import asyncio import logging import typing if typing.TYPE_CHECKING: from scribe.blockchain.network import LBCMainNet from scribe.blockchain.daemon import LBCDaemon def chunks(items, size): """Break up items, an iterable, into chunks of length size.""" for i in range(0, len(items), size): yield items[i: i + size] class Prefetcher: """Prefetches blocks (in the forward direction only).""" def __init__(self, daemon: 'LBCDaemon', coin: 'LBCMainNet', blocks_event: asyncio.Event): self.logger = logging.getLogger(__name__) self.daemon = daemon self.coin = coin self.blocks_event = blocks_event self.blocks = [] self.caught_up = False # Access to fetched_height should be protected by the semaphore self.fetched_height = None self.semaphore = asyncio.Semaphore() self.refill_event = asyncio.Event() # The prefetched block cache size. The min cache size has # little effect on sync time. self.cache_size = 0 self.min_cache_size = 10 * 1024 * 1024 # This makes the first fetch be 10 blocks self.ave_size = self.min_cache_size // 10 self.polling_delay = 0.5 async def main_loop(self, bp_height, started: asyncio.Event): """Loop forever polling for more blocks.""" await self.reset_height(bp_height) started.set() try: while True: # Sleep a while if there is nothing to prefetch await self.refill_event.wait() if not await self._prefetch_blocks(): await asyncio.sleep(self.polling_delay) except Exception as e: if not isinstance(e, asyncio.CancelledError): self.logger.exception("block fetcher loop crashed") raise finally: self.logger.info("block pre-fetcher is shutting down") def get_prefetched_blocks(self): """Called by block processor when it is processing queued blocks.""" blocks = self.blocks self.blocks = [] self.cache_size = 0 self.refill_event.set() return blocks async def reset_height(self, height): """Reset to prefetch blocks from the block processor's height. Used in blockchain reorganisations. This coroutine can be called asynchronously to the _prefetch_blocks coroutine so we must synchronize with a semaphore. """ async with self.semaphore: self.blocks.clear() self.cache_size = 0 self.fetched_height = height self.refill_event.set() daemon_height = await self.daemon.height() behind = daemon_height - height if behind > 0: self.logger.info(f'catching up to daemon height {daemon_height:,d} ' f'({behind:,d} blocks behind)') else: self.logger.info(f'caught up to daemon height {daemon_height:,d}') async def _prefetch_blocks(self): """Prefetch some blocks and put them on the queue. Repeats until the queue is full or caught up. """ daemon = self.daemon daemon_height = await daemon.height() async with self.semaphore: while self.cache_size < self.min_cache_size: # Try and catch up all blocks but limit to room in cache. # Constrain fetch count to between 0 and 500 regardless; # testnet can be lumpy. cache_room = self.min_cache_size // self.ave_size count = min(daemon_height - self.fetched_height, cache_room) count = min(500, max(count, 0)) if not count: self.caught_up = True return False first = self.fetched_height + 1 hex_hashes = await daemon.block_hex_hashes(first, count) if self.caught_up: self.logger.info('new block height {:,d} hash {}' .format(first + count-1, hex_hashes[-1])) blocks = await daemon.raw_blocks(hex_hashes) assert count == len(blocks) # Special handling for genesis block if first == 0: blocks[0] = self.coin.genesis_block(blocks[0]) self.logger.info(f'verified genesis block with hash {hex_hashes[0]}') # Update our recent average block size estimate size = sum(len(block) for block in blocks) if count >= 10: self.ave_size = size // count else: self.ave_size = (size + (10 - count) * self.ave_size) // 10 self.blocks.extend(blocks) self.cache_size += size self.fetched_height += count self.blocks_event.set() self.refill_event.clear() return True