async def test_merkle_cache_truncation():
max_length = 33
source = Source(max_length).hashes
for length in range(max_length - 2, max_length + 1):
for trunc_length in range(1, 20, 3):
cache = MerkleCache(merkle, source)
await cache.initialize(length)
cache.truncate(trunc_length)
assert cache.length <= trunc_length
for cp_length in range(1, length + 1, 3):
cp_hashes = await source(0, cp_length)
# All possible indices
for index in range(cp_length):
# Compare correct answer with cache
branch, root = merkle.branch_and_root(cp_hashes, index)
branch2, root2 = await cache.branch_and_root(
cp_length, index)
assert branch == branch2
assert root == root2
# Truncation is a no-op if longer
cache = MerkleCache(merkle, source)
await cache.initialize(10)
level = cache.level.copy()
for length in range(10, 13):
cache.truncate(length)
assert cache.level == level
assert cache.length == 10
def __init__(self, env):
self.logger = util.class_logger(__name__, self.__class__.__name__)
self.env = env
self.coin = env.coin
# Setup block header size handlers
if self.coin.STATIC_BLOCK_HEADERS:
self.header_offset = self.coin.static_header_offset
self.header_len = self.coin.static_header_len
else:
self.header_offset = self.dynamic_header_offset
self.header_len = self.dynamic_header_len
self.logger.info(f'switching current directory to {env.db_dir}')
os.chdir(env.db_dir)
self.db_class = db_class(self.env.db_engine)
self.history = History()
self.utxo_db = None
self.utxo_flush_count = 0
self.fs_height = -1
self.fs_tx_count = 0
self.db_height = -1
self.db_tx_count = 0
self.db_tip = None
self.tx_counts = None
self.last_flush = time.time()
self.last_flush_tx_count = 0
self.wall_time = 0
self.first_sync = True
self.db_version = -1
self.logger.info(f'using {self.env.db_engine} for DB backend')
# Header merkle cache
self.merkle = Merkle()
self.header_mc = MerkleCache(self.merkle, self.fs_block_hashes)
self.headers_file = util.LogicalFile('meta/headers', 2, 16000000)
self.tx_counts_file = util.LogicalFile('meta/txcounts', 2, 2000000)
self.hashes_file = util.LogicalFile('meta/hashes', 4, 16000000)
if not self.coin.STATIC_BLOCK_HEADERS:
self.headers_offsets_file = util.LogicalFile(
'meta/headers_offsets', 2, 16000000)
async def test_bad_extension():
length = 5
source = Source(length).hashes
cache = MerkleCache(merkle, source)
await cache.initialize(length)
level = cache.level.copy()
with pytest.raises(AssertionError):
await cache.branch_and_root(8, 0)
# The bad extension should not destroy the cache
assert cache.level == level
assert cache.length == length
async def test_truncation_bad():
cache = MerkleCache(merkle, Source(10).hashes)
await cache.initialize(10)
with pytest.raises(TypeError):
cache.truncate(1.0)
for n in (-1, 0):
with pytest.raises(ValueError):
cache.truncate(n)
async def time_it():
source = Source(500000).hashes
cp_length = 492000
import time
cache = MerkleCache(merkle, source)
await cache.initialize(cp_length)
cp_hashes = await source(0, cp_length)
brs2 = []
t1 = time.time()
for index in range(5, 400000, 500):
brs2.append(await cache.branch_and_root(cp_length, index))
t2 = time.time()
print(t2 - t1)
assert False
async def test_merkle_cache_bad():
length = 23
source = Source(length).hashes
cache = MerkleCache(merkle, source)
await cache.initialize(length)
await cache.branch_and_root(5, 3)
with pytest.raises(TypeError):
await cache.branch_and_root(5.0, 3)
with pytest.raises(TypeError):
await cache.branch_and_root(5, 3.0)
with pytest.raises(ValueError):
await cache.branch_and_root(0, -1)
with pytest.raises(ValueError):
await cache.branch_and_root(3, 3)
async def test_merkle_cache_extension():
source = Source(64).hashes
for length in range(14, 18):
for cp_length in range(30, 36):
cache = MerkleCache(merkle, source)
await cache.initialize(length)
cp_hashes = await source(0, cp_length)
