nimbus-eth1/nimbus/utils/lru_cache.nim

302 lines
11 KiB
Nim
Raw Normal View History

# Nimbus
# Copyright (c) 2018 Status Research & Development GmbH
# Licensed under either of
# * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
# http://www.apache.org/licenses/LICENSE-2.0)
# * MIT license ([LICENSE-MIT](LICENSE-MIT) or
# http://opensource.org/licenses/MIT)
# at your option. This file may not be copied, modified, or distributed except
# according to those terms.
## Hash as hash can: LRU cache
## ===========================
##
## This module provides a generic last-recently-used cache data structure.
##
## The implementation works with the same complexity as the worst case of a
## nim hash tables operation. This is is assumed to be O(1) in most cases
## (so long as the table does not degrade into one-bucket linear mode, or
## some bucket-adjustment algorithm takes over.)
##
## For consistency with every other data type in Nim these have value
## semantics, this means that `=` performs a deep copy of the LRU cache.
##
import
math,
eth/rlp,
stew/results,
tables
export
results
type
LruKey*[T,K] = ## User provided handler function, derives an
## LRU `key` from function argument `arg`. The
## `key` is used to index the cache data.
proc(arg: T): K {.gcsafe, raises: [Defect,CatchableError].}
LruValue*[T,V,E] = ## User provided handler function, derives an
## LRU `value` from function argument `arg`.
proc(arg: T): Result[V,E] {.gcsafe, raises: [Defect,CatchableError].}
LruItem*[K,V] = ## Doubly linked hash-tab item encapsulating
## the `value` (which is the result from
## `LruValue` handler function.
tuple[prv, nxt: K, value: V]
# There could be {.rlpCustomSerialization.} annotation for the tab field.
# As there was a problem with the automatic Rlp serialisation for generic
# type, the easier solution was an all manual read()/append() for the whole
# generic LruCacheData[K,V] type.
LruData[K,V] = object
maxItems: int ## Max number of entries
first, last: K ## Doubly linked item list queue
tab: Table[K,LruItem[K,V]] ## (`key`,encapsulated(`value`)) data table
LruCache*[T,K,V,E] = object
data*: LruData[K,V] ## Cache data, can be serialised
toKey: LruKey[T,K] ## Handler function, derives `key`
toValue: LruValue[T,V,E] ## Handler function, derives `value`
{.push raises: [Defect].}
# ------------------------------------------------------------------------------
# Private functions
# ------------------------------------------------------------------------------
proc `==`[K,V](a, b: var LruData[K,V]): bool =
a.maxItems == b.maxItems and
a.first == b.first and
a.last == b.last and
a.tab == b.tab
# ------------------------------------------------------------------------------
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
# Public constructor and reset
# ------------------------------------------------------------------------------
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
proc clearCache*[T,K,V,E](cache: var LruCache[T,K,V,E]; cacheInitSize = 0)
{.gcsafe, raises: [Defect].} =
## Reset/clear an initialised LRU cache. The cache will be re-allocated
## with `cacheInitSize` initial spaces if this is positive, or `cacheMaxItems`
## spaces (see `initLruCache()`) as a default.
var initSize = cacheInitSize
if initSize <= 0:
initSize = cache.data.maxItems
cache.data.first.reset
cache.data.last.reset
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
cache.data.tab = initTable[K,LruItem[K,V]](initSize.nextPowerOfTwo)
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
proc initCache*[T,K,V,E](cache: var LruCache[T,K,V,E];
toKey: LruKey[T,K], toValue: LruValue[T,V,E];
cacheMaxItems = 10; cacheInitSize = 0)
{.gcsafe, raises: [Defect].} =
## Initialise LRU cache. The handlers `toKey()` and `toValue()` are explained
## at the data type definition. The cache will be allocated with
## `cacheInitSize` initial spaces if this is positive, or `cacheMaxItems`
## spaces (see `initLruCache()`) as a default.
cache.data.maxItems = cacheMaxItems
cache.toKey = toKey
cache.toValue = toValue
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
cache.clearCache
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
# ------------------------------------------------------------------------------
# Public functions, basic mechanism
# ------------------------------------------------------------------------------
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
proc getItem*[T,K,V,E](lru: var LruCache[T,K,V,E];
2022-04-08 04:54:11 +00:00
arg: T; peekOk = false): Result[V,E]
