fix
[electrum-server.git] / backends / bitcoind / blockchain_processor.py
index 24aa0d0..e907e85 100644 (file)
@@ -3,6 +3,14 @@ import leveldb, urllib
 import deserialize
 import ast, time, threading, hashlib
 from Queue import Queue
+import traceback, sys, os, random
+
+
+
+Hash = lambda x: hashlib.sha256(hashlib.sha256(x).digest()).digest()
+hash_encode = lambda x: x[::-1].encode('hex')
+hash_decode = lambda x: x.decode('hex')[::-1]
+
 
 
 def rev_hex(s):
@@ -14,27 +22,57 @@ def int_to_hex(i, length=1):
     s = "0"*(2*length - len(s)) + s
     return rev_hex(s)
 
+def header_to_string(res):
+    pbh = res.get('prev_block_hash')
+    if pbh is None: pbh = '0'*64
+    s = int_to_hex(res.get('version'),4) \
+        + rev_hex(pbh) \
+        + rev_hex(res.get('merkle_root')) \
+        + int_to_hex(int(res.get('timestamp')),4) \
+        + int_to_hex(int(res.get('bits')),4) \
+        + int_to_hex(int(res.get('nonce')),4)
+    return s
+
+def header_from_string( s):
+    hex_to_int = lambda s: eval('0x' + s[::-1].encode('hex'))
+    h = {}
+    h['version'] = hex_to_int(s[0:4])
+    h['prev_block_hash'] = hash_encode(s[4:36])
+    h['merkle_root'] = hash_encode(s[36:68])
+    h['timestamp'] = hex_to_int(s[68:72])
+    h['bits'] = hex_to_int(s[72:76])
+    h['nonce'] = hex_to_int(s[76:80])
+    return h
 
-from processor import Processor, print_log
 
 
-class Blockchain2Processor(Processor):
 
-    def __init__(self, config):
+from processor import Processor, print_log
+
+class BlockchainProcessor(Processor):
+
+    def __init__(self, config, shared):
         Processor.__init__(self)
 
+        self.shared = shared
+        self.up_to_date = False
         self.watched_addresses = []
         self.history_cache = {}
         self.chunk_cache = {}
         self.cache_lock = threading.Lock()
+        self.headers_data = ''
 
+        self.mempool_addresses = {}
         self.mempool_hist = {}
-        self.known_mempool_hashes = []
+        self.mempool_hashes = []
+        self.mempool_lock = threading.Lock()
+
         self.address_queue = Queue()
+        self.dbpath = config.get('leveldb', 'path')
 
         self.dblock = threading.Lock()
         try:
-            self.db = leveldb.LevelDB(config.get('leveldb', 'path'))
+            self.db = leveldb.LevelDB(self.dbpath)
         except:
             traceback.print_exc(file=sys.stdout)
             self.shared.stop()
@@ -46,23 +84,39 @@ class Blockchain2Processor(Processor):
             config.get('bitcoind','port'))
 
         self.height = 0
+        self.is_test = False
         self.sent_height = 0
         self.sent_header = None
 
-        # catch_up first
+
         try:
-            hist = self.deserialize(self.db.Get('0'))
-            hh, self.height = hist[0] 
-            self.block_hashes = [hh]
+            hist = self.deserialize(self.db.Get('height'))
+            self.last_hash, self.height, _ = hist[0] 
             print_log( "hist", hist )
         except:
-            traceback.print_exc(file=sys.stdout)
+            #traceback.print_exc(file=sys.stdout)
+            print_log('initializing database')
             self.height = 0
-            self.block_hashes = [ '000000000019d6689c085ae165831e934ff763ae46a2a6c172b3f1b60a8ce26f' ]
+            self.last_hash = '000000000019d6689c085ae165831e934ff763ae46a2a6c172b3f1b60a8ce26f'
+
+        # catch_up headers
+        self.init_headers(self.height)
+
+        threading.Timer(0, lambda: self.catch_up(sync=False)).start()
+        while not shared.stopped() and not self.up_to_date:
+            try:
+                time.sleep(1)
+            except:
+                print "keyboard interrupt: stopping threads"
+                shared.stop()
+                sys.exit(0)
+
+        print_log( "blockchain is up to date." )
 
         threading.Timer(10, self.main_iteration).start()
 
