claim activations and takeovers (WIP)

This commit is contained in:
Jack Robison 2021-05-05 16:17:32 -04:00
parent aa3b18f848
commit 9a11ac06bf
No known key found for this signature in database
GPG key ID: DF25C68FE0239BB2
6 changed files with 572 additions and 124 deletions

View file

@ -174,7 +174,6 @@ class Outputs:
@classmethod
def to_bytes(cls, txo_rows, extra_txo_rows, offset=0, total=None, blocked: Censor = None) -> bytes:
extra_txo_rows = {row['claim_hash']: row for row in extra_txo_rows}
page = OutputsMessage()
page.offset = offset
if total is not None:
@ -221,7 +220,7 @@ class Outputs:
if resolve_result.canonical_url is not None:
txo_message.claim.canonical_url = resolve_result.canonical_url
if resolve_result.last_take_over_height is not None:
txo_message.claim.take_over_height = resolve_result.last_take_over_height
if resolve_result.last_takeover_height is not None:
txo_message.claim.take_over_height = resolve_result.last_takeover_height
if resolve_result.claims_in_channel is not None:
txo_message.claim.claims_in_channel = resolve_result.claims_in_channel

View file

@ -18,6 +18,9 @@ from lbry.crypto.hash import hash160
from lbry.wallet.server.leveldb import FlushData
from lbry.wallet.server.db import DB_PREFIXES
from lbry.wallet.server.db.claimtrie import StagedClaimtrieItem, StagedClaimtrieSupport, get_expiration_height
from lbry.wallet.server.db.claimtrie import get_takeover_name_ops, get_force_activate_ops, get_delay_for_name
from lbry.wallet.server.db.prefixes import PendingClaimActivationPrefixRow, Prefixes
from lbry.wallet.server.db.revertable import RevertablePut
from lbry.wallet.server.udp import StatusServer
if typing.TYPE_CHECKING:
from lbry.wallet.server.leveldb import LevelDB
@ -202,13 +205,12 @@ class BlockProcessor:
self.history_cache = {}
self.status_server = StatusServer()
self.effective_amount_changes = defaultdict(list)
self.pending_claims = {}
self.pending_claim_txos = {}
self.pending_claims: typing.Dict[Tuple[int, int], StagedClaimtrieItem] = {}
self.pending_claim_txos: typing.Dict[bytes, Tuple[int, int]] = {}
self.pending_supports = defaultdict(set)
self.pending_support_txos = {}
self.pending_abandon = set()
self.staged_pending_abandoned = {}
async def run_in_thread_with_lock(self, func, *args):
# Run in a thread to prevent blocking. Shielded so that
@ -239,7 +241,6 @@ class BlockProcessor:
try:
for block in blocks:
await self.run_in_thread_with_lock(self.advance_block, block)
print("advanced\n")
except:
self.logger.exception("advance blocks failed")
raise
@ -412,7 +413,8 @@ class BlockProcessor:
return None
def _add_claim_or_update(self, height: int, txo, script, tx_hash: bytes, idx: int, tx_count: int, txout,
spent_claims: typing.Dict[bytes, typing.Tuple[int, int, str]]) -> List['RevertableOp']:
spent_claims: typing.Dict[bytes, typing.Tuple[int, int, str]],
zero_delay_claims: typing.Dict[Tuple[str, bytes], Tuple[int, int]]) -> List['RevertableOp']:
try:
claim_name = txo.normalized_name
except UnicodeDecodeError:
@ -425,7 +427,13 @@ class BlockProcessor:
signing_channel_hash = None
channel_claims_count = 0
activation_height = 0
activation_delay = self.db.get_activation_delay(claim_hash, claim_name)
if activation_delay == 0:
zero_delay_claims[(claim_name, claim_hash)] = tx_count, idx
# else:
# print("delay activation ", claim_name, activation_delay, height)
activation_height = activation_delay + height
try:
signable = txo.signable
except: # google.protobuf.message.DecodeError: Could not parse JSON.
@ -455,9 +463,13 @@ class BlockProcessor:
root_idx = previous_claim.root_claim_tx_position
# prev_amount = previous_claim.amount
else:
root_tx_num, root_idx, prev_amount, _, _, _ = self.db.get_root_claim_txo_and_current_amount(
k, v = self.db.get_root_claim_txo_and_current_amount(
claim_hash
)
root_tx_num = v.root_tx_num
root_idx = v.root_position
prev_amount = v.amount
pending = StagedClaimtrieItem(
claim_name, claim_hash, txout.value, support_amount + txout.value,
activation_height, get_expiration_height(height), tx_count, idx, root_tx_num, root_idx,
@ -469,11 +481,25 @@ class BlockProcessor:
self.effective_amount_changes[claim_hash].append(txout.value)
return pending.get_add_claim_utxo_ops()
def _add_support(self, txo, txout, idx, tx_count) -> List['RevertableOp']:
def _add_support(self, height, txo, txout, idx, tx_count,
zero_delay_claims: typing.Dict[Tuple[str, bytes], Tuple[int, int]]) -> List['RevertableOp']:
supported_claim_hash = txo.claim_hash[::-1]
claim_info = self.db.get_root_claim_txo_and_current_amount(
supported_claim_hash
)
controlling_claim = None
supported_tx_num = supported_position = supported_activation_height = supported_name = None
if claim_info:
k, v = claim_info
supported_name = v.name
supported_tx_num = k.tx_num
supported_position = k.position
supported_activation_height = v.activation
controlling_claim = self.db.get_controlling_claim(v.name)
if supported_claim_hash in self.effective_amount_changes:
# print(f"\tsupport claim {supported_claim_hash.hex()} {starting_amount}+{txout.value}={starting_amount + txout.value}")
# print(f"\tsupport claim {supported_claim_hash.hex()} {txout.value}")
self.effective_amount_changes[supported_claim_hash].append(txout.value)
self.pending_supports[supported_claim_hash].add((tx_count, idx))
self.pending_support_txos[(tx_count, idx)] = supported_claim_hash, txout.value
@ -481,42 +507,84 @@ class BlockProcessor:
supported_claim_hash, tx_count, idx, txout.value
).get_add_support_utxo_ops()
elif supported_claim_hash not in self.pending_claims and supported_claim_hash not in self.pending_abandon:
if self.db.claim_exists(supported_claim_hash):
_, _, _, name, supported_tx_num, supported_pos = self.db.get_root_claim_txo_and_current_amount(
supported_claim_hash
)
# print(f"\tsupport claim {supported_claim_hash.hex()} {txout.value}")
ops = []
if claim_info:
starting_amount = self.db.get_effective_amount(supported_claim_hash)
if supported_claim_hash not in self.effective_amount_changes:
self.effective_amount_changes[supported_claim_hash].append(starting_amount)
self.effective_amount_changes[supported_claim_hash].append(txout.value)
