFrom(r io.Reader) (int64, error) {
// Read from a TeeReader to return the number of read bytes.
buf := bytes.Buffer{}
tr := io.TeeReader(r, &buf)
if err := (*wire.MsgTx)(tx).Deserialize(tr); err != nil {
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
if buf.Len() != 0 && err == io.EOF {
err = io.ErrUnexpectedEOF
}
return int64(buf.Len()), err
}
return int64((*wire.MsgTx)(tx).SerializeSize()), nil
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
}
func (tx *msgTx) WriteTo(w io.Writer) (int64, error) {
// Write to a buffer and then copy to w so the total number of bytes
// written can be returned to the caller. Writing to a to a
2014-05-30 19:46:41 +02:00
// bytes.Buffer never fails except for OOM panics, so check and panic
// on any unexpected non-nil returned errors.
buf := bytes.Buffer{}
if err := (*wire.MsgTx)(tx).Serialize(&buf); err != nil {
panic(err)
}
return io.Copy(w, &buf)
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
}
// ReadFrom reads a mined transaction output lookup key from r. The total
// number of bytes read is returned.
func (k *BlockOutputKey) ReadFrom(r io.Reader) (int64, error) {
var buf [4]byte
uint32Bytes := buf[:4]
// Read embedded BlockTxKey.
n64, err := k.BlockTxKey.ReadFrom(r)
if err != nil {
return n64, err
}
// Read output index (uint32).
n, err := io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
k.OutputIndex = byteOrder.Uint32(uint32Bytes)
return n64, nil
}
// WriteTo writes a mined transaction output lookup key to w. The total number
// of bytes written is returned.
func (k *BlockOutputKey) WriteTo(w io.Writer) (int64, error) {
var buf [4]byte
uint32Bytes := buf[:4]
// Write embedded BlockTxKey.
n64, err := k.BlockTxKey.WriteTo(w)
if err != nil {
return n64, err
}
// Write output index (uint32).
byteOrder.PutUint32(uint32Bytes, k.OutputIndex)
n, err := w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
return n64, nil
}
// ReadFrom reads a mined transaction lookup key from r. The total number of
// bytes read is returned.
func (k *BlockTxKey) ReadFrom(r io.Reader) (int64, error) {
var buf [4]byte
uint32Bytes := buf[:4]
// Read block index (as a uint32).
n, err := io.ReadFull(r, uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
k.BlockIndex = int(byteOrder.Uint32(uint32Bytes))
// Read block height (int32).
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
k.BlockHeight = int32(byteOrder.Uint32(uint32Bytes))
return n64, nil
}
// WriteTo writes a mined transaction lookup key to w. The total number of
// bytes written is returned.
func (k *BlockTxKey) WriteTo(w io.Writer) (int64, error) {
var buf [4]byte
uint32Bytes := buf[:4]
// Write block index (as a uint32).
byteOrder.PutUint32(uint32Bytes, uint32(k.BlockIndex))
n, err := w.Write(uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
// Write block height (int32).
byteOrder.PutUint32(uint32Bytes, uint32(k.BlockHeight))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
return n64, nil
}
func (u *unconfirmedStore) ReadFrom(r io.Reader) (int64, error) {
var buf [4]byte
uint32Bytes := buf[:4]
// Read length (as a uint32) of transaction record key/value pairs,
// followed by each transaction record.
n, err := io.ReadFull(r, uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
txCount := byteOrder.Uint32(uint32Bytes)
for i := uint32(0); i < txCount; i++ {
t := &txRecord{}
tmpn64, err := t.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
u.txs[*t.tx.Sha()] = t
}
// Read length (as a uint32) of key/value pairs in the
// spentBlockOutPoints and spentBlockOutPointKeys maps, followed by the
// outpoint, the block transaction lookup key, and the transaction hash
// of the spending transaction record.
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
spentBlockOutPointCount := byteOrder.Uint32(uint32Bytes)
for i := uint32(0); i < spentBlockOutPointCount; i++ {
// Read outpoint hash and index (uint32).
op := wire.OutPoint{}
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
n, err := io.ReadFull(r, op.Hash[:])
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
op.Index = byteOrder.Uint32(uint32Bytes)
// Read block transaction lookup key, and create the full block
// output key from it and the previously-read outpoint index.
opKey := BlockOutputKey{OutputIndex: op.Index}
tmpn64, err := opKey.BlockTxKey.ReadFrom(r)
n64 += tmpn64
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
// Read transaction record hash and check that it was previously
// read into the txs map. Use full record as the map value.
var txHash wire.ShaHash
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
n, err = io.ReadFull(r, txHash[:])
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
t, ok := u.txs[txHash]
if !ok {
return n64, fmt.Errorf("missing unconfirmed "+
"transaction record for transaction %v", txHash)
}
u.spentBlockOutPoints[opKey] = t
u.spentBlockOutPointKeys[op] = opKey
}
// Read length (as a uint32) of key/value pairs in the spentUnconfirmed
// map, followed by the outpoint and hash of the transaction record.