 
+
     def bitcoind(self, method, params=[]):
         postdata = dumps({"method": method, 'params': params, 'id':'jsonrpc'})
         respdata = urllib.urlopen(self.bitcoind_url, postdata).read()
@@ -74,24 +128,26 @@ class Blockchain2Processor(Processor):
 
     def serialize(self, h):
         s = ''
-        for txid, height in h:
-            s += txid + int_to_hex(height, 4)
+        for txid, txpos, height in h:
+            s += txid + int_to_hex(txpos, 4) + int_to_hex(height, 4)
         return s.decode('hex')
 
+
     def deserialize(self, s):
         h = []
         while s:
             txid = s[0:32].encode('hex')
-            height = s[32:36].encode('hex')
-            height = int( rev_hex( height ), 16 )
-            h.append( ( txid, height ) )
-            s = s[36:]
+            txpos = int( rev_hex( s[32:36].encode('hex') ), 16 )
+            height = int( rev_hex( s[36:40].encode('hex') ), 16 )
+            h.append( ( txid, txpos, height ) )
+            s = s[40:]
         return h
 
 
     def block2header(self, b):
         return {"block_height":b.get('height'), "version":b.get('version'), "prev_block_hash":b.get('previousblockhash'), 
-                "merkle_root":b.get('merkleroot'), "timestamp":b.get('time'), "bits":b.get('bits'), "nonce":b.get('nonce')}
+                "merkle_root":b.get('merkleroot'), "timestamp":b.get('time'), "bits":int(b.get('bits'),16), "nonce":b.get('nonce')}
+
 
     def get_header(self, height):
         block_hash = self.bitcoind('getblockhash', [height])
@@ -99,16 +155,100 @@ class Blockchain2Processor(Processor):
         return self.block2header(b)
     
 
-    def get_chunk(self):
+    def init_headers(self, db_height):
+        self.chunk_cache = {}
+        self.headers_filename = os.path.join( self.dbpath, 'blockchain_headers')
+
+        if os.path.exists(self.headers_filename):
+            height = os.path.getsize(self.headers_filename)/80 - 1   # the current height
+            if height > 0:
+                prev_hash = self.hash_header(self.read_header(height))
+            else:
+                prev_hash = None
+        else:
+            open(self.headers_filename,'wb').close()
+            prev_hash = None
+            height = -1
+
+        if height < db_height:
+            print_log( "catching up missing headers:", height, db_height)
+
+        try:
+            while height < db_height:
+                height = height + 1
+                header = self.get_header(height)
+                if height>1: 
+                    assert prev_hash == header.get('prev_block_hash')
+                self.write_header(header, sync=False)
+                prev_hash = self.hash_header(header)
+                if height%1000==0: print_log("headers file:",height)
+        except KeyboardInterrupt:
+            self.flush_headers()
+            sys.exit()
+
+        self.flush_headers()
+
+
+    def hash_header(self, header):
+        return rev_hex(Hash(header_to_string(header).decode('hex')).encode('hex'))
+
+
+    def read_header(self, block_height):
+        if os.path.exists(self.headers_filename):
+            f = open(self.headers_filename,'rb')
+            f.seek(block_height*80)
+            h = f.read(80)
+            f.close()
+            if len(h) == 80:
+                h = header_from_string(h)
+                return h
+
+
+    def read_chunk(self, index):
+        f = open(self.headers_filename,'rb')
+        f.seek(index*2016*80)
+        chunk = f.read(2016*80)
+        f.close()
+        return chunk.encode('hex')
+
+
+    def write_header(self, header, sync=True):
+        if not self.headers_data:
+            self.headers_offset = header.get('block_height')
+
+        self.headers_data += header_to_string(header).decode('hex')
+        if sync or len(self.headers_data) > 40*100:
+            self.flush_headers()
+
+    def pop_header(self):
+        # we need to do this only if we have not flushed
+        if self.headers_data:
+            self.headers_data = self.headers_data[:-40]
+
+    def flush_headers(self):
+        if not self.headers_data: return
+        f = open(self.headers_filename,'rb+')
+        f.seek(self.headers_offset*80)
+        f.write(self.headers_data)
+        f.close()
+        self.headers_data = ''
+
+
+    def get_chunk(self, i):
         # store them on disk; store the current chunk in memory
-        pass
+        chunk = self.chunk_cache.get(i)
+        if not chunk:
+            chunk = self.read_chunk(i)
+            self.chunk_cache[i] = chunk
+        return chunk
 