# All possible indices
for index in range(cp_length):
# Compare correct answer with cache
branch, root = merkle.branch_and_root(cp_hashes, index)
branch2, root2 = await cache.branch_and_root(cp_length, index)
assert branch == branch2
assert root == root2
async def test_merkle_cache():
lengths = (*range(1, 18), 31, 32, 33, 57)
source = Source(max(lengths)).hashes
for length in lengths:
cache = MerkleCache(merkle, source)
await cache.initialize(length)
# Simulate all possible checkpoints
for cp_length in range(1, length + 1):
cp_hashes = await source(0, cp_length)
# All possible indices
for index in range(cp_length):
# Compare correct answer with cache
branch, root = merkle.branch_and_root(cp_hashes, index)
branch2, root2 = await cache.branch_and_root(cp_length, index)
assert branch == branch2
assert root == root2
class DB(object):
'''Simple wrapper of the backend database for querying.
Performs no DB update, though the DB will be cleaned on opening if
it was shutdown uncleanly.
'''
DB_VERSIONS = [6, 7, 8]
class DBError(Exception):
'''Raised on general DB errors generally indicating corruption.'''
def __init__(self, env):
self.logger = util.class_logger(__name__, self.__class__.__name__)
self.env = env
self.coin = env.coin
# Setup block header size handlers
if self.coin.STATIC_BLOCK_HEADERS:
self.header_offset = self.coin.static_header_offset
self.header_len = self.coin.static_header_len
else:
self.header_offset = self.dynamic_header_offset
self.header_len = self.dynamic_header_len
self.logger.info(f'switching current directory to {env.db_dir}')
os.chdir(env.db_dir)
self.db_class = db_class(self.env.db_engine)
self.history = History()
self.utxo_db = None
self.utxo_flush_count = 0
self.fs_height = -1
self.fs_tx_count = 0
self.db_height = -1
self.db_tx_count = 0
self.db_tip = None
self.tx_counts = None
self.last_flush = time.time()
self.last_flush_tx_count = 0
self.wall_time = 0
self.first_sync = True
self.db_version = -1
self.logger.info(f'using {self.env.db_engine} for DB backend')
# Header merkle cache
self.merkle = Merkle()
self.header_mc = MerkleCache(self.merkle, self.fs_block_hashes)
self.headers_file = util.LogicalFile('meta/headers', 2, 16000000)
self.tx_counts_file = util.LogicalFile('meta/txcounts', 2, 2000000)
self.hashes_file = util.LogicalFile('meta/hashes', 4, 16000000)
if not self.coin.STATIC_BLOCK_HEADERS:
self.headers_offsets_file = util.LogicalFile(
'meta/headers_offsets', 2, 16000000)
async def _read_tx_counts(self):
if self.tx_counts is not None:
return
# tx_counts[N] has the cumulative number of txs at the end of
# height N. So tx_counts[0] is 1 - the genesis coinbase
size = (self.db_height + 1) * 8
tx_counts = self.tx_counts_file.read(0, size)
assert len(tx_counts) == size
self.tx_counts = array.array('Q', tx_counts)
if self.tx_counts:
assert self.db_tx_count == self.tx_counts[-1]
else:
assert self.db_tx_count == 0
async def _open_dbs(self, for_sync, compacting):
assert self.utxo_db is None
# First UTXO DB
self.utxo_db = self.db_class('utxo', for_sync)
if self.utxo_db.is_new:
self.logger.info('created new database')
self.logger.info('creating metadata directory')
os.mkdir('meta')
with util.open_file('COIN', create=True) as f:
f.write(f'ElectrumX databases and metadata for '
f'{self.coin.NAME} {self.coin.NET}'.encode())
if not self.coin.STATIC_BLOCK_HEADERS:
self.headers_offsets_file.write(0, bytes(8))
else:
self.logger.info(f'opened UTXO DB (for sync: {for_sync})')
self.read_utxo_state()
# Then history DB
self.utxo_flush_count = self.history.open_db(self.db_class, for_sync,
self.utxo_flush_count,
compacting)
self.clear_excess_undo_info()
# Read TX counts (requires meta directory)
await self._read_tx_counts()
async def open_for_compacting(self):
await self._open_dbs(True, True)
async def open_for_sync(self):
'''Open the databases to sync to the daemon.
