Caching the base58check encoded version of an `AccountID` has
performance advantages, because because of the computationally
heavy cost associated with the conversion, which requires the
application of SHA-256 twice.
This commit makes the cache significantly more efficient in terms
of memory used: it eliminates the map, using a vector with a size
that is determined by the configured size of the node, and a hash
function to directly map any given `AccountID` to a specific slot
in the cache; the eviction policy is simple: in case of collision
the existing entry is removed and replaced with the new data.
Previously, use of the cache was optional and required additional
effort by the programmer. Now the cache is automatic and does not
require any additional work or information.
The new cache also utilizes a 64-way spinlock, to help reduce any
contention that the pressure on the cache would impose.
* "A path is considered invalid if and only if it enters and exits an
address node through trust lines where No Ripple has been enabled for
that address." (https://xrpl.org/rippling.html#specifics)
* When loading trust lines for an account "Alice" which was reached
via a trust line that has the No Ripple flag set on Alice's side, do
not use or cache any of Alice's trust lines which have the No Ripple
flag set on Alice's side. For typical "end-user" accounts, this will
return no trust lines.
* Abort background path finding when closed or disconnected
* Exit pathfinding job thread if there are no requests left
* Don't bother creating the path find job if there are no requests
* Refactor to remove circular dependency between InfoSub and PathRequest
* Fix bug where incorrect max amount was set for XRP
* Fix bug where incorrect source currencies were set when XRP was the dst and a
sendmax amount was set
* Force jtx to request/receive the 2.0 API
* Force the JSON and WebSocket tests to use 2.0 API
* This specifically allows the Websocket to create 2.0 json/ripple
and get back a 2.0 response.
* Add test for malformed json2
* Add check for parse failure
* Add check for params to be in array form.
* Correct type-o discovered in tests due to stricter checking.
* Add API version to the WSClient & JSONRPCClient test
* Update source.dox with more headers
Replace Journal public data members with member function accessors
in order to make Journal lighter weight. The change makes a
Journal cheaper to pass by value.
Also add missing stream checks (e.g., calls to JLOG) to avoid
text processing that ultimately will not be stored in the log.
The IP address used to perform pathfinding operations is now charged an
additional resource increment for each source currency in the path set.
* NOTE: This charge is a local resource charge, not a transaction fee
charge.
Add a new algorithm for finding the liquidity in a payment path. There
is still a reverse and forward pass, but the forward pass starts at the
limiting step rather than the payment source. This insures the limiting
step is completely consumed rather than potentially leaving a 'dust'
amount in the forward pass.
Each step in a payment is either a book step, a direct step (account to
account step), or an xrp endpoint. Each step in the existing
implementation is a triple, where each element in the triple is either
an account of a book, for a total of eight step types.
Since accounts are considered in pairs, rather than triples, transfer
fees are handled differently. In V1 of payments, in the payment path
A -> gw ->B, if A redeems to gw, and gw issues to B, a transfer fee is
changed. In the new code, a transfer fee is changed even if A issues to
gw.
The View hierarchy of classes is reorganized to include new
classes with member functions moved and renamed, to solve
defects in the original design:
OpenView accumulates raw state and tx changes and
can be applied to the base. ApplyView accumulates changes
for a single transaction, including metadata, and can be
applied to an OpenView. The Sandbox allows changes with
the option to apply or throw them out. The PaymentSandbox
provides a sandbox with account credit deferral.
Call sites are changed to use the class appropriate for
the task.
Many functions and classes that used a Ledger now use a BasicView.
Calls to cachedRead are changed to call member read on the view,
note that this bypasses the SLECache optimization. To restore the
optimization, the BasicView passed at the top of call stacks
should be wrapped with a caching view, coming in future commits.
This tidies up the View interface and makes transaction
application a free function, with the removal of the
TransactionEngine class. A new class ApplyContext provides
all the state information needed to apply a Transactor. The
Transactor is refactored to perform all the processing
activities previously part of TransactionEngine.
The calculation of metadata from a MetaView is improved.
A new apply function performs all the steps for calculating
and inserting metadata into the tx map.
Transaction processing code path is passed a Config instead
of retrieving the global, and uses the Journal supplied in
the call to apply() consistently.
To support transaction processing and RPC operations, a
new POD type ViewInfo is added which consolidates static
information about open and closed ledgers, such as the ledger
sequence number or the closing times. Ledger and MetaView are
refactored to use this info.
The ViewInfo now contains the "open ledger" setting. The
tapOPEN_LEDGER ViewFlag is removed. The view property of
being an open ledger is obtained from the base or by using
the MetaView constructor which presents a closed ledger as
an open one.