 
-    def get_transaction(self, txid, block_height=-1):
+    def get_transaction(self, txid, block_height=-1, is_coinbase = False):
         raw_tx = self.bitcoind('getrawtransaction', [txid, 0, block_height])
         vds = deserialize.BCDataStream()
         vds.write(raw_tx.decode('hex'))
-        return deserialize.parse_Transaction(vds)
+        out = deserialize.parse_Transaction(vds, is_coinbase)
+        return out
 
 
     def get_history(self, addr, cache_only=False):
@@ -119,18 +259,24 @@ class Blockchain2Processor(Processor):
         with self.dblock:
             try:
                 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...
 
         # add memory pool
-        for txid in self.mempool_hist.get(addr,[]):
-            hist.append((txid, 0))
+        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)
+        # add something to distinguish between unused and empty addresses
+        if hist == [] and is_known: hist = ['*']
 
-        hist = map(lambda x: {'tx_hash':x[0], 'height':x[1]}, hist)
         with self.cache_lock: self.history_cache[addr] = hist
         return hist
 
@@ -140,18 +286,22 @@ class Blockchain2Processor(Processor):
         if cache_only and tx_points == -1: return -1
 
         if not tx_points: return None
+        if tx_points == ['*']: return '*'
         status = ''
         for tx in tx_points:
             status += tx.get('tx_hash') + ':%d:' % tx.get('height')
         return hashlib.sha256( status ).digest().encode('hex')
 
 
-    def get_merkle(self, target_hash, height):
+    def get_merkle(self, tx_hash, height):
 
         block_hash = self.bitcoind('getblockhash', [height])
         b = self.bitcoind('getblock', [block_hash])
-        merkle = b.get('tx')
-
+        tx_list = b.get('tx')
+        tx_pos = tx_list.index(tx_hash)
+        
+        merkle = map(hash_decode, tx_list)
+        target_hash = hash_decode(tx_hash)
         s = []
         while len(merkle) != 1:
             if len(merkle)%2: merkle.append( merkle[-1] )
@@ -159,10 +309,10 @@ class Blockchain2Processor(Processor):
             while merkle:
                 new_hash = Hash( merkle[0] + merkle[1] )
                 if merkle[0] == target_hash:
-                    s.append( merkle[1])
+                    s.append( hash_encode( merkle[1]))
                     target_hash = new_hash
                 elif merkle[1] == target_hash:
-                    s.append( merkle[0])
+                    s.append( hash_encode( merkle[0]))
                     target_hash = new_hash
                 n.append( new_hash )
                 merkle = merkle[2:]
@@ -173,108 +323,217 @@ class Blockchain2Processor(Processor):
         
 
 
-    def import_block(self, block, block_hash, block_height):
-        #print "importing block", block_hash, block_height
+    def add_to_history(self, addr, tx_hash, tx_pos, tx_height):
 
-        txlist = block.get('tx')
-        batch_list = {}
+        # keep it sorted
+        s = (tx_hash + int_to_hex(tx_pos, 4) + int_to_hex(tx_height, 4)).decode('hex')
 
-        for txid in txlist:
-            tx = self.get_transaction(txid, block_height)
-            for x in tx.get('inputs') + tx.get('outputs'):
-                addr = x.get('address')
-                serialized_hist = batch_list.get(addr)
-                if serialized_hist is None:
-                    try:
-                        serialized_hist = self.db.Get(addr)
-                    except: 
-                        serialized_hist = ''
-
-                s = (txid + int_to_hex(block_height, 4)).decode('hex')
-
-                found = False
-                for i in range(len(serialized_hist)/36):
-                    item = serialized_hist[-36*(1+i):]
-                    item = item[0:36]
-
-                    h = int( rev_hex( item[32:36].encode('hex') ), 16 )
-                    if h > block_height:
-                        txhash = item[0:32].encode('hex')
-                        print_log('warning: non-chronological order at', addr, (txhash, h), (txid, block_height))
-                        hist = self.deserialize(serialized_hist)
-                        print_log(hist)
-                        hist.sort( key=lambda tup: tup[1])
-                        while hist:
-                            last = hist[-1]
-                            if last[1] > block_height:
-                                hist = hist[0:-1]
-                            else:
-                                break
-                        found = (txhash, h) in hist
-                        print_log('new sorted hist', hist, found)
-                        serialized_hist = self.serialize(hist)
-                        break
-                    elif h < block_height:
-                        break
-                    elif item == s:
-                        found = True
-                        break
-
-                if not found:
-                    serialized_hist += s
-
-                batch_list[addr] = serialized_hist
-
-        # batch write
-        batch = leveldb.WriteBatch()
-        for addr, hist in batch_list.items():
-            batch.Put(addr, serialized_hist)
-        batch.Put('0', self.serialize( [(block_hash, block_height)] ) )
-        self.db.Write(batch, sync = True)
+        serialized_hist = self.batch_list[addr] 
+
+        l = len(serialized_hist)/40
+        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):]
+                break
+        else:
+            serialized_hist = s + serialized_hist
+
+        self.batch_list[addr] = serialized_hist
+
+        # backlink
+        txo = (tx_hash + int_to_hex(tx_pos, 4)).decode('hex')
+        self.batch_txio[txo] = addr
 
