self.address_queue = Queue()
self.dbpath = config.get('leveldb', 'path')
+ self.pruning_limit = config.getint('leveldb', 'pruning_limit')
+ self.db_version = 1 # increase this when database needs to be updated
self.dblock = threading.Lock()
try:
try:
hist = self.deserialize(self.db.Get('height'))
- self.last_hash, self.height, _ = hist[0]
- print_log("hist", hist)
+ self.last_hash, self.height, db_version = hist[0]
+ print_log("Database version", self.db_version)
+ print_log("Blockchain height", self.height)
except:
- #traceback.print_exc(file=sys.stdout)
+ traceback.print_exc(file=sys.stdout)
print_log('initializing database')
self.height = 0
self.last_hash = '000000000019d6689c085ae165831e934ff763ae46a2a6c172b3f1b60a8ce26f'
+ db_version = self.db_version
+
+ # check version
+ if self.db_version != db_version:
+ print_log("Your database '%s' is deprecated. Please create a new database"%self.dbpath)
+ self.shared.stop()
+ return
# catch_up headers
self.init_headers(self.height)
def serialize(self, h):
s = ''
for txid, txpos, height in h:
- s += txid + int_to_hex(txpos, 4) + int_to_hex(height, 4)
- return s.decode('hex')
+ s += self.serialize_item(txid, txpos, height)
+ return s
+
+ def serialize_item(self, txid, txpos, height, spent=chr(0)):
+ s = (txid + int_to_hex(txpos, 4) + int_to_hex(height, 3)).decode('hex') + spent
+ return s
+
+ def deserialize_item(self,s):
+ txid = s[0:32].encode('hex')
+ txpos = int(rev_hex(s[32:36].encode('hex')), 16)
+ height = int(rev_hex(s[36:39].encode('hex')), 16)
+ spent = s[39:40]
+ return (txid, txpos, height, spent)
def deserialize(self, s):
h = []
while s:
- txid = s[0:32].encode('hex')
- txpos = int(rev_hex(s[32:36].encode('hex')), 16)
- height = int(rev_hex(s[36:40].encode('hex')), 16)
+ txid, txpos, height, spent = self.deserialize_item(s[0:40])
h.append((txid, txpos, height))
- s = s[40:]
+ if spent == chr(1):
+ txid, txpos, height, spent = self.deserialize_item(s[40:80])
+ h.append((txid, txpos, height))
+ s = s[80:]
return h
def block2header(self, b):
with self.dblock:
try:
- hash_160 = bc_address_to_hash_160(addr)
- hist = self.deserialize(self.db.Get(hash_160))
+ hist = self.deserialize(self.db.Get(addr))
is_known = True
except:
hist = []
is_known = False
- # should not be necessary
- hist.sort(key=lambda tup: tup[1])
- # check uniqueness too...
+ # sort history, because redeeming transactions are next to the corresponding txout
+ hist.sort(key=lambda tup: tup[2])
# add memory pool
with self.mempool_lock:
for txid in self.mempool_hist.get(addr, []):
hist.append((txid, 0, 0))
- hist = map(lambda x: {'tx_hash': x[0], 'height': x[2]}, hist)
+ # uniqueness
+ hist = set(map(lambda x: (x[0], x[2]), hist))
+
+ # convert to dict
+ hist = map(lambda x: {'tx_hash': x[0], 'height': x[1]}, hist)
+
# add something to distinguish between unused and empty addresses
if hist == [] and is_known:
hist = ['*']
return {"block_height": height, "merkle": s, "pos": tx_pos}
+
def add_to_history(self, addr, tx_hash, tx_pos, tx_height):
# keep it sorted
- s = (tx_hash + int_to_hex(tx_pos, 4) + int_to_hex(tx_height, 4)).decode('hex')
+ s = self.serialize_item(tx_hash, tx_pos, tx_height) + 40*chr(0)
+ assert len(s) == 80
serialized_hist = self.batch_list[addr]
- l = len(serialized_hist)/40
+ l = len(serialized_hist)/80
for i in range(l-1, -1, -1):
- item = serialized_hist[40*i:40*(i+1)]
- item_height = int(rev_hex(item[36:40].encode('hex')), 16)
- if item_height < tx_height:
- serialized_hist = serialized_hist[0:40*(i+1)] + s + serialized_hist[40*(i+1):]
+ item = serialized_hist[80*i:80*(i+1)]
+ item_height = int(rev_hex(item[36:39].encode('hex')), 16)
+ if item_height <= tx_height:
+ serialized_hist = serialized_hist[0:80*(i+1)] + s + serialized_hist[80*(i+1):]