View, MetaView:
* Fix missing includes
* Add apply free function
* Use Journal in TransactionEngine
* Use BasicView in TransactionEngine
* inline NetworkOPs::batchApply
* Add shallow_copy, open_ledger MetaView ctor tags
* Add ViewInfo with open flag, seq, close times
* Make parent_ a reference
* Tidy up ctor arguments and base_ name
* Remove tapOPEN_LEDGER
* add assert to MetaView::apply
* ViewInfo comment
* Throw, pass Journal in txInsert
* Add BasicView::txCount
TransactionEngine:
* Add apply
* Make TransactionEngine private
* Refactor MetaView::apply and apply()
* Rename to TxMeta
* Refactor treatment of metadata in MetaView, TransactionEngine
* Rename to ApplyContext
* Use ApplyContext& in Transactor
* Pass Config in ApplyContext
* Declare Transactor classes in headers
* Use view flags in Transactor
This shores up the View interface support for contextual
transaction processing by putting params in the View, and
provides support for replacing the open ledger with the
open MetaView.
Transaction metadata is now part of the View interface.
Stacked MetaViews correctly apply their transaction
metadata to the parent.
* Add lastCloseTime to View
* Add insertTx to View, implement in MetaView
* Add View::txExists for transaction checking
* Add Fees to View, cache fees in Ledger and MetaView
* Use ViewFlags in View
* Use tapENABLE_TESTING flag for features
* Use cached Fees in View
* Rename to ViewFlags
* Move FreezeHandling to View.h, remove ViewAPIBasics.h
* Remove BasicView::parent hack
* Remove calls to getLedger in Transactors
Handle legacy (ripple_path_find) requests that don't specify a ledger
the same way regular path_find requests are. This provides a
performance improvement for these requests and reduces the problem
of server busy errors.
Conflicts:
src/ripple/app/paths/PathRequest.cpp
All AccountID functionality is removed from RippleAddress and
replaced with free functions. The AccountID to string conversion
cache is factored out as an explicit type with an instance in
the Application object. New base58 conversion functions are used,
with no dependence on OpenSSL.
All types and free functions related to AccountID are consolidated
into one header file. Routines to operate on "tokens" are also
introduced and consolidated into a single header file.
A token one of the cryptographic primitives used in Ripple:
Secret Seed
Server Public Key
Server Secret Key
Account ID
Account Public Key
Account Private Key
and these deprecated primitives:
Account Family Seed
Account Family Generator
Member functions and free functions on Ledger and LedgerEntrySet are
rewritten in terms of new abstract interfaces `BasicView` and `View`,
representing the set of non-decomposable primitives necessary to read
and write state map items in a ledger, and to overlay a discardable
view onto a Ledger that can calculate metadata during transaction
processing. const-correctness is enforced through the parameter and
return types.
The MetaView now supports multi-level stacking: A MetaView can be
stacked on top of either a Ledger or another MetaView, up to any
number of levels.
The getSLEi member function is removed. The CachedView wrapper
replaces it, wrapping a View such that any function called with a
CachedView will go through the SLECache.
* Add BasicView, View, CachedView
* Rename LedgerEntrySet to MetaView
* Factor out free functions
* Consolidate free functions in ViewAPI
* Remove unused class members and free functions
* Remove duplicate:
This changes behavior to fix an apparent bug. The
copy will now correctly inherit mParams instead
of reverting to tapNONE
* Tidy up LedgerEntrySet declarations
* Tidy up TransactionEngine
* Tidy PathCursor declarations
* Add LedgerEntrySet::apply
* Add LedgerEntrySet ctor
* Add Keylet, keylet namespace
* Add defaulted copy members
* Use optional in TransactionEngine
* Use optional<LedgerEntrySet> in PathState
* Return shared_ptr in Ledger::fetch
* Don't call entryCache with zero
* Deprecate invalidate
* Remove default constructor
* Remove unused container API
* Remove CountedObject base class
* Remove insert, clear
* Remove entryCreate overload
* Remove unused and tidy up STLedgerEntry
* Make getEntry private and tidy
* Replace members with adjustOwnerCount free function
* Replace accountFunds with funds free function
This performs a deep refactor on the Ledger class and its supporting
classes, in preparation for the move to shared_ptr<SLE const> in
places where the SLE is immutable and we are currently using
shared_ptr<SLE>.
Member functions are converted to free functions, the SLECache is an
explicit parameter, one line convenience functions are removed to
streamline the interface. Some callers are changed to use <SLE const>
instead of <SLE>
SLECache:
* Moved to its own header file
RippleState:
* Remove unused functions
* Store the SLE as const
* Simplify callers
AccountState:
* Remove unused members
* Simplify existing members
Ledger:
* Replace writeBack with insert and update
* Remove unused functions
* Remove LedgerStateParams
* Move getLastFullLedger to Application
* add entryCacheI, exists, fetch, erase
* Use boost::optional where it makes sense
* Make member functions free functions
Free functions:
* fetch: cache-aware SLE retrieval
* forEachItem, forEachItemAfter
* (various)