supported_amount = self._get_pending_effective_amount(supported_claim_hash)
if controlling_claim and supported_claim_hash != controlling_claim.claim_hash:
if supported_amount + txo.amount > self._get_pending_effective_amount(controlling_claim.claim_hash):
# takeover could happen
if (supported_name, supported_claim_hash) not in zero_delay_claims:
takeover_delay = get_delay_for_name(height - supported_activation_height)
if takeover_delay == 0:
zero_delay_claims[(supported_name, supported_claim_hash)] = (
supported_tx_num, supported_position
)
else:
ops.append(
RevertablePut(
*Prefixes.pending_activation.pack_item(
height + takeover_delay, supported_tx_num, supported_position,
supported_claim_hash, supported_name
)
)
)
self.pending_supports[supported_claim_hash].add((tx_count, idx))
self.pending_support_txos[(tx_count, idx)] = supported_claim_hash, txout.value
# print(f"\tsupport claim {supported_claim_hash.hex()} {starting_amount}+{txout.value}={starting_amount + txout.value}")
return StagedClaimtrieSupport(
ops.extend(StagedClaimtrieSupport(
supported_claim_hash, tx_count, idx, txout.value
).get_add_support_utxo_ops()
).get_add_support_utxo_ops())
return ops
else:
print(f"\tthis is a wonky tx, contains unlinked support for non existent {supported_claim_hash.hex()}")
return []
def _add_claim_or_support(self, height: int, tx_hash: bytes, tx_count: int, idx: int, txo, txout, script,
spent_claims: typing.Dict[bytes, Tuple[int, int, str]]) -> List['RevertableOp']:
spent_claims: typing.Dict[bytes, Tuple[int, int, str]],
zero_delay_claims: typing.Dict[Tuple[str, bytes], Tuple[int, int]]) -> List['RevertableOp']:
if script.is_claim_name or script.is_update_claim:
return self._add_claim_or_update(height, txo, script, tx_hash, idx, tx_count, txout, spent_claims)
return self._add_claim_or_update(height, txo, script, tx_hash, idx, tx_count, txout, spent_claims,
zero_delay_claims)
elif script.is_support_claim or script.is_support_claim_data:
return self._add_support(txo, txout, idx, tx_count)
return self._add_support(height, txo, txout, idx, tx_count, zero_delay_claims)
return []
def _spend_support(self, txin):
def _remove_support(self, txin, zero_delay_claims):
txin_num = self.db.transaction_num_mapping[txin.prev_hash]
supported_name = None
if (txin_num, txin.prev_idx) in self.pending_support_txos:
spent_support, support_amount = self.pending_support_txos.pop((txin_num, txin.prev_idx))
supported_name = self._get_pending_claim_name(spent_support)
self.pending_supports[spent_support].remove((txin_num, txin.prev_idx))
else:
spent_support, support_amount = self.db.get_supported_claim_from_txo(txin_num, txin.prev_idx)
if spent_support:
supported_name = self._get_pending_claim_name(spent_support)
if spent_support and support_amount is not None and spent_support not in self.pending_abandon:
# print(f"\tspent support for {spent_support.hex()} -{support_amount} ({txin_num}, {txin.prev_idx})")
controlling = self.db.get_controlling_claim(supported_name)
if controlling:
bid_queue = {
claim_hash: self._get_pending_effective_amount(claim_hash)
for claim_hash in self.db.get_claims_for_name(supported_name)
if claim_hash not in self.pending_abandon
}
bid_queue[spent_support] -= support_amount
sorted_claims = sorted(
list(bid_queue.keys()), key=lambda claim_hash: bid_queue[claim_hash], reverse=True
)
if controlling.claim_hash == spent_support and sorted_claims.index(controlling.claim_hash) > 0:
print("takeover due to abandoned support")
# print(f"\tspent support for {spent_support.hex()} -{support_amount} ({txin_num}, {txin.prev_idx}) {supported_name}")
if spent_support not in self.effective_amount_changes:
assert spent_support not in self.pending_claims
prev_effective_amount = self.db.get_effective_amount(spent_support)
@ -527,7 +595,7 @@ class BlockProcessor:
).get_spend_support_txo_ops()
return []
def _spend_claim(self, txin, spent_claims):
def _remove_claim(self, txin, spent_claims, zero_delay_claims):
txin_num = self.db.transaction_num_mapping[txin.prev_hash]
if (txin_num, txin.prev_idx) in self.pending_claims:
spent = self.pending_claims[(txin_num, txin.prev_idx)]
@ -540,7 +608,7 @@ class BlockProcessor:
)
if not spent_claim_hash_and_name: # txo is not a claim
return []
prev_claim_hash, txi_len_encoded_name = spent_claim_hash_and_name
prev_claim_hash = spent_claim_hash_and_name.claim_hash
prev_signing_hash = self.db.get_channel_for_claim(prev_claim_hash)
prev_claims_in_channel_count = None
@ -551,8 +619,14 @@ class BlockProcessor:
prev_effective_amount = self.db.get_effective_amount(
prev_claim_hash
)
claim_root_tx_num, claim_root_idx, prev_amount, name, tx_num, position = self.db.get_root_claim_txo_and_current_amount(prev_claim_hash)
activation_height = 0
k, v = self.db.get_root_claim_txo_and_current_amount(prev_claim_hash)
claim_root_tx_num = v.root_tx_num
claim_root_idx = v.root_position
prev_amount = v.amount
name = v.name
tx_num = k.tx_num
position = k.position
activation_height = v.activation
height = bisect_right(self.db.tx_counts, tx_num)
spent = StagedClaimtrieItem(
name, prev_claim_hash, prev_amount, prev_effective_amount,
@ -564,23 +638,29 @@ class BlockProcessor:
if spent.claim_hash not in self.effective_amount_changes:
self.effective_amount_changes[spent.claim_hash].append(spent.effective_amount)
self.effective_amount_changes[spent.claim_hash].append(-spent.amount)
if (name, spent.claim_hash) in zero_delay_claims:
zero_delay_claims.pop((name, spent.claim_hash))
return spent.get_spend_claim_txo_ops()
def _spend_claim_or_support(self, txin, spent_claims):
spend_claim_ops = self._spend_claim(txin, spent_claims)
def _remove_claim_or_support(self, txin, spent_claims, zero_delay_claims):
spend_claim_ops = self._remove_claim(txin, spent_claims, zero_delay_claims)
if spend_claim_ops:
return spend_claim_ops
return self._spend_support(txin)
return self._remove_support(txin, zero_delay_claims)
def _abandon(self, spent_claims):
def _abandon(self, spent_claims) -> typing.Tuple[List['RevertableOp'], typing.Set[str]]:
# Handle abandoned claims
ops = []
controlling_claims = {}
need_takeover = set()
for abandoned_claim_hash, (prev_tx_num, prev_idx, name) in spent_claims.items():