When syncing we want to reserve a lot of open files for the
synchronization. When serving clients we want the open files for
serving network connections.
'''
await self._open_dbs(True, False)
async def open_for_serving(self):
'''Open the databases for serving. If they are already open they are
closed first.
'''
if self.utxo_db:
self.logger.info('closing DBs to re-open for serving')
self.utxo_db.close()
self.history.close_db()
self.utxo_db = None
await self._open_dbs(False, False)
# Header merkle cache
async def populate_header_merkle_cache(self):
self.logger.info('populating header merkle cache...')
length = max(1, self.db_height - self.env.reorg_limit)
start = time.time()
await self.header_mc.initialize(length)
elapsed = time.time() - start
self.logger.info(f'header merkle cache populated in {elapsed:.1f}s')
async def header_branch_and_root(self, length, height):
return await self.header_mc.branch_and_root(length, height)
# Flushing
def assert_flushed(self, flush_data):
'''Asserts state is fully flushed.'''
assert flush_data.tx_count == self.fs_tx_count == self.db_tx_count
assert flush_data.height == self.fs_height == self.db_height
assert flush_data.tip == self.db_tip
assert not flush_data.headers
assert not flush_data.block_tx_hashes
assert not flush_data.adds
assert not flush_data.deletes
assert not flush_data.undo_infos
self.history.assert_flushed()
def flush_dbs(self, flush_data, flush_utxos, estimate_txs_remaining):
'''Flush out cached state. History is always flushed; UTXOs are
flushed if flush_utxos.'''
if flush_data.height == self.db_height:
self.assert_flushed(flush_data)
return
start_time = time.time()
prior_flush = self.last_flush
tx_delta = flush_data.tx_count - self.last_flush_tx_count
# Flush to file system
self.flush_fs(flush_data)
# Then history
self.flush_history()
# Flush state last as it reads the wall time.
with self.utxo_db.write_batch() as batch:
if flush_utxos:
self.flush_utxo_db(batch, flush_data)
self.flush_state(batch)
# Update and put the wall time again - otherwise we drop the
# time it took to commit the batch
self.flush_state(self.utxo_db)
elapsed = self.last_flush - start_time
self.logger.info(f'flush #{self.history.flush_count:,d} took '
f'{elapsed:.1f}s. Height {flush_data.height:,d} '
f'txs: {flush_data.tx_count:,d} ({tx_delta:+,d})')
# Catch-up stats
if self.utxo_db.for_sync:
flush_interval = self.last_flush - prior_flush
tx_per_sec_gen = int(flush_data.tx_count / self.wall_time)
tx_per_sec_last = 1 + int(tx_delta / flush_interval)
eta = estimate_txs_remaining() / tx_per_sec_last
self.logger.info(f'tx/sec since genesis: {tx_per_sec_gen:,d}, '
f'since last flush: {tx_per_sec_last:,d}')
self.logger.info(f'sync time: {formatted_time(self.wall_time)} '
f'ETA: {formatted_time(eta)}')
def flush_fs(self, flush_data):
'''Write headers, tx counts and block tx hashes to the filesystem.
The first height to write is self.fs_height + 1. The FS
metadata is all append-only, so in a crash we just pick up
again from the height stored in the DB.