break
else:
serialized_hist = s + serialized_hist
txo = (tx_hash + int_to_hex(tx_pos, 4)).decode('hex')
self.batch_txio[txo] = addr
- def remove_from_history(self, addr, tx_hash, tx_pos):
- txi = (tx_hash + int_to_hex(tx_pos, 4)).decode('hex')
- if addr is None:
- try:
- addr = self.batch_txio[txi]
- except:
- raise BaseException(tx_hash, tx_pos)
+ def revert_add_to_history(self, addr, tx_hash, tx_pos, tx_height):
+
+ serialized_hist = self.batch_list[addr]
+ s = self.serialize_item(tx_hash, tx_pos, tx_height) + 40*chr(0)
+ if serialized_hist.find(s) == -1: raise
+ serialized_hist = serialized_hist.replace(s, '')
+ self.batch_list[addr] = serialized_hist
+
+
+
+ def prune_history(self, addr, undo):
+ # remove items that have bit set to one
+ if undo.get(addr) is None: undo[addr] = []
+
+ serialized_hist = self.batch_list[addr]
+ l = len(serialized_hist)/80
+ for i in range(l):
+ if len(serialized_hist)/80 < self.pruning_limit: break
+ item = serialized_hist[80*i:80*(i+1)]
+ if item[39:40] == chr(1):
+ assert item[79:80] == chr(2)
+ serialized_hist = serialized_hist[0:80*i] + serialized_hist[80*(i+1):]
+ undo[addr].append(item) # items are ordered
+ self.batch_list[addr] = serialized_hist
+
+
+ def revert_prune_history(self, addr, undo):
+ # restore removed items
serialized_hist = self.batch_list[addr]
- l = len(serialized_hist)/40
+ if undo.get(addr) is not None:
+ itemlist = undo.pop(addr)
+ else:
+ return
+
+ if not itemlist: return
+
+ l = len(serialized_hist)/80
+ tx_item = ''
+ for i in range(l-1, -1, -1):
+ if tx_item == '':
+ if not itemlist:
+ break
+ else:
+ tx_item = itemlist.pop(-1) # get the last element
+ tx_height = int(rev_hex(tx_item[36:39].encode('hex')), 16)
+
+ item = serialized_hist[80*i:80*(i+1)]
+ item_height = int(rev_hex(item[36:39].encode('hex')), 16)
+
+ if item_height < tx_height:
+ serialized_hist = serialized_hist[0:80*(i+1)] + tx_item + serialized_hist[80*(i+1):]
+ tx_item = ''
+
+ else:
+ serialized_hist = ''.join(itemlist) + tx_item + serialized_hist
+
+ self.batch_list[addr] = serialized_hist
+
+
+ def set_spent_bit(self, addr, txi, is_spent, txid=None, index=None, height=None):
+ serialized_hist = self.batch_list[addr]
+ l = len(serialized_hist)/80
for i in range(l):
- item = serialized_hist[40*i:40*(i+1)]
+ item = serialized_hist[80*i:80*(i+1)]
if item[0:36] == txi:
- height = int(rev_hex(item[36:40].encode('hex')), 16)
- serialized_hist = serialized_hist[0:40*i] + serialized_hist[40*(i+1):]
+ if is_spent:
+ new_item = item[0:39] + chr(1) + self.serialize_item(txid, index, height, chr(2))
+ else:
+ new_item = item[0:39] + chr(0) + chr(0)*40
+ serialized_hist = serialized_hist[0:80*i] + new_item + serialized_hist[80*(i+1):]
break
else:
+ self.shared.stop()
hist = self.deserialize(serialized_hist)
- raise BaseException("prevout not found", addr, hist, tx_hash, tx_pos)
+ raise BaseException("prevout not found", addr, hist, txi.encode('hex'))
self.batch_list[addr] = serialized_hist
- return height, addr
+
+
+ def unset_spent_bit(self, addr, txi):
+ self.set_spent_bit(addr, txi, False)
+ self.batch_txio[txi] = addr
+
def deserialize_block(self, block):
txlist = block.get('tx')
t00 = time.time()
+ # undo info
+ if revert:
+ undo_info = self.get_undo_info(block_height)
+ else:
+ undo_info = {}
+
+
if not revert:
# read addresses of tx inputs
for tx in txdict.values():
for txi in block_inputs:
try:
addr = self.db.Get(txi)
- except:
+ except KeyError:
# the input could come from the same block
continue
+ except:
+ traceback.print_exc(file=sys.stdout)
+ self.shared.stop()
+ raise
+
self.batch_txio[txi] = addr
addr_to_read.append(addr)
for x in tx.get('outputs'):