// Use this hash as the lookup key for the full transaction record
// previously read into the txs map.
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
spentUnconfirmedCount := byteOrder.Uint32(uint32Bytes)
for i := uint32(0); i < spentUnconfirmedCount; i++ {
// Read outpoint hash and index (uint32).
op := wire.OutPoint{}
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
n, err := io.ReadFull(r, op.Hash[:])
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
n, err = io.ReadFull(r, uint32Bytes)
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
op.Index = byteOrder.Uint32(uint32Bytes)
// Read transaction record hash and check that it was previously
// read into the txs map. Use full record as the map value.
var txHash wire.ShaHash
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
n, err = io.ReadFull(r, txHash[:])
n64 += int64(n)
if err != nil {
if err == io.EOF {
err = io.ErrUnexpectedEOF
}
return n64, err
}
t, ok := u.txs[txHash]
if !ok {
return n64, fmt.Errorf("missing unconfirmed "+
"transaction record for transaction %v", txHash)
}
u.spentUnconfirmed[op] = t
}
// Recreate the previousOutpoints map. For each transaction record
// saved in the txs map, map each previous outpoint to the record
// itself.
for _, t := range u.txs {
for _, input := range t.tx.MsgTx().TxIn {
u.previousOutpoints[input.PreviousOutPoint] = t
Another day, another tx store implementation. The last transaction store was a great example of how not to write scalable software. For a variety of reasons, it was very slow at processing transaction inserts. Among them: 1) Every single transaction record being saved in a linked list (container/list), and inserting into this list would be an O(n) operation so that records could be ordered by receive date. 2) Every single transaction in the above mentioned list was iterated over in order to find double spends which must be removed. It is silly to do this check for mined transactions, which already have been checked for this by btcd. Worse yet, if double spends were found, the list would be iterated a second (or third, or fourth) time for each removed transaction. 3) All spend tracking for signed-by-wallet transactions was found on each transaction insert, even if the now spent previous transaction outputs were known by the caller. This list could keep going on, but you get the idea. It was bad. To resolve these issues a new transaction store had to be implemented. The new implementation: 1) Tracks mined and unmined transactions in different data structures. Mined transactions are cheap to track because the required double spend checks have already been performed by the chain server, and double spend checks are only required to be performed on newly-inserted mined transactions which may conflict with previous unmined transactions. 2) Saves mined transactions grouped by block first, and then by their transaction index. Lookup keys for mined transactions are simply the block height (in the best chain, that's all we save) and index of the transaction in the block. This makes looking up any arbitrary transaction almost an O(1) operation (almost, because block height and block indexes are mapped to their slice indexes with a Go map). 3) Saves records in each transaction for whether the outputs are wallet credits (spendable by wallet) and for whether inputs debit from previous credits. Both structures point back to the source or spender (credits point to the transaction that spends them, or nil for unspent credits, and debits include keys to lookup the transaction credits they spent. While complicated to keep track of, this greatly simplifies the spent tracking for transactions across rollbacks and transaction removals. 4) Implements double spend checking as an almost O(1) operation. A Go map is used to map each previous outpoint for all unconfirmed transactions to the unconfirmed tx record itself. Checking for double spends on confirmed transaction inserts only involves looking up each previous outpoint of the inserted tx in this map. If a double spend is found, removal is simplified by only removing the transaction and its spend chain from store maps, rather than iterating a linked list several times over to remove each dead transaction in the spend chain. 5) Allows the caller to specify the previous credits which are spent by a debiting transaction. When a transaction is created by wallet, the previous outputs are already known, and by passing their record types to the AddDebits method, lookups for each previously unspent credit are omitted. 6) Bookkeeps all blocks with transactions with unspent credits, and bookkeeps the transaction indexes of all transactions with unspent outputs for a single block. For the case where the caller adding a debit record does not know what credits a transaction debits from, these bookkeeping structures allow the store to only consider known unspent transactions, rather than searching through both spent and unspents. 7) Saves amount deltas for the entire balance as a result of each block, due to transactions within that block. This improves the performance of calculating the full balance by not needing to iterate over every transaction, and then every credit, to determine if a credit is spent or unspent. When transactions are moved from unconfirmed to a block structure, the amount deltas are incremented by the amount of all transaction credits (both spent and unspent) and debited by the total amount the transaction spends from previous wallet credits. For the common case of calculating a balance with just one confirmation, the only involves iterating over each block structure and adding the (possibly negative) amount delta. Coinbase rewards are saved similarly, but with a different amount variable so they can be seperatly included or excluded. Due to all of the changes in how the store internally works, the serialization format has changed. To simplify the serialization logic, support for reading the last store file version has been removed. Past this change, a rescan (run automatically) will be required to rebuild the transaction history.