'''
prior_tx_count = (self.tx_counts[self.fs_height]
if self.fs_height >= 0 else 0)
assert len(flush_data.block_tx_hashes) == len(flush_data.headers)
assert flush_data.height == self.fs_height + len(flush_data.headers)
assert flush_data.tx_count == (self.tx_counts[-1]
if self.tx_counts else 0)
assert len(self.tx_counts) == flush_data.height + 1
hashes = b''.join(flush_data.block_tx_hashes)
flush_data.block_tx_hashes.clear()
assert len(hashes) % 32 == 0
assert len(hashes) // 32 == flush_data.tx_count - prior_tx_count
# Write the headers, tx counts, and tx hashes
start_time = time.time()
height_start = self.fs_height + 1
offset = self.header_offset(height_start)
self.headers_file.write(offset, b''.join(flush_data.headers))
self.fs_update_header_offsets(offset, height_start, flush_data.headers)
flush_data.headers.clear()
offset = height_start * self.tx_counts.itemsize
self.tx_counts_file.write(offset,
self.tx_counts[height_start:].tobytes())
offset = prior_tx_count * 32
self.hashes_file.write(offset, hashes)
self.fs_height = flush_data.height
self.fs_tx_count = flush_data.tx_count
if self.utxo_db.for_sync:
elapsed = time.time() - start_time
self.logger.info(f'flushed filesystem data in {elapsed:.2f}s')
def flush_history(self):
self.history.flush()
def flush_utxo_db(self, batch, flush_data):
'''Flush the cached DB writes and UTXO set to the batch.'''
# Care is needed because the writes generated by flushing the
# UTXO state may have keys in common with our write cache or
# may be in the DB already.
start_time = time.time()
add_count = len(flush_data.adds)
spend_count = len(flush_data.deletes) // 2
# Spends
batch_delete = batch.delete
for key in sorted(flush_data.deletes):
batch_delete(key)
flush_data.deletes.clear()
# New UTXOs
batch_put = batch.put
for key, value in flush_data.adds.items():
# suffix = tx_idx + tx_num
hashX = value[:-13]
suffix = key[-4:] + value[-13:-8]
batch_put(b'h' + key[:4] + suffix, hashX)
batch_put(b'u' + hashX + suffix, value[-8:])
flush_data.adds.clear()
# New undo information
self.flush_undo_infos(batch_put, flush_data.undo_infos)
flush_data.undo_infos.clear()
if self.utxo_db.for_sync:
block_count = flush_data.height - self.db_height
tx_count = flush_data.tx_count - self.db_tx_count
elapsed = time.time() - start_time
self.logger.info(f'flushed {block_count:,d} blocks with '
f'{tx_count:,d} txs, {add_count:,d} UTXO adds, '
f'{spend_count:,d} spends in '
f'{elapsed:.1f}s, committing...')
self.utxo_flush_count = self.history.flush_count
self.db_height = flush_data.height
self.db_tx_count = flush_data.tx_count
self.db_tip = flush_data.tip
def flush_state(self, batch):
'''Flush chain state to the batch.'''
now = time.time()
self.wall_time += now - self.last_flush
self.last_flush = now
self.last_flush_tx_count = self.fs_tx_count
self.write_utxo_state(batch)
def flush_backup(self, flush_data, touched):
'''Like flush_dbs() but when backing up. All UTXOs are flushed.'''