# print(f"\tabandon lbry://{name}#{abandoned_claim_hash.hex()} {prev_tx_num} {prev_idx}")
if (prev_tx_num, prev_idx) in self.pending_claims:
pending = self.pending_claims.pop((prev_tx_num, prev_idx))
self.staged_pending_abandoned[pending.claim_hash] = pending
claim_root_tx_num = pending.root_claim_tx_num
claim_root_idx = pending.root_claim_tx_position
prev_amount = pending.amount
@ -588,9 +668,12 @@ class BlockProcessor:
prev_effective_amount = pending.effective_amount
prev_claims_in_channel_count = pending.claims_in_channel_count
else:
claim_root_tx_num, claim_root_idx, prev_amount, _, _, _ = self.db.get_root_claim_txo_and_current_amount(
k, v = self.db.get_root_claim_txo_and_current_amount(
abandoned_claim_hash
)
claim_root_tx_num = v.root_tx_num
claim_root_idx = v.root_position
prev_amount = v.amount
prev_signing_hash = self.db.get_channel_for_claim(abandoned_claim_hash)
prev_claims_in_channel_count = None
if prev_signing_hash:
@ -601,6 +684,13 @@ class BlockProcessor:
abandoned_claim_hash
)
if name not in controlling_claims:
controlling_claims[name] = self.db.get_controlling_claim(name)
controlling = controlling_claims[name]
if controlling and controlling.claim_hash == abandoned_claim_hash:
need_takeover.add(name)
# print("needs takeover")
for (support_tx_num, support_tx_idx) in self.pending_supports[abandoned_claim_hash]:
_, support_amount = self.pending_support_txos.pop((support_tx_num, support_tx_idx))
ops.extend(
@ -628,7 +718,7 @@ class BlockProcessor:
self.effective_amount_changes.pop(abandoned_claim_hash)
self.pending_abandon.add(abandoned_claim_hash)
# print(f"\tabandoned lbry://{name}#{abandoned_claim_hash.hex()}")
# print(f"\tabandoned lbry://{name}#{abandoned_claim_hash.hex()}, {len(need_takeover)} names need takeovers")
ops.extend(
StagedClaimtrieItem(
name, abandoned_claim_hash, prev_amount, prev_effective_amount,
@ -636,20 +726,120 @@ class BlockProcessor:
claim_root_idx, prev_signing_hash, prev_claims_in_channel_count
).get_abandon_ops(self.db.db)
)
return ops
return ops, need_takeover
def _expire_claims(self, height: int):
def _expire_claims(self, height: int, zero_delay_claims):
expired = self.db.get_expired_by_height(height)
spent_claims = {}
ops = []
names_needing_takeover = set()
for expired_claim_hash, (tx_num, position, name, txi) in expired.items():
if (tx_num, position) not in self.pending_claims:
ops.extend(self._spend_claim(txi, spent_claims))
ops.extend(self._remove_claim(txi, spent_claims, zero_delay_claims))
if expired:
# do this to follow the same content claim removing pathway as if a claim (possible channel) was abandoned
abandon_ops, _names_needing_takeover = self._abandon(spent_claims)
if abandon_ops:
ops.extend(abandon_ops)
names_needing_takeover.update(_names_needing_takeover)
ops.extend(self._abandon(spent_claims))
return ops, names_needing_takeover
def _get_pending_claim_amount(self, claim_hash: bytes) -> int:
if claim_hash in self.pending_claim_txos:
return self.pending_claims[self.pending_claim_txos[claim_hash]].amount
return self.db.get_claim_amount(claim_hash)
def _get_pending_claim_name(self, claim_hash: bytes) -> str:
assert claim_hash is not None
if claim_hash in self.pending_claims:
return self.pending_claims[claim_hash].name
claim = self.db.get_claim_from_txo(claim_hash)
return claim.name
def _get_pending_effective_amount(self, claim_hash: bytes) -> int:
claim_amount = self._get_pending_claim_amount(claim_hash) or 0
support_amount = self.db.get_support_amount(claim_hash) or 0
return claim_amount + support_amount + sum(
self.pending_support_txos[support_txnum, support_n][1]
for (support_txnum, support_n) in self.pending_supports.get(claim_hash, [])
) # TODO: subtract pending spend supports
def _get_name_takeover_ops(self, height: int, name: str,
activated_claims: typing.Set[bytes]) -> List['RevertableOp']:
controlling = self.db.get_controlling_claim(name)
if not controlling or controlling.claim_hash in self.pending_abandon:
# print("no controlling claim for ", name)
bid_queue = {
claim_hash: self._get_pending_effective_amount(claim_hash) for claim_hash in activated_claims
}
winning_claim = max(bid_queue, key=lambda k: bid_queue[k])
if winning_claim in self.pending_claim_txos:
s = self.pending_claims[self.pending_claim_txos[winning_claim]]
else:
s = self.db.make_staged_claim_item(winning_claim)
ops = []
if s.activation_height > height:
ops.extend(get_force_activate_ops(
name, s.tx_num, s.position, s.claim_hash, s.root_claim_tx_num, s.root_claim_tx_position,
s.amount, s.effective_amount, s.activation_height, height
))
ops.extend(get_takeover_name_ops(name, winning_claim, height))
return ops
else:
# print(f"current controlling claim for {name}#{controlling.claim_hash.hex()}")
controlling_effective_amount = self._get_pending_effective_amount(controlling.claim_hash)
bid_queue = {
claim_hash: self._get_pending_effective_amount(claim_hash) for claim_hash in activated_claims
}
highest_newly_activated = max(bid_queue, key=lambda k: bid_queue[k])
if bid_queue[highest_newly_activated] > controlling_effective_amount:
# print(f"takeover controlling claim for {name}#{controlling.claim_hash.hex()}")
return get_takeover_name_ops(name, highest_newly_activated, height, controlling)
print(bid_queue[highest_newly_activated], controlling_effective_amount)
# print("no takeover")
return []
def _get_takeover_ops(self, height: int, zero_delay_claims) -> List['RevertableOp']:
ops = []
pending = defaultdict(set)
# get non delayed takeovers for new names
for (name, claim_hash) in zero_delay_claims:
if claim_hash not in self.pending_abandon:
pending[name].add(claim_hash)
# print("zero delay activate", name, claim_hash.hex())
# get takeovers from claims activated at this block
for activated in self.db.get_activated_claims_at_height(height):
if activated.claim_hash not in self.pending_abandon:
pending[activated.name].add(activated.claim_hash)
# print("delayed activate")
# get takeovers from supports for controlling claims being abandoned
for abandoned_claim_hash in self.pending_abandon:
if abandoned_claim_hash in self.staged_pending_abandoned:
abandoned = self.staged_pending_abandoned[abandoned_claim_hash]
controlling = self.db.get_controlling_claim(abandoned.name)