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
{.gcsafe, raises: [Defect,CatchableError].} =
## If the key `lru.toKey(arg)` is a cached key, the associated value will
## be returnd. If the `peekOK` argument equals `false`, the associated
## key-value pair will have been moved to the end of the LRU queue.
##
## If the key `lru.toKey(arg)` is not a cached key and the LRU queue has at
## least `cacheMaxItems` entries (see `initLruCache()`, the first key-value
## pair will be removed from the LRU queue. Then the value the pair
## (`lru.toKey(arg)`,`lru.toValue(arg)`) will be appended to the LRU queue
## and the value part returned.
##
let key = lru.toKey(arg)
# Relink item if already in the cache => move to last position
if lru.data.tab.hasKey(key):
let lruItem = lru.data.tab[key]
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
if peekOk or key == lru.data.last:
# Nothing to do
return ok(lruItem.value)
# Unlink key Item
if key == lru.data.first:
lru.data.first = lruItem.nxt
else:
lru.data.tab[lruItem.prv].nxt = lruItem.nxt
lru.data.tab[lruItem.nxt].prv = lruItem.prv
# Append key item
lru.data.tab[lru.data.last].nxt = key
lru.data.tab[key].prv = lru.data.last
lru.data.last = key
return ok(lruItem.value)
# Calculate value, pass through error unless OK
let rcValue = ? lru.toValue(arg)
# Limit number of cached items
if lru.data.maxItems <= lru.data.tab.len:
# Delete oldest/first entry
var nextKey = lru.data.tab[lru.data.first].nxt
lru.data.tab.del(lru.data.first)
lru.data.first = nextKey
# Add cache entry
var tabItem: LruItem[K,V]
# Initialise empty queue
if lru.data.tab.len == 0:
lru.data.first = key
lru.data.last = key
else:
# Append queue item
lru.data.tab[lru.data.last].nxt = key
tabItem.prv = lru.data.last
lru.data.last = key
tabItem.value = rcValue
lru.data.tab[key] = tabItem
result = ok(rcValue)
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
# ------------------------------------------------------------------------------
# Public functions, cache info
# ------------------------------------------------------------------------------
proc hasKey*[T,K,V,E](lru: var LruCache[T,K,V,E]; arg: T): bool {.gcsafe.} =
## Check whether the `arg` argument is cached
let key = lru.toKey(arg)
lru.data.tab.hasKey(key)
proc firstKey*[T,K,V,E](lru: var LruCache[T,K,V,E]): K {.gcsafe.} =
## Returns the key of the first item in the LRU queue, or the reset
## value it the cache is empty.
if 0 < lru.data.tab.len:
result = lru.data.first
proc lastKey*[T,K,V,E](lru: var LruCache[T,K,V,E]): K {.gcsafe.} =
## Returns the key of the last item in the LRU queue, or the reset
## value it the cache is empty.
if 0 < lru.data.tab.len:
result = lru.data.last
proc maxLen*[T,K,V,E](lru: var LruCache[T,K,V,E]): int {.gcsafe.} =
## Maximal number of cache entries.
lru.data.maxItems
proc len*[T,K,V,E](lru: var LruCache[T,K,V,E]): int {.gcsafe.} =
## Return the number of elements in the cache.
lru.data.tab.len
# ------------------------------------------------------------------------------
# Public functions, advanced features
# ------------------------------------------------------------------------------
proc setItem*[T,K,V,E](lru: var LruCache[T,K,V,E]; arg: T; value: V): bool
{.gcsafe, raises: [Defect,CatchableError].} =
## Update entry with key `lru.toKey(arg)` by `value`. Reurns `true` if the
## key exists in the database, and false otherwise.
##
## This function allows for simlifying the `toValue()` function (see
## `initLruCache()`) to provide a placeholder only and later fill this
## slot with this `setLruItem()` function.
let key = lru.toKey(arg)
if lru.data.tab.hasKey(key):
lru.data.tab[key].value = value
return true
2022-04-08 04:54:11 +00:00
Feature/implement poa processing (#748) * re-shuffled Clique functions why: Due to the port from the go-sources, the interface logic is not optimal for nimbus. The main visible function is currently snapshot() and most of the _procurement_ of this function result has been moved to a sub-directory. * run eip-225 Clique test against p2p/chain.persistBlocks() why: Previously, loading the test block chains was fugdged with the purpose only to fill the database. As it is now clear how nimbus works on Goerli, the same can be achieved with a more realistic scenario. details: Eventually these tests will be pre-cursor to the reply tests for the Goerli chain supporting TDD approach with more simple cases. * fix exception annotations for executor module why: needed for exception tracking details: main annoyance are vmState methods (in state.nim) which potentially throw a base level Exception (a proc would only throws CatchableError) * split p2p/chain into sub-modules and fix exception annotations why: make space for implementing PoA stuff * provide over-loadable Clique PRNG why: There is a PRNG provided for generating reproducible number sequences. The functions which employ the PRNG to generate time slots were ported ported from the go-implementation. They are currently unused. * implement trusted signer assembly in p2p/chain.persistBlocks() details: * PoA processing moved there at the end of a transaction. Currently, there is no action (eg. transaction rollback) if this fails. * The unit tests with staged blocks work ok. In particular, there should be tests with to-be-rejected blocks. * TODO: 1.Optimise throughput/cache handling; 2.Verify headers * fix statement cast in pool.nim * added table features to LRU cache why: Clique uses the LRU cache using a mixture of volatile online items from the LRU cache and database checkpoints for hard synchronisation. For performance, Clique needs more table like features. details: First, last, and query key added, as well as efficient random delete added. Also key-item pair iterator added for debugging. * re-factored LRU snapshot caching why: Caching was sub-optimal (aka. bonkers) in that it skipped over memory caches in many cases and so mostly rebuild the snapshot from the last on-disk checkpoint. details; The LRU snapshot toValue() handler has been moved into the module clique_snapshot. This is for the fact that toValue() is not supposed to see the whole LRU cache database. So there must be a higher layer working with the the whole LRU cache and the on-disk checkpoint database. also: some clean up todo: The code still assumes that the block headers are valid in itself. This is particular important when an epoch header (aka re-sync header) is processed as it must contain the PoA result of all previous headers. So blocks need to be verified when they come in before used for PoA processing. * fix some snapshot cache fringe cases why: Must not index empty sequences in clique_snapshot module
2021-07-14 15:13:27 +00:00
proc delItem*[T,K,V,E](lru: var LruCache[T,K,V,E]; arg: T): bool
{.gcsafe, discardable, raises: [Defect,KeyError].} =
## Delete the `arg` argument from cached. That way, the LRU cache can
## be re-purposed as a sequence with efficient random delete facility.
let key = lru.toKey(arg)
# Relink item if already in the cache => move to last position
if lru.data.tab.hasKey(key):
let lruItem = lru.data.tab[key]
# Unlink key Item
if lru.data.tab.len == 1:
lru.data.first.reset
lru.data.last.reset
elif key == lru.data.last:
lru.data.last = lruItem.prv
elif key == lru.data.first:
lru.data.first = lruItem.nxt
else:
lru.data.tab[lruItem.prv].nxt = lruItem.nxt
lru.data.tab[lruItem.nxt].prv = lruItem.prv
lru.data.tab.del(key)
return true
iterator keyItemPairs*[T,K,V,E](lru: var LruCache[T,K,V,E]): (K,LruItem[K,V])
{.gcsafe, raises: [Defect,CatchableError].} =
## Cycle through all (key,lruItem) pairs in chronological order.
if 0 < lru.data.tab.len:
var key = lru.data.first
for _ in 0 ..< lru.data.tab.len - 1:
var item = lru.data.tab[key]
yield (key, item)
key = item.nxt
yield (key, lru.data.tab[key])
if key != lru.data.last:
raiseAssert "Garbled LRU cache next/prv references"
# ------------------------------------------------------------------------------
# Public functions, RLP support
# ------------------------------------------------------------------------------
proc `==`*[T,K,V,E](a, b: var LruCache[T,K,V,E]): bool =
## Returns `true` if both argument LRU caches contain the same data
## regardless of `toKey()`/`toValue()` handler functions.
a.data == b.data
proc append*[K,V](rw: var RlpWriter; data: LruData[K,V]) {.
inline, raises: [Defect,KeyError].} =
## Generic support for `rlp.encode(lru.data)` for serialising the data
## part of an LRU cache.
rw.append(data.maxItems)
rw.append(data.first)
rw.append(data.last)
rw.startList(data.tab.len)
# store keys in LRU order
if 0 < data.tab.len:
var key = data.first
for _ in 0 ..< data.tab.len - 1:
var value = data.tab[key]
rw.append((key, value))
key = value.nxt
rw.append((key, data.tab[key]))
if key != data.last:
raiseAssert "Garbled LRU cache next/prv references"
proc read*[K,V](rlp: var Rlp; Q: type LruData[K,V]): Q {.
inline, raises: [Defect,RlpError].} =
## Generic support for `rlp.decode(bytes)` for loading the data part
## of an LRU cache from a serialised data stream.
result.maxItems = rlp.read(int)
result.first = rlp.read(K)
result.last = rlp.read(K)
for w in rlp.items:
let (key,value) = w.read((K,LruItem[K,V]))
result.tab[key] = value
# ------------------------------------------------------------------------------
# End
# ------------------------------------------------------------------------------