txo = (txid + int_to_hex(x.get('index'), 4)).decode('hex')
block_outputs.append(txo)
+ addr_to_read.append( x.get('address') )
+
+ undo = undo_info.get(txid)
+ for i, x in enumerate(tx.get('inputs')):
+ addr = undo['prev_addr'][i]
+ addr_to_read.append(addr)
+
+
+
+
# read histories of addresses
for txid, tx in txdict.items():
for x in tx.get('outputs'):
- hash_160 = bc_address_to_hash_160(x.get('address'))
- addr_to_read.append(hash_160)
+ addr_to_read.append(x.get('address'))
addr_to_read.sort()
for addr in addr_to_read:
try:
self.batch_list[addr] = self.db.Get(addr)
- except:
+ except KeyError:
self.batch_list[addr] = ''
+ except:
+ traceback.print_exc(file=sys.stdout)
+ self.shared.stop()
+ raise
- if revert:
- undo_info = self.get_undo_info(block_height)
- # print "undo", block_height, undo_info
- else:
- undo_info = {}
# process
t1 = time.time()
if revert:
tx_hashes = tx_hashes[::-1]
+
+
for txid in tx_hashes: # must be ordered
tx = txdict[txid]
if not revert:
- undo = []
- for x in tx.get('inputs'):
- prevout_height, prevout_addr = self.remove_from_history(None, x.get('prevout_hash'), x.get('prevout_n'))
- undo.append((prevout_height, prevout_addr))
- undo_info[txid] = undo
+ undo = { 'prev_addr':[] } # contains the list of pruned items for each address in the tx; also, 'prev_addr' is a list of prev addresses
+
+ prev_addr = []
+ for i, x in enumerate(tx.get('inputs')):
+ txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex')
+ addr = self.batch_txio[txi]
+ # add redeem item to the history.
+ # add it right next to the input txi? this will break history sorting, but it's ok if I neglect tx inputs during search
+ self.set_spent_bit(addr, txi, True, txid, i, block_height)
+
+ # when I prune, prune a pair
+ self.prune_history(addr, undo)
+ prev_addr.append(addr)
+
+ undo['prev_addr'] = prev_addr
+
+ # here I add only the outputs to history; maybe I want to add inputs too (that's in the other loop)
for x in tx.get('outputs'):
- hash_160 = bc_address_to_hash_160(x.get('address'))
- self.add_to_history(hash_160, txid, x.get('index'), block_height)
+ addr = x.get('address')
+ self.add_to_history(addr, txid, x.get('index'), block_height)
+ self.prune_history(addr, undo) # prune here because we increased the length of the history
+
+ undo_info[txid] = undo
else:
+
+ undo = undo_info.pop(txid)
+
for x in tx.get('outputs'):
- hash_160 = bc_address_to_hash_160(x.get('address'))
- self.remove_from_history(hash_160, txid, x.get('index'))
+ addr = x.get('address')
+ self.revert_prune_history(addr, undo)
+ self.revert_add_to_history(addr, txid, x.get('index'), block_height)
+
+ prev_addr = undo.pop('prev_addr')
+ for i, x in enumerate(tx.get('inputs')):
+ addr = prev_addr[i]
+ self.revert_prune_history(addr, undo)
+ txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex')
+ self.unset_spent_bit(addr, txi)
- i = 0
- for x in tx.get('inputs'):
- prevout_height, prevout_addr = undo_info.get(txid)[i]
- i += 1
+ assert undo == {}
- # read the history into batch list
- if self.batch_list.get(prevout_addr) is None:
- self.batch_list[prevout_addr] = self.db.Get(prevout_addr)
+ if revert:
+ assert undo_info == {}
- # re-add them to the history
- self.add_to_history(prevout_addr, x.get('prevout_hash'), x.get('prevout_n'), prevout_height)
- # print_log("new hist for", hash_160_to_bc_address(prevout_addr), self.deserialize(self.batch_list[prevout_addr]) )
# write
max_len = 0
batch = leveldb.WriteBatch()
for addr, serialized_hist in self.batch_list.items():
batch.Put(addr, serialized_hist)
- l = len(serialized_hist)
+ l = len(serialized_hist)/80
if l > max_len:
max_len = l
max_addr = addr
else:
# restore spent inputs
for txio, addr in self.batch_txio.items():
+ # print "restoring spent input", repr(txio)