if controlling and controlling.claim_hash == abandoned_claim_hash and abandoned.name not in pending:
pending[abandoned.name].update(self.db.get_claims_for_name(abandoned.name))
else:
k, v = self.db.get_root_claim_txo_and_current_amount(abandoned_claim_hash)
controlling_claim = self.db.get_controlling_claim(v.name)
if controlling_claim and abandoned_claim_hash == controlling_claim.claim_hash and v.name not in pending:
pending[v.name].update(self.db.get_claims_for_name(v.name))
# print("check abandoned winning")
# get takeovers from controlling claims being abandoned
for name, activated_claims in pending.items():
ops.extend(self._get_name_takeover_ops(height, name, activated_claims))
return ops
def advance_block(self, block):
# print("advance ", height)
height = self.height + 1
txs: List[Tuple[Tx, bytes]] = block.transactions
block_hash = self.coin.header_hash(block.header)
@ -672,7 +862,8 @@ class BlockProcessor:
append_hashX_by_tx = hashXs_by_tx.append
hashX_from_script = self.coin.hashX_from_script
# unchanged_effective_amounts = {k: sum(v) for k, v in self.effective_amount_changes.items()}
zero_delay_claims: typing.Dict[Tuple[str, bytes], Tuple[int, int]] = {}
abandoned_or_expired_controlling = set()
for tx, tx_hash in txs:
spent_claims = {}
@ -690,7 +881,7 @@ class BlockProcessor:
undo_info_append(cache_value)
append_hashX(cache_value[:-12])
spend_claim_or_support_ops = self._spend_claim_or_support(txin, spent_claims)
spend_claim_or_support_ops = self._remove_claim_or_support(txin, spent_claims, zero_delay_claims)
if spend_claim_or_support_ops:
claimtrie_stash_extend(spend_claim_or_support_ops)
@ -708,15 +899,16 @@ class BlockProcessor:
txo = Output(txout.value, script)
claim_or_support_ops = self._add_claim_or_support(
height, tx_hash, tx_count, idx, txo, txout, script, spent_claims
height, tx_hash, tx_count, idx, txo, txout, script, spent_claims, zero_delay_claims
)
if claim_or_support_ops:
claimtrie_stash_extend(claim_or_support_ops)
# Handle abandoned claims
abandon_ops = self._abandon(spent_claims)
abandon_ops, abandoned_controlling_need_takeover = self._abandon(spent_claims)
if abandon_ops:
claimtrie_stash_extend(abandon_ops)
abandoned_or_expired_controlling.update(abandoned_controlling_need_takeover)
append_hashX_by_tx(hashXs)
update_touched(hashXs)
@ -725,11 +917,17 @@ class BlockProcessor:
tx_count += 1
# handle expired claims
expired_ops = self._expire_claims(height)
expired_ops, expired_need_takeover = self._expire_claims(height, zero_delay_claims)
if expired_ops:
print(f"************\nexpire claims at block {height}\n************")
# print(f"************\nexpire claims at block {height}\n************")
abandoned_or_expired_controlling.update(expired_need_takeover)
claimtrie_stash_extend(expired_ops)
# activate claims and process takeovers
takeover_ops = self._get_takeover_ops(height, zero_delay_claims)
if takeover_ops:
claimtrie_stash_extend(takeover_ops)
# self.db.add_unflushed(hashXs_by_tx, self.tx_count)
_unflushed = self.db.hist_unflushed
_count = 0
@ -789,7 +987,6 @@ class BlockProcessor:
coin = self.coin
for raw_block in raw_blocks:
self.logger.info("backup block %i", self.height)
print("backup", self.height)
# Check and update self.tip
block = coin.block(raw_block, self.height)
header_hash = coin.header_hash(block.header)

View file

@ -15,6 +15,9 @@ class DB_PREFIXES(enum.Enum):
claim_effective_amount_prefix = b'D'
claim_expiration = b'O'
claim_takeover = b'P'
pending_activation = b'Q'
undo_claimtrie = b'M'
HISTORY_PREFIX = b'A'

View file

@ -2,7 +2,7 @@ import typing
from typing import Optional
from lbry.wallet.server.db.revertable import RevertablePut, RevertableDelete, RevertableOp, delete_prefix
from lbry.wallet.server.db import DB_PREFIXES
from lbry.wallet.server.db.prefixes import Prefixes
from lbry.wallet.server.db.prefixes import Prefixes, ClaimTakeoverValue
nOriginalClaimExpirationTime = 262974
nExtendedClaimExpirationTime = 2102400
@ -13,6 +13,12 @@ nMaxTakeoverWorkaroundHeight = 658300 # targeting 30 Oct 2019
nWitnessForkHeight = 680770 # targeting 11 Dec 2019
nAllClaimsInMerkleForkHeight = 658310 # targeting 30 Oct 2019
proportionalDelayFactor = 32
maxTakeoverDelay = 4032
def get_delay_for_name(blocks_of_continuous_ownership: int) -> int:
return min(blocks_of_continuous_ownership // proportionalDelayFactor, maxTakeoverDelay)
def get_expiration_height(last_updated_height: int) -> int:
if last_updated_height < nExtendedClaimExpirationForkHeight:
@ -57,18 +63,21 @@ class StagedClaimtrieSupport(typing.NamedTuple):
def get_update_effective_amount_ops(name: str, new_effective_amount: int, prev_effective_amount: int, tx_num: int,
position: int, root_tx_num: int, root_position: int, claim_hash: bytes,
activation_height: int, prev_activation_height: int,
signing_hash: Optional[bytes] = None,
claims_in_channel_count: Optional[int] = None):
assert root_position != root_tx_num, f"{tx_num} {position} {root_tx_num} {root_tx_num}"
ops = [
RevertableDelete(
*Prefixes.claim_effective_amount.pack_item(
name, prev_effective_amount, tx_num, position, claim_hash, root_tx_num, root_position
name, prev_effective_amount, tx_num, position, claim_hash, root_tx_num, root_position,
prev_activation_height
)
),
RevertablePut(
*Prefixes.claim_effective_amount.pack_item(
name, new_effective_amount, tx_num, position, claim_hash, root_tx_num, root_position
name, new_effective_amount, tx_num, position, claim_hash, root_tx_num, root_position,
activation_height
)
)
]
@ -88,6 +97,89 @@ def get_update_effective_amount_ops(name: str, new_effective_amount: int, prev_e
return ops
def get_takeover_name_ops(name: str, claim_hash: bytes, takeover_height: int,
previous_winning: Optional[ClaimTakeoverValue] = None):
if previous_winning:
# print(f"takeover previously owned {name} - {claim_hash.hex()} at {takeover_height}")
return [
RevertableDelete(
*Prefixes.claim_takeover.pack_item(
name, previous_winning.claim_hash, previous_winning.height
)
),
RevertablePut(
*Prefixes.claim_takeover.pack_item(
name, claim_hash, takeover_height
)
)
]
# print(f"takeover {name} - {claim_hash[::-1].hex()} at {takeover_height}")
return [
RevertablePut(