assert not flush_data.headers
assert not flush_data.block_tx_hashes
assert flush_data.height < self.db_height
self.history.assert_flushed()
start_time = time.time()
tx_delta = flush_data.tx_count - self.last_flush_tx_count
self.backup_fs(flush_data.height, flush_data.tx_count)
self.history.backup(touched, flush_data.tx_count)
with self.utxo_db.write_batch() as batch:
self.flush_utxo_db(batch, flush_data)
# Flush state last as it reads the wall time.
self.flush_state(batch)
elapsed = self.last_flush - start_time
self.logger.info(f'backup flush #{self.history.flush_count:,d} took '
f'{elapsed:.1f}s. Height {flush_data.height:,d} '
f'txs: {flush_data.tx_count:,d} ({tx_delta:+,d})')
def fs_update_header_offsets(self, offset_start, height_start, headers):
if self.coin.STATIC_BLOCK_HEADERS:
return
offset = offset_start
offsets = []
for h in headers:
offset += len(h)
offsets.append(pack_le_uint64(offset))
# For each header we get the offset of the next header, hence we
# start writing from the next height
pos = (height_start + 1) * 8
self.headers_offsets_file.write(pos, b''.join(offsets))
def dynamic_header_offset(self, height):
assert not self.coin.STATIC_BLOCK_HEADERS
offset, = unpack_le_uint64(
self.headers_offsets_file.read(height * 8, 8))
return offset
def dynamic_header_len(self, height):
return self.dynamic_header_offset(height + 1)\
- self.dynamic_header_offset(height)
def backup_fs(self, height, tx_count):
'''Back up during a reorg. This just updates our pointers.'''
self.fs_height = height
self.fs_tx_count = tx_count
# Truncate header_mc: header count is 1 more than the height.
self.header_mc.truncate(height + 1)
async def raw_header(self, height):
'''Return the binary header at the given height.'''
header, n = await self.read_headers(height, 1)
if n != 1:
raise IndexError(f'height {height:,d} out of range')
return header
async def read_headers(self, start_height, count):
'''Requires start_height >= 0, count >= 0. Reads as many headers as
are available starting at start_height up to count. This
would be zero if start_height is beyond self.db_height, for
example.
Returns a (binary, n) pair where binary is the concatenated
binary headers, and n is the count of headers returned.
'''
if start_height < 0 or count < 0:
raise self.DBError(f'{count:,d} headers starting at '
f'{start_height:,d} not on disk')
def read_headers():
# Read some from disk
disk_count = max(0, min(count, self.db_height + 1 - start_height))
if disk_count:
offset = self.header_offset(start_height)
size = self.header_offset(start_height + disk_count) - offset
return self.headers_file.read(offset, size), disk_count
return b'', 0
return await run_in_thread(read_headers)
def fs_tx_hash(self, tx_num):
'''Return a pair (tx_hash, tx_height) for the given tx number.
If the tx_height is not on disk, returns (None, tx_height).'''
tx_height = bisect_right(self.tx_counts, tx_num)
if tx_height > self.db_height:
tx_hash = None
else:
tx_hash = self.hashes_file.read(tx_num * 32, 32)
return tx_hash, tx_height
def fs_tx_hashes_at_blockheight(self, block_height):
'''Return a list of tx_hashes at given block height,
in the same order as in the block.
'''
if block_height > self.db_height:
raise self.DBError(
f'block {block_height:,d} not on disk (>{self.db_height:,d})')
assert block_height >= 0
if block_height > 0:
first_tx_num = self.tx_counts[block_height - 1]
else:
first_tx_num = 0
num_txs_in_block = self.tx_counts[block_height] - first_tx_num
tx_hashes = self.hashes_file.read(first_tx_num * 32,
num_txs_in_block * 32)
assert num_txs_in_block == len(tx_hashes) // 32
return [
tx_hashes[idx * 32:(idx + 1) * 32]
for idx in range(num_txs_in_block)
]
async def tx_hashes_at_blockheight(self, block_height):
return await run_in_thread(self.fs_tx_hashes_at_blockheight,
block_height)
async def fs_block_hashes(self, height, count):
headers_concat, headers_count = await self.read_headers(height, count)
if headers_count != count:
raise self.DBError('only got {:,d} headers starting at {:,d}, not '
'{:,d}'.format(headers_count, height, count))
offset = 0
headers = []
for n in range(count):
hlen = self.header_len(height + n)
headers.append(headers_concat[offset:offset + hlen])
offset += hlen
return [self.coin.header_hash(header) for header in headers]
async def limited_history(self, hashX, *, limit=1000):
'''Return an unpruned, sorted list of (tx_hash, height) tuples of
confirmed transactions that touched the address, earliest in
the blockchain first. Includes both spending and receiving
transactions. By default returns at most 1000 entries. Set
limit to None to get them all.