-        # invalidate cache
-        for addr in batch_list.keys(): self.update_history_cache(addr)
 
-        return len(txlist)
+    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)
+        
+        serialized_hist = self.batch_list[addr]
+
+        l = len(serialized_hist)/40
+        for i in range(l):
+            item = serialized_hist[40*i:40*(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):]
+                break
+        else:
+            hist = self.deserialize(serialized_hist)
+            raise BaseException("prevout not found", addr, hist, tx_hash, tx_pos)
 
+        self.batch_list[addr] = serialized_hist
+        return height, addr
 
-    def revert_block(self, block, block_hash, block_height):
 
+    def deserialize_block(self, block):
         txlist = block.get('tx')
-        batch_list = {}
+        tx_hashes = []  # ordered txids
+        txdict = {}     # deserialized tx
+        is_coinbase = True
+        for raw_tx in txlist:
+            tx_hash = hash_encode(Hash(raw_tx.decode('hex')))
+            tx_hashes.append(tx_hash)
+            vds = deserialize.BCDataStream()
+            vds.write(raw_tx.decode('hex'))
+            tx = deserialize.parse_Transaction(vds, is_coinbase)
+            txdict[tx_hash] = tx
+            is_coinbase = False
+        return tx_hashes, txdict
 
-        for txid in txlist:
-            tx = self.get_transaction(txid, block_height)
-            for x in tx.get('inputs') + tx.get('outputs'):
+    def get_undo_info(self, height):
+        s = self.db.Get("undo%d"%(height%100))
+        return eval(s)
 
-                addr = x.get('address')
+    def write_undo_info(self, batch, height, undo_info):
+        batch.Put("undo%d"%(height%100), repr(undo_info))
+
+
+    def import_block(self, block, block_hash, block_height, sync, revert=False):
 
-                hist = batch_list.get(addr)
-                if hist is None:
-                    try:
-                        hist = self.deserialize(self.db.Get(addr))
-                    except: 
-                        hist = []
+        self.batch_list = {}  # address -> history
+        self.batch_txio = {}  # transaction i/o -> address
 
-                if (txid, block_height) in hist:
-                    hist.remove( (txid, block_height) )
-                else:
-                    print "error: txid not found during block revert", txid, block_height
+        block_inputs = []
+        block_outputs = []
+        addr_to_read = []
 