*Prefixes.claim_takeover.pack_item(
name, claim_hash, takeover_height
)
)
]
def get_force_activate_ops(name: str, tx_num: int, position: int, claim_hash: bytes, root_claim_tx_num: int,
root_claim_tx_position: int, amount: int, effective_amount: int,
prev_activation_height: int, new_activation_height: int):
return [
# delete previous
RevertableDelete(
*Prefixes.claim_effective_amount.pack_item(
name, effective_amount, tx_num, position, claim_hash,
root_claim_tx_num, root_claim_tx_position, prev_activation_height
)
),
RevertableDelete(
*Prefixes.claim_to_txo.pack_item(
claim_hash, tx_num, position, root_claim_tx_num, root_claim_tx_position,
amount, prev_activation_height, name
)
),
RevertableDelete(
*Prefixes.claim_short_id.pack_item(
name, claim_hash, root_claim_tx_num, root_claim_tx_position, tx_num,
position, prev_activation_height
)
),
RevertableDelete(
*Prefixes.pending_activation.pack_item(
prev_activation_height, tx_num, position, claim_hash, name
)
),
# insert new
RevertablePut(
*Prefixes.claim_effective_amount.pack_item(
name, effective_amount, tx_num, position, claim_hash,
root_claim_tx_num, root_claim_tx_position, new_activation_height
)
),
RevertablePut(
*Prefixes.claim_to_txo.pack_item(
claim_hash, tx_num, position, root_claim_tx_num, root_claim_tx_position,
amount, new_activation_height, name
)
),
RevertablePut(
*Prefixes.claim_short_id.pack_item(
name, claim_hash, root_claim_tx_num, root_claim_tx_position, tx_num,
position, new_activation_height
)
),
RevertablePut(
*Prefixes.pending_activation.pack_item(
new_activation_height, tx_num, position, claim_hash, name
)
)
]
class StagedClaimtrieItem(typing.NamedTuple):
name: str
claim_hash: bytes
@ -119,21 +211,21 @@ class StagedClaimtrieItem(typing.NamedTuple):
op(
*Prefixes.claim_effective_amount.pack_item(
self.name, self.effective_amount, self.tx_num, self.position, self.claim_hash,
self.root_claim_tx_num, self.root_claim_tx_position
self.root_claim_tx_num, self.root_claim_tx_position, self.activation_height
)
),
# claim tip by claim hash
op(
*Prefixes.claim_to_txo.pack_item(
self.claim_hash, self.tx_num, self.position, self.root_claim_tx_num, self.root_claim_tx_position,
self.amount, self.name
self.amount, self.activation_height, self.name
)
),
# short url resolution
op(
*Prefixes.claim_short_id.pack_item(
self.name, self.claim_hash, self.root_claim_tx_num, self.root_claim_tx_position, self.tx_num,
self.position
self.position, self.activation_height
)
),
# claim hash by txo
@ -146,6 +238,12 @@ class StagedClaimtrieItem(typing.NamedTuple):
self.expiration_height, self.tx_num, self.position, self.claim_hash,
self.name
)
),
# claim activation
op(
*Prefixes.pending_activation.pack_item(
self.activation_height, self.tx_num, self.position, self.claim_hash, self.name
)
)
]
if self.signing_hash and self.claims_in_channel_count is not None:
@ -187,4 +285,3 @@ class StagedClaimtrieItem(typing.NamedTuple):
delete_supports_ops = delete_prefix(db, DB_PREFIXES.claim_to_support.value + self.claim_hash)
invalidate_channel_ops = self.get_invalidate_channel_ops(db)
return delete_short_id_ops + delete_claim_ops + delete_supports_ops + invalidate_channel_ops

View file

@ -46,6 +46,7 @@ class EffectiveAmountValue(typing.NamedTuple):
claim_hash: bytes
root_tx_num: int
root_position: int
activation: int
class ClaimToTXOKey(typing.NamedTuple):
@ -58,6 +59,7 @@ class ClaimToTXOValue(typing.NamedTuple):
root_tx_num: int
root_position: int
amount: int
activation: int
name: str
@ -81,6 +83,7 @@ class ClaimShortIDKey(typing.NamedTuple):
class ClaimShortIDValue(typing.NamedTuple):
tx_num: int
position: int
activation: int
class ClaimToChannelKey(typing.NamedTuple):
@ -134,10 +137,30 @@ class ClaimExpirationValue(typing.NamedTuple):
name: str
class ClaimTakeoverKey(typing.NamedTuple):
name: str
class ClaimTakeoverValue(typing.NamedTuple):
claim_hash: bytes
height: int
class PendingActivationKey(typing.NamedTuple):
height: int
tx_num: int
position: int
class PendingActivationValue(typing.NamedTuple):
claim_hash: bytes
name: str
class EffectiveAmountPrefixRow(PrefixRow):
prefix = DB_PREFIXES.claim_effective_amount_prefix.value
key_struct = struct.Struct(b'>QLH')
value_struct = struct.Struct(b'>20sLH')
value_struct = struct.Struct(b'>20sLHL')
@classmethod
def pack_key(cls, name: str, effective_amount: int, tx_num: int, position: int):
@ -160,20 +183,20 @@ class EffectiveAmountPrefixRow(PrefixRow):
return EffectiveAmountValue(*super().unpack_value(data))
@classmethod
def pack_value(cls, claim_hash: bytes, root_tx_num: int, root_position: int) -> bytes:
return super().pack_value(claim_hash, root_tx_num, root_position)
def pack_value(cls, claim_hash: bytes, root_tx_num: int, root_position: int, activation: int) -> bytes:
return super().pack_value(claim_hash, root_tx_num, root_position, activation)
@classmethod
def pack_item(cls, name: str, effective_amount: int, tx_num: int, position: int, claim_hash: bytes,
root_tx_num: int, root_position: int):
root_tx_num: int, root_position: int, activation: int):
return cls.pack_key(name, effective_amount, tx_num, position), \
cls.pack_value(claim_hash, root_tx_num, root_position)
cls.pack_value(claim_hash, root_tx_num, root_position, activation)
class ClaimToTXOPrefixRow(PrefixRow):
prefix = DB_PREFIXES.claim_to_txo.value
key_struct = struct.Struct(b'>20sLH')
value_struct = struct.Struct(b'>LHQ')
value_struct = struct.Struct(b'>LHQL')
@classmethod
def pack_key(cls, claim_hash: bytes, tx_num: int, position: int):
@ -190,21 +213,21 @@ class ClaimToTXOPrefixRow(PrefixRow):
)
@classmethod
def unpack_value(cls, data: bytes) ->ClaimToTXOValue:
root_tx_num, root_position, amount = cls.value_struct.unpack(data[:14])
name_len = int.from_bytes(data[14:16], byteorder='big')
name = data[16:16 + name_len].decode()
return ClaimToTXOValue(root_tx_num, root_position, amount, name)
def unpack_value(cls, data: bytes) -> ClaimToTXOValue:
root_tx_num, root_position, amount, activation = cls.value_struct.unpack(data[:18])
name_len = int.from_bytes(data[18:20], byteorder='big')
name = data[20:20 + name_len].decode()
return ClaimToTXOValue(root_tx_num, root_position, amount, activation, name)