'''
def read_history():
tx_nums = list(self.history.get_txnums(hashX, limit))
fs_tx_hash = self.fs_tx_hash
return [fs_tx_hash(tx_num) for tx_num in tx_nums]
while True:
history = await run_in_thread(read_history)
if all(hash is not None for hash, height in history):
return history
self.logger.warning(f'limited_history: tx hash '
f'not found (reorg?), retrying...')
await sleep(0.25)
# -- Undo information
def min_undo_height(self, max_height):
'''Returns a height from which we should store undo info.'''
return max_height - self.env.reorg_limit + 1
def undo_key(self, height):
'''DB key for undo information at the given height.'''
return b'U' + pack_be_uint32(height)
def read_undo_info(self, height):
'''Read undo information from a file for the current height.'''
return self.utxo_db.get(self.undo_key(height))
def flush_undo_infos(self, batch_put, undo_infos):
'''undo_infos is a list of (undo_info, height) pairs.'''
for undo_info, height in undo_infos:
batch_put(self.undo_key(height), b''.join(undo_info))
def raw_block_prefix(self):
return 'meta/block'
def raw_block_path(self, height):
return f'{self.raw_block_prefix()}{height:d}'
def read_raw_block(self, height):
'''Returns a raw block read from disk. Raises FileNotFoundError
if the block isn't on-disk.'''
with util.open_file(self.raw_block_path(height)) as f:
return f.read(-1)
def write_raw_block(self, block, height):
'''Write a raw block to disk.'''
with util.open_truncate(self.raw_block_path(height)) as f:
f.write(block)
# Delete old blocks to prevent them accumulating
try:
del_height = self.min_undo_height(height) - 1
os.remove(self.raw_block_path(del_height))
except FileNotFoundError:
pass
def clear_excess_undo_info(self):
'''Clear excess undo info. Only most recent N are kept.'''
prefix = b'U'
min_height = self.min_undo_height(self.db_height)
keys = []
for key, _hist in self.utxo_db.iterator(prefix=prefix):
height, = unpack_be_uint32(key[-4:])
if height >= min_height:
break
keys.append(key)
if keys:
with self.utxo_db.write_batch() as batch:
for key in keys:
batch.delete(key)
self.logger.info(f'deleted {len(keys):,d} stale undo entries')
# delete old block files
prefix = self.raw_block_prefix()
paths = [
path for path in glob(f'{prefix}[0-9]*')
if len(path) > len(prefix) and int(path[len(prefix):]) < min_height
]
if paths:
for path in paths:
try:
os.remove(path)
except FileNotFoundError:
pass
self.logger.info(f'deleted {len(paths):,d} stale block files')
# -- UTXO database
def read_utxo_state(self):
state = self.utxo_db.get(b'state')
if not state:
self.db_height = -1
self.db_tx_count = 0
self.db_tip = b'\0' * 32
self.db_version = max(self.DB_VERSIONS)
self.utxo_flush_count = 0
self.wall_time = 0
self.first_sync = True
else:
state = ast.literal_eval(state.decode())
if not isinstance(state, dict):
raise self.DBError('failed reading state from DB')
self.db_version = state['db_version']
if self.db_version not in self.DB_VERSIONS:
raise self.DBError('your UTXO DB version is {} but this '
'software only handles versions {}'.format(
self.db_version, self.DB_VERSIONS))
# backwards compat
genesis_hash = state['genesis']
if isinstance(genesis_hash, bytes):
genesis_hash = genesis_hash.decode()
if genesis_hash != self.coin.GENESIS_HASH:
raise self.DBError(
'DB genesis hash {} does not match coin {}'.format(
genesis_hash, self.coin.GENESIS_HASH))
self.db_height = state['height']
self.db_tx_count = state['tx_count']
self.db_tip = state['tip']