batch.Put(txio, addr)
# delete spent outputs
for txo in block_outputs:
batch.Delete(txo)
# add the max
- batch.Put('height', self.serialize([(block_hash, block_height, 0)]))
+ batch.Put('height', self.serialize([(block_hash, block_height, self.db_version)]))
# actual write
self.db.Write(batch, sync=sync)
"read:%0.2f " % (t1 - t00),
"proc:%.2f " % (t2-t1),
"write:%.2f " % (t3-t2),
- "max:", max_len, hash_160_to_bc_address(max_addr))
+ "max:", max_len, max_addr)
- for h160 in self.batch_list.keys():
- addr = hash_160_to_bc_address(h160)
+ for addr in self.batch_list.keys():
self.invalidate_cache(addr)
def add_request(self, request):
address = params[1]
if password == self.config.get('server', 'password'):
self.watched_addresses.remove(address)
- print_log('unsubscribed', address)
+ # print_log('unsubscribed', address)
result = "ok"
else:
print_log('incorrect password')
tx_height = params[1]
result = self.get_merkle(tx_hash, tx_height)
except BaseException, e:
- error = str(e) + ': ' + tx_hash
- print_log("error:", error)
+ error = str(e) + ': ' + repr(params)
+ print_log("get_merkle error:", error)
elif method == 'blockchain.transaction.get':
try:
height = params[1]
result = self.bitcoind('getrawtransaction', [tx_hash, 0, height])
except BaseException, e:
- error = str(e) + ': ' + tx_hash
- print_log("error:", error)
+ error = str(e) + ': ' + repr(params)
+ print_log("tx get error:", error)
else:
error = "unknown method:%s" % method
return -1
if error:
- response = {'id': message_id, 'error': error}
+ self.push_response({'id': message_id, 'error': error})
elif result != '':
- response = {'id': message_id, 'result': result}
- self.push_response(response)
+ self.push_response({'id': message_id, 'result': result})
def watch_address(self, addr):
if addr not in self.watched_addresses:
for x in tx.get('inputs'):
txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex')
try:
- h160 = self.db.Get(txi)
- addr = hash_160_to_bc_address(h160)
+ addr = self.db.Get(txi)
except:
- continue
+ tx_prev = self.get_mempool_transaction(x.get('prevout_hash'))
+ try:
+ addr = tx_prev['outputs'][x.get('prevout_n')]['address']
+ if not addr: continue
+ except:
+ continue
l = self.mempool_addresses.get(tx_hash, [])
if addr not in l:
l.append(addr)
if tx_hash not in self.mempool_hashes:
self.mempool_addresses.pop(tx_hash)
- # rebuild histories
+ # rebuild mempool histories
new_mempool_hist = {}
for tx_hash, addresses in self.mempool_addresses.items():
for addr in addresses:
h.append(tx_hash)
new_mempool_hist[addr] = h
- for addr in new_mempool_hist.keys():
- if addr in self.mempool_hist.keys():
+ # invalidate cache for mempool addresses whose mempool history has changed
+ new_mempool_hist_keys = new_mempool_hist.keys()
+ self_mempool_hist_keys = self.mempool_hist.keys()
+
+ for addr in new_mempool_hist_keys:
+ if addr in self_mempool_hist_keys:
if self.mempool_hist[addr] != new_mempool_hist[addr]:
self.invalidate_cache(addr)
else:
self.invalidate_cache(addr)
+ # invalidate cache for addresses that are removed from mempool ?
+ # this should not be necessary if they go into a block, but they might not
+ for addr in self_mempool_hist_keys:
+ if addr not in new_mempool_hist_keys:
+ self.invalidate_cache(addr)
+
+
with self.mempool_lock:
self.mempool_hist = new_mempool_hist
def invalidate_cache(self, address):
with self.cache_lock:
- if 'address' in self.history_cache:
+ if address in self.history_cache:
print_log("cache: invalidating", address)
self.history_cache.pop(address)
if address in self.watched_addresses:
+ # TODO: update cache here. if new value equals cached value, do not send notification
self.address_queue.put(address)
def main_iteration(self):