hkalodner released this
Nov 7, 2018
Expanded iterator and range functionality to return NumPy arrays.
Many methods and properties of BlockSci objects return range or iterator objects such as blocksci.TxRange. These objects allow vectorized operations over sequences of BlockSci objects. Their API matches up with the API of their member objects, and thus blocksci.TxRange has almost the same set of methods as blocksci.Tx. These methods will efficiently call the given method over all items in the range or iterator. Depending on the return type of the method, the result will either be another range, a NumPy array, or a python list. For further information, look for these classes in the reference.
Add custom BlockSci pickler to enable sending and receiving serialized BlockSci objects. This means that returning BlockSci objects from the multiprocessing interface now works correctly.
Enhance the change address heuristics interface
Change address heuristics are now composible in order to form new customized heuristics using the blocksci.heuristics.change.ChangeHeuristic interface. These can be used in combination with the new clustering interface described below.
Incorporate clustering module into main BlockSci library
The formerly external clustering module is now avaiable as blocksci.cluster. Further, it is now possible to generate new clusterings through the python interface using the ~blocksci.cluster.ClusterManager.create_clustering method. Users can select their choice of change address heuristic in order to experiment with different clustering strategies.
Simplified build system
BlockSci's install process no longer requires the compilation of any external dependencies to compile on Ubuntu 16.04. The BlockSci library no longer has any public dependencies so compiling against it will not require linking against anything else.
The CMake build script has now been updated to install a Config file which allows you to use find_package(blocksci) to import BlockSci's target's into your build script. This makes it much easier to build libraries that use BlockSci as a dependency.
The BlockSci python module has been moved into a separate module to allow for a simple SetupTools or pip based install process: pip install -e pyblocksci. The main BlockSci library must be installed first for this to work.
Finally, install instructions for the mac have been added along with Ubuntu 16.06 instructions.
Updated mempool recorder and integrated it into BlockSci interface.
For instructions on running the mempool recorder and using the data it produces, see the setup section.
Improve and clean up auto generated API reference.
All method signatures display correct types and all properties display the type of the returned value. Further, all types link to their definition in the documentation.
hkalodner released this
Mar 15, 2018
Safe incremental updates
Following an number of enhancements BlockSci is now capable of safely performing incremental updates. The AWS distribution of BlockSci now includes a Bitcoin full node and will automatically update the BlockChain once per hour. For local installations of BlockSci, see the readme for setup instructions.
Introduced new concept of Equivalent Addresses which includes two types of equivalences, Type Equivalent and Script Equivalent. Type equivalent refers to two addresses using the same secret in a different way such as how a single pubkey could be used for a Pay to Pubkey Hash address and a Pay To Witness Pubkey Hash address. Script Equivalent refers to a Pay tp Script Hash address being equivalent to the address it contains. Address.equiv() and the EquivAddress class were added to support these concepts. See the documentation for more information.
Enabled the opening of multiple Blockchain objects in the same notebook by removing internal usage of Singleton pattern.
Proper handling of segwit tx and block size distinctions. This included updating the parser to store the size of each transaction excluding segwit data and as as supporting the 3 new notions of size that segwit introduced.
Proper handling of bech32 addresses.
Blockchain.address_from_string() now supports lookup of bech32 addresses.
Address objects now display the correct human readable address depending on the address type.
Improved initial chain parsing from 24 hours down to 12 hours and reduced in parser data size due to unification of the hash index database and parser address hash index database.
Updated to new data version for the parser output requiring a rerun of the blocksci_parser.
In order to allow multiple blockchain objects. All constructors and factory methods were removed with parallel methods added to the chain object. For instance Tx(hash) is now chain.tx_with_hash(hash).
Removed Address.script and merged its functionality into Address
Modified Address.outs(), Address.balance(), and related functions to only return results for places on the Blockchain where that address appeared in a top level context (Not wrapped inside another address).
Renamed various methods from using script in their name to address in order to reflect updated terminology.
Removed ScriptType since its functionality was superseded by EquivAddress
.. _Issue #43: #43 .. _Issue #50: #50 .. _Issue #53: #53
hkalodner released this
Mar 9, 2018
· 5 commits to master since this release
Version 0.4 introduces full bech32 address support, adds segwit size support, and fixes a bug which had been preventing use of continuous incremental blockchain updates.