self.utxo_flush_count = state['utxo_flush_count']
self.wall_time = state['wall_time']
self.first_sync = state['first_sync']
# These are our state as we move ahead of DB state
self.fs_height = self.db_height
self.fs_tx_count = self.db_tx_count
self.last_flush_tx_count = self.fs_tx_count
# Upgrade DB
if self.db_version != max(self.DB_VERSIONS):
self.upgrade_db()
# Log some stats
self.logger.info('UTXO DB version: {:d}'.format(self.db_version))
self.logger.info('coin: {}'.format(self.coin.NAME))
self.logger.info('network: {}'.format(self.coin.NET))
self.logger.info('height: {:,d}'.format(self.db_height))
self.logger.info('tip: {}'.format(hash_to_hex_str(self.db_tip)))
self.logger.info('tx count: {:,d}'.format(self.db_tx_count))
if self.utxo_db.for_sync:
self.logger.info(f'flushing DB cache at {self.env.cache_MB:,d} MB')
if self.first_sync:
self.logger.info('sync time so far: {}'.format(
util.formatted_time(self.wall_time)))
def upgrade_db(self):
self.logger.info(f'UTXO DB version: {self.db_version}')
self.logger.info('Upgrading your DB; this can take some time...')
def upgrade_u_prefix(prefix):
count = 0
with self.utxo_db.write_batch() as batch:
batch_delete = batch.delete
batch_put = batch.put
# Key: b'u' + address_hashX + tx_idx + tx_num
for db_key, db_value in self.utxo_db.iterator(prefix=prefix):
if len(db_key) == 21:
return
break
if self.db_version == 6:
for db_key, db_value in self.utxo_db.iterator(
prefix=prefix):
count += 1
batch_delete(db_key)
batch_put(db_key[:14] + b'\0\0' + db_key[14:] + b'\0',
db_value)
else:
for db_key, db_value in self.utxo_db.iterator(
prefix=prefix):
count += 1
batch_delete(db_key)
batch_put(db_key + b'\0', db_value)
return count
last = time.time()
count = 0
for cursor in range(65536):
prefix = b'u' + pack_be_uint16(cursor)
count += upgrade_u_prefix(prefix)
now = time.time()
if now > last + 10:
last = now
self.logger.info(f'DB 1 of 3: {count:,d} entries updated, '
f'{cursor * 100 / 65536:.1f}% complete')
self.logger.info('DB 1 of 3 upgraded successfully')
def upgrade_h_prefix(prefix):
count = 0
with self.utxo_db.write_batch() as batch:
batch_delete = batch.delete
batch_put = batch.put
# Key: b'h' + compressed_tx_hash + tx_idx + tx_num
for db_key, db_value in self.utxo_db.iterator(prefix=prefix):
if len(db_key) == 14:
return
break
if self.db_version == 6:
for db_key, db_value in self.utxo_db.iterator(
prefix=prefix):
count += 1
batch_delete(db_key)
batch_put(db_key[:7] + b'\0\0' + db_key[7:] + b'\0',
db_value)
else:
for db_key, db_value in self.utxo_db.iterator(
prefix=prefix):
count += 1
batch_delete(db_key)
batch_put(db_key + b'\0', db_value)
return count
last = time.time()
count = 0
for cursor in range(65536):
prefix = b'h' + pack_be_uint16(cursor)
count += upgrade_h_prefix(prefix)
now = time.time()
if now > last + 10:
last = now
self.logger.info(f'DB 2 of 3: {count:,d} entries updated, '
f'{cursor * 100 / 65536:.1f}% complete')
# Upgrade tx_counts file
size = (self.db_height + 1) * 8
tx_counts = self.tx_counts_file.read(0, size)
if len(tx_counts) == (self.db_height + 1) * 4:
tx_counts = array.array('I', tx_counts)
tx_counts = array.array('Q', tx_counts)
self.tx_counts_file.write(0, tx_counts.tobytes())
self.db_version = max(self.DB_VERSIONS)
with self.utxo_db.write_batch() as batch:
self.write_utxo_state(batch)
self.logger.info('DB 2 of 3 upgraded successfully')
def write_utxo_state(self, batch):
'''Write (UTXO) state to the batch.'''