-                batch_list[addr] = hist
+        # deserialize transactions
+        t0 = time.time()
+        tx_hashes, txdict = self.deserialize_block(block)
+
+        t00 = time.time()
+
+
+        if not revert:
+            # read addresses of tx inputs
+            for tx in txdict.values():
+                for x in tx.get('inputs'):
+                    txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex')
+                    block_inputs.append(txi)
+
+            block_inputs.sort()
+            for txi in block_inputs:
+                try:
+                    addr = self.db.Get(txi)
+                except:
+                    # the input could come from the same block
+                    continue
+                self.batch_txio[txi] = addr
+                addr_to_read.append(addr)
+
+        else:
+            for txid, tx in txdict.items():
+                for x in tx.get('outputs'):
+                    txo = (txid + int_to_hex(x.get('index'), 4)).decode('hex')
+                    block_outputs.append(txo)
+            
+
+
+        # read histories of addresses
+        for txid, tx in txdict.items():
+            for x in tx.get('outputs'):
+                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: 
+                self.batch_list[addr] = ''
+
+
+        if revert: 
+            undo_info = self.get_undo_info(block_height)
+            print "undo", block_height, undo_info
+        else: undo_info = {}
+
+        # process
+        t1 = time.time()
+
+        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
+
+                for x in tx.get('outputs'):
+                    self.add_to_history( x.get('address'), txid, x.get('index'), block_height)
+                    
+            else:
+                for x in tx.get('outputs'):
+                    self.remove_from_history( x.get('address'), txid, x.get('index'))
+
+                i = 0
+                for x in tx.get('inputs'):
+                    prevout_height, prevout_addr = undo_info.get(txid)[i]
+                    i += 1
+
+                    # read the history into batch list
+                    self.batch_list[prevout_addr] = self.db.Get(prevout_addr)
+                    # re-add them to the history
+                    self.add_to_history( prevout_addr, x.get('prevout_hash'), x.get('prevout_n'), prevout_height)
+                    print "new hist for", prevout_addr, self.deserialize(self.batch_list[prevout_addr])
+
+        # write
+        max_len = 0
+        max_addr = ''
+        t2 = time.time()
 
-        # batch write
         batch = leveldb.WriteBatch()
-        for addr, hist in batch_list.items():
-            batch.Put(addr, self.serialize(hist))
-        batch.Put('0', self.serialize( [(block_hash, block_height)] ) )
-        self.db.Write(batch, sync = True)
+        for addr, serialized_hist in self.batch_list.items():
+            batch.Put(addr, serialized_hist)
+            l = len(serialized_hist)
+            if l > max_len:
+                max_len = l
+                max_addr = addr
+
+        if not revert:
+            # add new created outputs
+            for txio, addr in self.batch_txio.items():
+                batch.Put(txio, addr)
+            # delete spent inputs
+            for txi in block_inputs:
+                batch.Delete(txi)
+            # add undo info 
+            self.write_undo_info(batch, block_height, undo_info)
+        else:
+            # restore spent inputs
+            for txio, addr in self.batch_txio.items():
+                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)] ) )
+
+        # actual write
+        self.db.Write(batch, sync = sync)
 
-        # invalidate cache
-        for addr in batch_list.keys(): self.update_history_cache(addr)
+        t3 = time.time()
+        if t3 - t0 > 10 and not sync: 
+            print_log("block", block_height, 
+                      "parse:%0.2f "%(t00 - t0), 
+                      "read:%0.2f "%(t1 - t00), 
+                      "proc:%.2f "%(t2-t1), 
+                      "write:%.2f "%(t3-t2), 
+                      "max:", max_len, max_addr)
 
-        return len(txlist)
+        for addr in self.batch_list.keys(): self.invalidate_cache(addr)
 
 
 
@@ -294,13 +553,13 @@ class Blockchain2Processor(Processor):
         result = None
         error = None
 
-        if method == 'blockchain2.numblocks.subscribe':
+        if method == 'blockchain.numblocks.subscribe':
             result = self.height
 
-        elif method == 'blockchain2.headers.subscribe':
+        elif method == 'blockchain.headers.subscribe':
             result = self.header
 
-        elif method == 'blockchain2.address.subscribe':
+        elif method == 'blockchain.address.subscribe':
             try:
                 address = params[0]
                 result = self.get_status(address, cache_only)
@@ -309,16 +568,16 @@ class Blockchain2Processor(Processor):
                 error = str(e) + ': ' + address
                 print_log( "error:", error )
 
-        elif method == 'blockchain2.address.subscribe2':
+        elif method == 'blockchain.address.subscribe2':
             try:
                 address = params[0]
-                result = self.get_status2(address, cache_only)
+                result = self.get_status(address, cache_only)
                 self.watch_address(address)
             except BaseException, e:
                 error = str(e) + ': ' + address
                 print_log( "error:", error )
 
-        elif method == 'blockchain2.address.get_history':
+        elif method == 'blockchain.address.get_history2':
             try:
                 address = params[0]
                 result = self.get_history( address, cache_only )
@@ -326,7 +585,7 @@ class Blockchain2Processor(Processor):
                 error = str(e) + ': ' + address
                 print_log( "error:", error )
 
-        elif method == 'blockchain2.block.get_header':
+        elif method == 'blockchain.block.get_header':
             if cache_only: 
                 result = -1
             else:
@@ -337,7 +596,7 @@ class Blockchain2Processor(Processor):
                     error = str(e) + ': %d'% height
                     print_log( "error:", error )
                     