@classmethod
def pack_value(cls, root_tx_num: int, root_position: int, amount: int, name: str) -> bytes:
return cls.value_struct.pack(root_tx_num, root_position, amount) + length_encoded_name(name)
def pack_value(cls, root_tx_num: int, root_position: int, amount: int, activation: int, name: str) -> bytes:
return cls.value_struct.pack(root_tx_num, root_position, amount, activation) + length_encoded_name(name)
@classmethod
def pack_item(cls, claim_hash: bytes, tx_num: int, position: int, root_tx_num: int, root_position: int,
amount: int, name: str):
amount: int, activation: int, name: str):
return cls.pack_key(claim_hash, tx_num, position), \
cls.pack_value(root_tx_num, root_position, amount, name)
cls.pack_value(root_tx_num, root_position, amount, activation, name)
class TXOToClaimPrefixRow(PrefixRow):
@ -240,15 +263,15 @@ class TXOToClaimPrefixRow(PrefixRow):
class ClaimShortIDPrefixRow(PrefixRow):
prefix = DB_PREFIXES.claim_short_id_prefix.value
key_struct = struct.Struct(b'>20sLH')
value_struct = struct.Struct(b'>LH')
value_struct = struct.Struct(b'>LHL')
@classmethod
def pack_key(cls, name: str, claim_hash: bytes, root_tx_num: int, root_position: int):
return cls.prefix + length_encoded_name(name) + cls.key_struct.pack(claim_hash, root_tx_num, root_position)
@classmethod
def pack_value(cls, tx_num: int, position: int):
return super().pack_value(tx_num, position)
def pack_value(cls, tx_num: int, position: int, activation: int):
return super().pack_value(tx_num, position, activation)
@classmethod
def unpack_key(cls, key: bytes) -> ClaimShortIDKey:
@ -263,9 +286,9 @@ class ClaimShortIDPrefixRow(PrefixRow):
@classmethod
def pack_item(cls, name: str, claim_hash: bytes, root_tx_num: int, root_position: int,
tx_num: int, position: int):
tx_num: int, position: int, activation: int):
return cls.pack_key(name, claim_hash, root_tx_num, root_position), \
cls.pack_value(tx_num, position)
cls.pack_value(tx_num, position, activation)
class ClaimToChannelPrefixRow(PrefixRow):
@ -418,6 +441,63 @@ class ClaimExpirationPrefixRow(PrefixRow):
return cls.unpack_key(key), cls.unpack_value(value)
class ClaimTakeoverPrefixRow(PrefixRow):
prefix = DB_PREFIXES.claim_takeover.value
value_struct = struct.Struct(b'>20sL')
@classmethod
def pack_key(cls, name: str):
return cls.prefix + length_encoded_name(name)
@classmethod
def pack_value(cls, claim_hash: bytes, takeover_height: int):
return super().pack_value(claim_hash, takeover_height)
@classmethod
def unpack_key(cls, key: bytes) -> ClaimTakeoverKey:
assert key[:1] == cls.prefix
name_len = int.from_bytes(key[1:3], byteorder='big')
name = key[3:3 + name_len].decode()
return ClaimTakeoverKey(name)
@classmethod
def unpack_value(cls, data: bytes) -> ClaimTakeoverValue:
return ClaimTakeoverValue(*super().unpack_value(data))
@classmethod
def pack_item(cls, name: str, claim_hash: bytes, takeover_height: int):
return cls.pack_key(name), cls.pack_value(claim_hash, takeover_height)
class PendingClaimActivationPrefixRow(PrefixRow):
prefix = DB_PREFIXES.pending_activation.value
key_struct = struct.Struct(b'>LLH')
@classmethod
def pack_key(cls, height: int, tx_num: int, position: int):
return super().pack_key(height, tx_num, position)
@classmethod
def unpack_key(cls, key: bytes) -> PendingActivationKey:
return PendingActivationKey(*super().unpack_key(key))
@classmethod
def pack_value(cls, claim_hash: bytes, name: str) -> bytes:
return claim_hash + length_encoded_name(name)
@classmethod
def unpack_value(cls, data: bytes) -> PendingActivationValue:
claim_hash = data[:20]
name_len = int.from_bytes(data[20:22], byteorder='big')
name = data[22:22 + name_len].decode()
return PendingActivationValue(claim_hash, name)
@classmethod
def pack_item(cls, height: int, tx_num: int, position: int, claim_hash: bytes, name: str):
return cls.pack_key(height, tx_num, position), \
cls.pack_value(claim_hash, name)
class Prefixes:
claim_to_support = ClaimToSupportPrefixRow
support_to_claim = SupportToClaimPrefixRow
@ -432,4 +512,7 @@ class Prefixes:
claim_effective_amount = EffectiveAmountPrefixRow
claim_expiration = ClaimExpirationPrefixRow
claim_takeover = ClaimTakeoverPrefixRow
pending_activation = PendingClaimActivationPrefixRow
# undo_claimtrie = b'M'

View file

@ -16,6 +16,8 @@ import time
import typing
import struct
import attr
import zlib
import base64
from typing import Optional, Iterable
from functools import partial
from asyncio import sleep
@ -34,9 +36,10 @@ from lbry.wallet.server.util import formatted_time, pack_be_uint16, unpack_be_ui
from lbry.wallet.server.storage import db_class
from lbry.wallet.server.db.revertable import RevertablePut, RevertableDelete, RevertableOp, delete_prefix
from lbry.wallet.server.db import DB_PREFIXES
from lbry.wallet.server.db.prefixes import Prefixes
from lbry.wallet.server.db.prefixes import Prefixes, PendingActivationValue, ClaimTakeoverValue, ClaimToTXOValue
from lbry.wallet.server.db.claimtrie import StagedClaimtrieItem, get_update_effective_amount_ops, length_encoded_name
from lbry.wallet.server.db.claimtrie import get_expiration_height
from lbry.wallet.server.db.claimtrie import get_expiration_height, get_delay_for_name
from lbry.wallet.server.db.elasticsearch import SearchIndex
@ -79,7 +82,7 @@ class FlushData:
adds = attr.ib()
deletes = attr.ib()
tip = attr.ib()
undo_claimtrie = attr.ib()
undo = attr.ib()
class ResolveResult(typing.NamedTuple):
@ -89,6 +92,7 @@ class ResolveResult(typing.NamedTuple):
position: int
tx_hash: bytes
height: int
amount: int
short_url: str
is_controlling: bool
canonical_url: str
@ -97,12 +101,14 @@ class ResolveResult(typing.NamedTuple):
expiration_height: int
effective_amount: int
support_amount: int
last_take_over_height: Optional[int]
last_takeover_height: Optional[int]
claims_in_channel: Optional[int]
channel_hash: Optional[bytes]
reposted_claim_hash: Optional[bytes]
OptionalResolveResultOrError = Optional[typing.Union[ResolveResult, LookupError, ValueError]]
DB_STATE_STRUCT = struct.Struct(b'>32sLL32sHLBBlll')
DB_STATE_STRUCT_SIZE = 92
@ -183,12 +189,10 @@ class LevelDB:
self.search_index = SearchIndex(self.env.es_index_prefix, self.env.database_query_timeout)