2014-05-05 23:12:05 +02:00
}
}
return n64, nil
}
func (u *unconfirmedStore) WriteTo(w io.Writer) (int64, error) {
var buf [4]byte
uint32Bytes := buf[:4]
// Write length of key/values pairs in txs map, followed by each
// transaction record.
byteOrder.PutUint32(uint32Bytes, uint32(len(u.txs)))
n, err := w.Write(uint32Bytes)
n64 := int64(n)
if err != nil {
return n64, err
}
for _, t := range u.txs {
tmpn64, err := t.WriteTo(w)
n64 += tmpn64
if err != nil {
return n64, err
}
}
// Write length (as a uint32) of key/value pairs in the
// spentBlockOutPoints and spentBlockOutPointKeys maps (these lengths
// must be equal), followed by the outpoint, the block transaction
// lookup key, and the hash of the transaction record.
if len(u.spentBlockOutPoints) != len(u.spentBlockOutPointKeys) {
return n64, errors.New("spent block tx maps lengths differ")
}
byteOrder.PutUint32(uint32Bytes, uint32(len(u.spentBlockOutPoints)))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
for op, opKey := range u.spentBlockOutPointKeys {
// Write outpoint hash and the index (uint32).
n, err := w.Write(op.Hash[:])
n64 += int64(n)
if err != nil {
return n64, err
}
byteOrder.PutUint32(uint32Bytes, op.Index)
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
// Write the block transaction lookup key. This is not the full
// output key, as the index has already been serialized as part
// of the outpoint written above.
tmpn64, err := opKey.BlockTxKey.WriteTo(w)
n64 += tmpn64
if err != nil {
return n64, err
}
// Lookup transaction record and write the transaction hash.
t, ok := u.spentBlockOutPoints[opKey]
if !ok {
return n64, MissingCreditError(opKey)
}
n, err = w.Write(t.tx.Sha()[:])
n64 += int64(n)
if err != nil {
return n64, err
}
}
// Write length (as a uint32) of key/value pairs in the spentUnconfirmed
// map, followed by the outpoint and hash of the transaction record.
byteOrder.PutUint32(uint32Bytes, uint32(len(u.spentUnconfirmed)))
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
for op, t := range u.spentUnconfirmed {
// Write outpoint hash and the index (uint32).
n, err := w.Write(op.Hash[:])
n64 += int64(n)
if err != nil {
return n64, err
}
byteOrder.PutUint32(uint32Bytes, op.Index)
n, err = w.Write(uint32Bytes)
n64 += int64(n)
if err != nil {
return n64, err
}
// Write transaction record hash.
n, err = w.Write(t.tx.Sha()[:])
n64 += int64(n)
if err != nil {
return n64, err
}
}
// The previousOutpoints map is intentionally not written, as it can
// be fully recreated by iterating each transaction record and adding
// a key/value pair for each prevous outpoint. This is performed when
// reading the unconfirmed store. This makes reads slightly more
// expensive, but writing faster, and as writes are far more common in
// application use, this was deemed to be an acceptable tradeoff.
return n64, nil
}
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
// MarkDirty marks that changes have been made to the transaction store.
// This should be run after any modifications are performed to the store
// or any of its records.
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
func (s *Store) MarkDirty() {
s.mtx.Lock()
defer s.mtx.Unlock()
s.dirty = true
}
// WriteIfDirty writes the entire transaction store to permanent storage if
// the dirty flag has been set (see MarkDirty).
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
func (s *Store) WriteIfDirty() error {
s.mtx.RLock()
if !s.dirty {
s.mtx.RUnlock()
return nil
}
// TempFile creates the file 0600, so no need to chmod it.
fi, err := ioutil.TempFile(s.dir, s.file)
if err != nil {
s.mtx.RUnlock()
return err
}
fiPath := fi.Name()
_, err = s.writeTo(fi)
if err != nil {
s.mtx.RUnlock()
fi.Close()
return err
}
err = fi.Sync()
if err != nil {
s.mtx.RUnlock()
fi.Close()
return err
}
fi.Close()
err = rename.Atomic(fiPath, s.path)
s.mtx.RUnlock()
if err == nil {
s.mtx.Lock()
s.dirty = false
s.mtx.Unlock()
}
return err
}
// OpenDir opens a new transaction store from the specified directory.
// If the file does not exist, the error from the os package will be
// returned, and can be checked with os.IsNotExist to differentiate missing
// file errors from others (including deserialization).
func OpenDir(dir string) (*Store, error) {
path := filepath.Join(dir, filename)
fi, err := os.OpenFile(path, os.O_RDONLY, 0)
if err != nil {
return nil, err
}
defer fi.Close()
store := new(Store)
_, err = store.ReadFrom(fi)
if err != nil {
return nil, err
}
store.path = path
store.dir = dir
return store, nil
}