state = {
'genesis': self.coin.GENESIS_HASH,
'height': self.db_height,
'tx_count': self.db_tx_count,
'tip': self.db_tip,
'utxo_flush_count': self.utxo_flush_count,
'wall_time': self.wall_time,
'first_sync': self.first_sync,
'db_version': self.db_version,
}
batch.put(b'state', repr(state).encode())
def set_flush_count(self, count):
self.utxo_flush_count = count
with self.utxo_db.write_batch() as batch:
self.write_utxo_state(batch)
async def all_utxos(self, hashX):
'''Return all UTXOs for an address sorted in no particular order.'''
def read_utxos():
utxos = []
utxos_append = utxos.append
# Key: b'u' + address_hashX + tx_idx + tx_num
# Value: the UTXO value as a 64-bit unsigned integer
prefix = b'u' + hashX
for db_key, db_value in self.utxo_db.iterator(prefix=prefix):
tx_pos, = unpack_le_uint32(db_key[-9:-5])
tx_num, = unpack_le_uint64(db_key[-5:] + bytes(3))
value, = unpack_le_uint64(db_value)
tx_hash, height = self.fs_tx_hash(tx_num)
utxos_append(UTXO(tx_num, tx_pos, tx_hash, height, value))
return utxos
while True:
utxos = await run_in_thread(read_utxos)
if all(utxo.tx_hash is not None for utxo in utxos):
return utxos
self.logger.warning(f'all_utxos: tx hash not '
f'found (reorg?), retrying...')
await sleep(0.25)
async def lookup_utxos(self, prevouts):
'''For each prevout, lookup it up in the DB and return a (hashX,
value) pair or None if not found.
Used by the mempool code.
'''
def lookup_hashXs():
'''Return (hashX, suffix) pairs, or None if not found,
for each prevout.
'''
def lookup_hashX(tx_hash, tx_idx):
idx_packed = pack_le_uint32(tx_idx)
# Key: b'h' + compressed_tx_hash + tx_idx + tx_num
# Value: hashX
prefix = b'h' + tx_hash[:4] + idx_packed
# Find which entry, if any, the TX_HASH matches.
for db_key, hashX in self.utxo_db.iterator(prefix=prefix):
tx_num_packed = db_key[-5:]
tx_num, = unpack_le_uint64(tx_num_packed + bytes(3))
hash, _height = self.fs_tx_hash(tx_num)
if hash == tx_hash:
return hashX, idx_packed + tx_num_packed
return None, None
return [lookup_hashX(*prevout) for prevout in prevouts]
def lookup_utxos(hashX_pairs):
def lookup_utxo(hashX, suffix):
if not hashX:
# This can happen when the daemon is a block ahead
# of us and has mempool txs spending outputs from
# that new block
return None
# Key: b'u' + address_hashX + tx_idx + tx_num
# Value: the UTXO value as a 64-bit unsigned integer
key = b'u' + hashX + suffix
db_value = self.utxo_db.get(key)
if not db_value:
# This can happen if the DB was updated between
# getting the hashXs and getting the UTXOs
return None
value, = unpack_le_uint64(db_value)
return hashX, value
return [lookup_utxo(*hashX_pair) for hashX_pair in hashX_pairs]
hashX_pairs = await run_in_thread(lookup_hashXs)
return await run_in_thread(lookup_utxos, hashX_pairs)