def claim_hash_and_name_from_txo(self, tx_num: int, tx_idx: int):
claim_hash_and_name = self.db.get(
DB_PREFIXES.txo_to_claim.value + TXO_STRUCT_pack(tx_num, tx_idx)
)
claim_hash_and_name = self.db.get(Prefixes.txo_to_claim.pack_key(tx_num, tx_idx))
if not claim_hash_and_name:
return
return claim_hash_and_name[:CLAIM_HASH_LEN], claim_hash_and_name[CLAIM_HASH_LEN:]
return Prefixes.txo_to_claim.unpack_value(claim_hash_and_name)
def get_supported_claim_from_txo(self, tx_num: int, position: int) -> typing.Tuple[Optional[bytes], Optional[int]]:
key = Prefixes.support_to_claim.pack_key(tx_num, position)
@ -213,20 +217,22 @@ class LevelDB:
unpacked_k = Prefixes.claim_to_support.unpack_key(k)
unpacked_v = Prefixes.claim_to_support.unpack_value(v)
supports.append((unpacked_k.tx_num, unpacked_k.position, unpacked_v.amount))
return supports
def _prepare_resolve_result(self, tx_num: int, position: int, claim_hash: bytes, name: str, root_tx_num: int,
root_position: int) -> ResolveResult:
root_position: int, activation_height: int) -> ResolveResult:
controlling_claim = self.get_controlling_claim(name)
tx_hash = self.total_transactions[tx_num]
height = bisect_right(self.tx_counts, tx_num)
created_height = bisect_right(self.tx_counts, root_tx_num)
last_take_over_height = 0
activation_height = created_height
last_take_over_height = controlling_claim.height
expiration_height = get_expiration_height(height)
support_amount = self.get_support_amount(claim_hash)
effective_amount = self.get_effective_amount(claim_hash)
claim_amount = self.get_claim_txo_amount(claim_hash, tx_num, position)
effective_amount = support_amount + claim_amount
channel_hash = self.get_channel_for_claim(claim_hash)
claims_in_channel = None
@ -235,27 +241,35 @@ class LevelDB:
if channel_hash:
channel_vals = self.get_root_claim_txo_and_current_amount(channel_hash)
if channel_vals:
_, _, _, channel_name, _, _ = channel_vals
channel_name = channel_vals[1].name
claims_in_channel = self.get_claims_in_channel_count(channel_hash)
canonical_url = f'{channel_name}#{channel_hash.hex()}/{name}#{claim_hash.hex()}'
return ResolveResult(
name, claim_hash, tx_num, position, tx_hash, height, short_url=short_url,
is_controlling=False, canonical_url=canonical_url, last_take_over_height=last_take_over_height,
claims_in_channel=claims_in_channel, creation_height=created_height, activation_height=activation_height,
name, claim_hash, tx_num, position, tx_hash, height, claim_amount, short_url=short_url,
is_controlling=controlling_claim.claim_hash == claim_hash, canonical_url=canonical_url,
last_takeover_height=last_take_over_height, claims_in_channel=claims_in_channel,
creation_height=created_height, activation_height=activation_height,
expiration_height=expiration_height, effective_amount=effective_amount, support_amount=support_amount,
channel_hash=channel_hash, reposted_claim_hash=None
)
def _resolve(self, normalized_name: str, claim_id: Optional[str] = None,
amount_order: int = 1) -> Optional[ResolveResult]:
amount_order: Optional[int] = None) -> Optional[ResolveResult]:
"""
:param normalized_name: name
:param claim_id: partial or complete claim id
:param amount_order: '$<value>' suffix to a url, defaults to 1 (winning) if no claim id modifier is provided
"""
if not amount_order and not claim_id:
# winning resolution
controlling = self.get_controlling_claim(normalized_name)
if not controlling:
return
return self._fs_get_claim_by_hash(controlling.claim_hash)
encoded_name = length_encoded_name(normalized_name)
amount_order = max(int(amount_order or 1), 1)
if claim_id:
# resolve by partial/complete claim id
short_claim_hash = bytes.fromhex(claim_id)
@ -263,19 +277,22 @@ class LevelDB:
for k, v in self.db.iterator(prefix=prefix):
key = Prefixes.claim_short_id.unpack_key(k)
claim_txo = Prefixes.claim_short_id.unpack_value(v)
return self._prepare_resolve_result(claim_txo.tx_num, claim_txo.position, key.claim_hash, key.name,
key.root_tx_num, key.root_position)
return self._prepare_resolve_result(
claim_txo.tx_num, claim_txo.position, key.claim_hash, key.name, key.root_tx_num,
key.root_position, claim_txo.activation
)
return
# resolve by amount ordering, 1 indexed
for idx, (k, v) in enumerate(self.db.iterator(prefix=DB_PREFIXES.claim_effective_amount_prefix.value + encoded_name)):
for idx, (k, v) in enumerate(self.db.iterator(
prefix=DB_PREFIXES.claim_effective_amount_prefix.value + encoded_name)):
if amount_order > idx + 1:
continue
key = Prefixes.claim_effective_amount.unpack_key(k)
claim_val = Prefixes.claim_effective_amount.unpack_value(v)
return self._prepare_resolve_result(
key.tx_num, key.position, claim_val.claim_hash, key.name, claim_val.root_tx_num,
claim_val.root_position
claim_val.root_position, claim_val.activation
)
return
@ -293,7 +310,7 @@ class LevelDB:
return
return list(sorted(candidates, key=lambda item: item[1]))[0]
def _fs_resolve(self, url):
def _fs_resolve(self, url) -> typing.Tuple[OptionalResolveResultOrError, OptionalResolveResultOrError]:
try:
parsed = URL.parse(url)
except ValueError as e:
@ -326,7 +343,7 @@ class LevelDB:
return resolved_stream, resolved_channel
async def fs_resolve(self, url):
async def fs_resolve(self, url) -> typing.Tuple[OptionalResolveResultOrError, OptionalResolveResultOrError]:
return await asyncio.get_event_loop().run_in_executor(self.executor, self._fs_resolve, url)
def _fs_get_claim_by_hash(self, claim_hash):
@ -335,7 +352,7 @@ class LevelDB:
unpacked_v = Prefixes.claim_to_txo.unpack_value(v)
return self._prepare_resolve_result(
unpacked_k.tx_num, unpacked_k.position, unpacked_k.claim_hash, unpacked_v.name,
unpacked_v.root_tx_num, unpacked_v.root_position
unpacked_v.root_tx_num, unpacked_v.root_position, unpacked_v.activation
)
async def fs_getclaimbyid(self, claim_id):
@ -352,17 +369,21 @@ class LevelDB:
for k, v in self.db.iterator(prefix=DB_PREFIXES.claim_to_txo.value + claim_hash):
unpacked_k = Prefixes.claim_to_txo.unpack_key(k)
unpacked_v = Prefixes.claim_to_txo.unpack_value(v)
return unpacked_v.root_tx_num, unpacked_v.root_position, unpacked_v.amount, unpacked_v.name,\
unpacked_k.tx_num, unpacked_k.position
return unpacked_k, unpacked_v