-        elif method == 'blockchain2.block.get_chunk':
+        elif method == 'blockchain.block.get_chunk':
             if cache_only:
                 result = -1
             else:
@@ -348,12 +607,12 @@ class Blockchain2Processor(Processor):
                     error = str(e) + ': %d'% index
                     print_log( "error:", error)
 
-        elif method == 'blockchain2.transaction.broadcast':
-            txo = self.bitcoind('sendrawtransaction', params[0])
+        elif method == 'blockchain.transaction.broadcast':
+            txo = self.bitcoind('sendrawtransaction', params)
             print_log( "sent tx:", txo )
             result = txo 
 
-        elif method == 'blockchain2.transaction.get_merkle':
+        elif method == 'blockchain.transaction.get_merkle':
             if cache_only:
                 result = -1
             else:
@@ -365,7 +624,7 @@ class Blockchain2Processor(Processor):
                     error = str(e) + ': ' + tx_hash
                     print_log( "error:", error )
                     
-        elif method == 'blockchain2.transaction.get':
+        elif method == 'blockchain.transaction.get':
             try:
                 tx_hash = params[0]
                 height = params[1]
@@ -393,11 +652,7 @@ class Blockchain2Processor(Processor):
 
 
 
-    def last_hash(self):
-        return self.block_hashes[-1]
-
-
-    def catch_up(self):
+    def catch_up(self, sync = True):
 
         t1 = time.time()
 
@@ -407,38 +662,46 @@ class Blockchain2Processor(Processor):
             info = self.bitcoind('getinfo')
             bitcoind_height = info.get('blocks')
             bitcoind_block_hash = self.bitcoind('getblockhash', [bitcoind_height])
-            if self.last_hash() == bitcoind_block_hash: break
+            if self.last_hash == bitcoind_block_hash: 
+                self.up_to_date = True
+                break
 
             # not done..
-            block_hash = self.bitcoind('getblockhash', [self.height+1])
-            block = self.bitcoind('getblock', [block_hash])
+            self.up_to_date = False
+            next_block_hash = self.bitcoind('getblockhash', [self.height+1])
+            next_block = self.bitcoind('getblock', [next_block_hash, 1])
 
-            if block.get('previousblockhash') == self.last_hash():
+            revert = (random.randint(1, 100)==1) if self.is_test else False
+            if (next_block.get('previousblockhash') == self.last_hash) and not revert:
 
-                self.import_block(block, block_hash, self.height+1)
+                self.import_block(next_block, next_block_hash, self.height+1, sync)
+                self.height = self.height + 1
+                self.write_header(self.block2header(next_block), sync)
+                self.last_hash = next_block_hash
 
-                if (self.height+1)%100 == 0: 
+                if (self.height)%100 == 0 and not sync: 
                     t2 = time.time()
-                    print_log( "bc2: block %d (%.3fs)"%( self.height+1, t2 - t1 ) )
+                    print_log( "catch_up: block %d (%.3fs)"%( self.height, t2 - t1 ) )
                     t1 = t2
-
-                self.height = self.height + 1
-                self.block_hashes.append(block_hash)
-                self.block_hashes = self.block_hashes[-10:]
                     
             else:
                 # revert current block
-                print_log( "bc2: reorg", self.height, block.get('previousblockhash'), self.last_hash() )
-                block_hash = self.last_hash()
-                block = self.bitcoind('getblock', [block_hash])
+                block = self.bitcoind('getblock', [self.last_hash, 1])
+                print_log( "blockchain reorg", self.height, block.get('previousblockhash'), self.last_hash )
+                self.import_block(block, self.last_hash, self.height, sync, revert=True)
+                self.pop_header()
+                self.flush_headers()
+
                 self.height = self.height -1
-                self.block_hashes.remove(block_hash)
-                self.revert_block(block, self.last_hash(), self.height)
+
+                # read previous header from disk
+                self.header = self.read_header(self.height)
+                self.last_hash = self.hash_header(self.header)
         
 
-        self.header = self.block2header(self.bitcoind('getblock', [self.last_hash()]))
+        self.header = self.block2header(self.bitcoind('getblock', [self.last_hash]))
+
 