def make_staged_claim_item(self, claim_hash: bytes) -> StagedClaimtrieItem:
root_tx_num, root_idx, value, name, tx_num, idx = self.db.get_root_claim_txo_and_current_amount(
claim_hash
)
def make_staged_claim_item(self, claim_hash: bytes) -> Optional[StagedClaimtrieItem]:
claim_info = self.get_root_claim_txo_and_current_amount(claim_hash)
k, v = claim_info
root_tx_num = v.root_tx_num
root_idx = v.root_position
value = v.amount
name = v.name
tx_num = k.tx_num
idx = k.position
height = bisect_right(self.tx_counts, tx_num)
effective_amount = self.db.get_support_amount(claim_hash) + value
effective_amount = self.get_support_amount(claim_hash) + value
signing_hash = self.get_channel_for_claim(claim_hash)
activation_height = 0
activation_height = v.activation
if signing_hash:
count = self.get_claims_in_channel_count(signing_hash)
else:
@ -372,17 +393,36 @@ class LevelDB:
root_tx_num, root_idx, signing_hash, count
)
def get_effective_amount(self, claim_hash):
def get_claim_txo_amount(self, claim_hash: bytes, tx_num: int, position: int) -> Optional[int]:
v = self.db.get(Prefixes.claim_to_txo.pack_key(claim_hash, tx_num, position))
if v:
return Prefixes.claim_to_txo.unpack_value(v).amount
def get_claim_from_txo(self, claim_hash: bytes) -> Optional[ClaimToTXOValue]:
assert claim_hash
for v in self.db.iterator(prefix=DB_PREFIXES.claim_to_txo.value + claim_hash, include_key=False):
return Prefixes.claim_to_txo.unpack_value(v).amount + self.get_support_amount(claim_hash)
fnord
return None
return Prefixes.claim_to_txo.unpack_value(v)
def get_claim_amount(self, claim_hash: bytes) -> Optional[int]:
claim = self.get_claim_from_txo(claim_hash)
if claim:
return claim.amount
def get_effective_amount(self, claim_hash: bytes):
return (self.get_claim_amount(claim_hash) or 0) + self.get_support_amount(claim_hash)
def get_update_effective_amount_ops(self, claim_hash: bytes, effective_amount: int):
claim_info = self.get_root_claim_txo_and_current_amount(claim_hash)
if not claim_info:
return []
root_tx_num, root_position, amount, name, tx_num, position = claim_info
root_tx_num = claim_info[1].root_tx_num
root_position = claim_info[1].root_position
amount = claim_info[1].amount
name = claim_info[1].name
tx_num = claim_info[0].tx_num
position = claim_info[0].position
activation = claim_info[1].activation
signing_hash = self.get_channel_for_claim(claim_hash)
claims_in_channel_count = None
if signing_hash:
@ -390,7 +430,8 @@ class LevelDB:
prev_effective_amount = self.get_effective_amount(claim_hash)
return get_update_effective_amount_ops(
name, effective_amount, prev_effective_amount, tx_num, position,
root_tx_num, root_position, claim_hash, signing_hash, claims_in_channel_count
root_tx_num, root_position, claim_hash, activation, activation, signing_hash,
claims_in_channel_count
)
def get_claims_in_channel_count(self, channel_hash) -> int:
@ -415,6 +456,35 @@ class LevelDB:
)
return expired
def get_controlling_claim(self, name: str) -> Optional[ClaimTakeoverValue]:
controlling = self.db.get(Prefixes.claim_takeover.pack_key(name))
if not controlling:
return
return Prefixes.claim_takeover.unpack_value(controlling)
def get_claims_for_name(self, name: str):
claim_hashes = set()
for k in self.db.iterator(prefix=Prefixes.claim_short_id.prefix + length_encoded_name(name),
include_value=False):
claim_hashes.add(Prefixes.claim_short_id.unpack_key(k).claim_hash)
return claim_hashes
def get_activated_claims_at_height(self, height: int) -> typing.Set[PendingActivationValue]:
claims = set()
prefix = Prefixes.pending_activation.prefix + height.to_bytes(4, byteorder='big')
for _v in self.db.iterator(prefix=prefix, include_key=False):
v = Prefixes.pending_activation.unpack_value(_v)
claims.add(v)
return claims
def get_activation_delay(self, claim_hash: bytes, name: str) -> int:
controlling = self.get_controlling_claim(name)
if not controlling:
return 0
if claim_hash == controlling.claim_hash:
return 0
return get_delay_for_name(self.db_height - controlling.height)
async def _read_tx_counts(self):
if self.tx_counts is not None:
return
@ -685,9 +755,9 @@ class LevelDB:
batch_delete(staged_change.key)
flush_data.claimtrie_stash.clear()
for undo_claims, height in flush_data.undo_claimtrie:
batch_put(DB_PREFIXES.undo_claimtrie.value + util.pack_be_uint64(height), undo_claims)
flush_data.undo_claimtrie.clear()
for undo_ops, height in flush_data.undo:
batch_put(DB_PREFIXES.undo_claimtrie.value + util.pack_be_uint64(height), undo_ops)
flush_data.undo.clear()
self.fs_height = flush_data.height
self.fs_tx_count = flush_data.tx_count
@ -788,11 +858,10 @@ class LevelDB:
claim_reorg_height = self.fs_height
# print("flush undos", flush_data.undo_claimtrie)
for (ops, height) in reversed(flush_data.undo_claimtrie):
claimtrie_ops = RevertableOp.unpack_stack(ops)
print("%i undo ops for %i" % (len(claimtrie_ops), height))
for op in reversed(claimtrie_ops):
print("REWIND", op)
for (packed_ops, height) in reversed(flush_data.undo):
undo_ops = RevertableOp.unpack_stack(packed_ops)
for op in reversed(undo_ops):
# print("REWIND", op)
if op.is_put:
batch_put(op.key, op.value)
else:
@ -800,7 +869,7 @@ class LevelDB:
batch_delete(DB_PREFIXES.undo_claimtrie.value + util.pack_be_uint64(claim_reorg_height))
claim_reorg_height -= 1
flush_data.undo_claimtrie.clear()
flush_data.undo.clear()
flush_data.claimtrie_stash.clear()
while self.fs_height > flush_data.height:
@ -828,9 +897,9 @@ class LevelDB:
batch_put(key, value)
# New undo information
for undo_info, height in flush_data.undo_infos:
for undo_info, height in flush_data.undo:
batch.put(self.undo_key(height), b''.join(undo_info))
flush_data.undo_infos.clear()
flush_data.undo.clear()
# Spends
for key in sorted(flush_data.deletes):
@ -1023,9 +1092,9 @@ class LevelDB:
"""Returns a height from which we should store undo info."""
return max_height - self.env.reorg_limit + 1
def undo_key(self, height):
def undo_key(self, height: int) -> bytes:
"""DB key for undo information at the given height."""
return UNDO_PREFIX + pack('>I', height)
return DB_PREFIXES.UNDO_PREFIX.value + pack('>I', height)
def read_undo_info(self, height):
"""Read undo information from a file for the current height."""