-        
 
             
     def memorypool_update(self):
@@ -446,51 +709,95 @@ class Blockchain2Processor(Processor):
         mempool_hashes = self.bitcoind('getrawmempool')
 
         for tx_hash in mempool_hashes:
-            if tx_hash in self.known_mempool_hashes: continue
-            self.known_mempool_hashes.append(tx_hash)
+            if tx_hash in self.mempool_hashes: continue
 
             tx = self.get_transaction(tx_hash)
             if not tx: continue
 
-            for x in tx.get('inputs') + tx.get('outputs'):
+            for x in tx.get('inputs'):
+                txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex')
+                try:
+                    addr = self.db.Get(txi)    
+                except:
+                    continue
+                l = self.mempool_addresses.get(tx_hash, [])
+                if addr not in l: 
+                    l.append( addr )
+                    self.mempool_addresses[tx_hash] = l
+
+            for x in tx.get('outputs'):
                 addr = x.get('address')
-                hist = self.mempool_hist.get(addr, [])
-                if tx_hash not in hist: 
-                    hist.append( tx_hash )
-                    self.mempool_hist[addr] = hist
-                    self.update_history_cache(addr)
+                l = self.mempool_addresses.get(tx_hash, [])
+                if addr not in l: 
+                    l.append( addr )
+                    self.mempool_addresses[tx_hash] = l
+
+            self.mempool_hashes.append(tx_hash)
+
+        # remove older entries from mempool_hashes
+        self.mempool_hashes = mempool_hashes
+
+        # remove deprecated entries from mempool_addresses
+        for tx_hash, addresses in self.mempool_addresses.items():
+            if tx_hash not in self.mempool_hashes:
+                self.mempool_addresses.pop(tx_hash)
+
+        # rebuild histories
+        new_mempool_hist = {}
+        for tx_hash, addresses in self.mempool_addresses.items():
+            for addr in addresses:
+                h = new_mempool_hist.get(addr, [])
+                if tx_hash not in h: 
+                    h.append( tx_hash )
+                new_mempool_hist[addr] = h
+
+        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)
+
+        with self.mempool_lock:
+            self.mempool_hist = new_mempool_hist
 
-        self.known_mempool_hashes = mempool_hashes
 
 
-    def update_history_cache(self, address):
+    def invalidate_cache(self, address):
         with self.cache_lock:
             if self.history_cache.has_key(address):
                 print_log( "cache: invalidating", address )
                 self.history_cache.pop(address)
 
+        if address in self.watched_addresses:
+            self.address_queue.put(address)
+
 
 
     def main_iteration(self):
 
         if self.shared.stopped(): 
-            print_log( "bc2 terminating")
+            print_log( "blockchain processor terminating" )
             return
 
         with self.dblock:
             t1 = time.time()
             self.catch_up()
             t2 = time.time()
-            print_log( "blockchain: %d (%.3fs)"%( self.height+1, t2 - t1 ) )
+
         self.memorypool_update()
+        t3 = time.time()
+        # print "mempool:", len(self.mempool_addresses), len(self.mempool_hist), "%.3fs"%(t3 - t2)
+
 
         if self.sent_height != self.height:
             self.sent_height = self.height
-            self.push_response({ 'id': None, 'method':'blockchain2.numblocks.subscribe', 'params':[self.height] })
+            self.push_response({ 'id': None, 'method':'blockchain.numblocks.subscribe', 'params':[self.height] })
 
         if self.sent_header != self.header:
+            print_log( "blockchain: %d (%.3fs)"%( self.height, t2 - t1 ) )
             self.sent_header = self.header
-            self.push_response({ 'id': None, 'method':'blockchain2.headers.subscribe', 'params':[self.header] })
+            self.push_response({ 'id': None, 'method':'blockchain.headers.subscribe', 'params':[self.header] })
 
         while True:
             try:
@@ -500,12 +807,13 @@ class Blockchain2Processor(Processor):
             if addr in self.watched_addresses:
                 status = self.get_status( addr )
                 self.push_response({ 'id': None, 'method':'blockchain.address.subscribe', 'params':[addr, status] })
+                self.push_response({ 'id': None, 'method':'blockchain.address.subscribe2', 'params':[addr, status] })
 
 
         if not self.shared.stopped(): 
             threading.Timer(10, self.main_iteration).start()
         else:
-            print_log( "bc2 terminating" )
+            print_log( "blockchain processor terminating" )