paint-brush
Rusty Chains: A Basic Blockchain Implementation Written in Pure Rustby@mereep
12,874 reads
12,874 reads

Rusty Chains: A Basic Blockchain Implementation Written in Pure Rust

by mereepAugust 19th, 2020
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

The next project at my work will be fueled by blockchain technology. The best way to overcome the level of just repeating buzzwords and to get an actual level of understanding is to get the hands dirty: Implementing the thing from scratch. We will use the rather new programming language Rust to do so. Rust is very detailed in the implementation part by not only showcasing the code but explaining its ideas and syntax. When following up, you will have a natural understanding of blockchains by creating an extensible blockchain prototype and learn some usable basics in Rust.

Companies Mentioned

Mention Thumbnail
Mention Thumbnail

Coins Mentioned

Mention Thumbnail
Mention Thumbnail
featured image - Rusty Chains: A Basic Blockchain Implementation Written in Pure Rust
mereep HackerNoon profile picture

A hands-on tutorial on blockchain basics, taxonomy and Rust.

What is inside the story?

In this article, I will take you a bit along my journey into blockchain. The next project at my work will be fueled by blockchain technology. Since I had (have) no real idea about that topic apart from some general high level ideas, I started to grind for information on the topic and finally decided that the best way to overcome the level of just repeating buzzwords and to get an actual level of understanding is to get the hands dirty: Implementing the thing from scratch.

Not everything which would be needed to deploy your own public blockchain, but a working part of the base technology which already resembles a fully functional (local) blockchain. We will use the rather new programming language Rust to do so. Since the language does not have a huge user base like Java, C++ or Python, that module will be very detailed in the implementation part by not only showcasing the code but explaining its ideas and syntax.

There is a lot to cover. When following up, you will have a natural understanding of blockchains by creating an extensible blockchain prototype and learn some usable basics in Rust. Who knows what that might be good for in the future? But don’t worry, I will try to guide you step by step through the whole thing instead of just dropping code snippets.

Prerequisites

To understand the article the only thing you will have to bring with you is a basic level of programming experience and the wish to dive into the topic of blockchains. Working knowledge specifically in Rust is not strictly necessary. I will explain the relevant concepts on the go.

However, an installation of Rust[1] and a proper code editor would be a plus (e.g.., CLion or Visual Studio Code) if you want to try the code and follow up with the details is recommended.

Fast Track 🏃

Here are the relevant pieces:

Fundamentals

The first part of the article will guide you through some basics and my personal view of the technology in order to have everyone at the same page.

What is a Blockchain? 

First things first: let us talk a little bit about the blockchain itself (no really: first get yourself a coffee or tea ☕ we have work to do). If we want to talk about blockchains, we first should talk about Distributed Ledgers. You have probably heard this term in the same breath as blockchain.

Maybe you have come around the peer-to-peer filesharing protocol BitTorrent[2]. This protocol enables filesharing in a decentralized manner. Normally, when downloading a file from somewhere over the internet, you will send a request to a specific server which stores the whole file and transmits the data to your device. This is classical client server architecture. BitTorrent however will share a single file across multiple actors / peers at the same time. If you download a file from the Torrent network, you effectively will download it from multiple sources and eventually becoming a source yourself by participating in the network.

Distributed ledgers share some common characteristics. The data (i.e., the ledger) - which generally is not just a simple file but rather a kind of database or really whatever you want it to be - is distributed across multiple actors / peers. There is no single authority in charge for keeping, modifying or distributing that ledger. Rather, all the connected peers keep a copy of the very same ledger.

This idea comes with a lot of problems and questions like: Who can add information to the network? Who can edit it? Who can read it? How much data can be added? Who pays for the storage? Who pays for the processing? Which data is added first? How to recognize failures? How to prevent malicious participants?

How authentication is managed? What about network latency? That’s a lot of complicated tech to cover. All those questions need some form of consensus which all participating peers agree on. Consensus is a rather long topic on its own. There are different types of consensus algorithms suiting different use cases[3] all having their own advantages and disadvantages.

In this article we will implement a sort of a Proof of Work variant, which is used in known public blockchains like Ethereum (might switch to Proof of Stake in future versions) or Bitcoin. We will come back on that in Chapter II of the writeup.

Having that said, what is the connection between distributed ledgers and blockchains? To be honest: Taxonomy wise — it’s not that simple[4]:

The term blockchain is an ill-defined buzzword

If we want to visualize the relationship in a Venn diagram it might look like that:

So, is a blockchain a distributed ledger? Well it might be; sort of. Does it need to be? Not necessarily. My take on it:

The blockchain itself is just a special kind of ledger

It is a special kind of ledger which remembers all states-transitions it ever went through. Imagine having a regular database, where every change / transaction you commit will be recorded such that the final state of the database can always be obtained by executing all the recorded transactions in order. The final state in the context of blockchains is referred as World State.

Since this state does not contain the whole history of transactions, it is typically much smaller than the actual blockchain. However, it is not strictly necessary to store the world state since it always could be retrieved by replaying all transactions. This chain of ordered transactions is secured in a way that it is impossible to change any intermediate transaction (previous chain parts) without breaking the whole thing apart. This is often referred as immutable state.

If you share that blockchain around multiple parties (including the whole world if public) by taking in account the previously mentioned features on consensus, well then you have a distributed ledger where the ledger is that blockchain.

But why is it that people talk about blockchains when they are talking about concrete realizations like Bitcoin or Ethereum? Well, the blockchain itself is a part of all of them. I would rather refer to them as Blockchain Frameworks and Blockchain Ecosystems. 

The common understanding of blockchain actually refers to blockchain frameworks and blockchain ecosystems

Marketing decided that this is the same, but keep in mind: it really isn’t. Blockchain Frameworks and Ecosystems typically offer a lot of technology and software, mostly together with a couple of of additional components offering different sets of features, possibly also with an existing user base / network. There are a lot of blockchain frameworks in the wild, which may or may not be compatible to another.

Next to the already mentioned Ethereum and Bitcoin, there are prominent representatives like Hyperledger Fabric, Hyperledger Besu, Substrate, Quora or Corda just to name a view; all having different features and use cases / audiences[5].

What we need blockchains for? (Use Cases)

This question is not so easy to answer and might also be somewhat opinionated. Blockchain Frameworks can be categorized into at least two different flavors: public blockchains and private blockchains. Those are not fundamentally different, but their components and use cases differ.

Public blockchains are basically accessible by anyone. People can join and participate if they speak the same language, i.e., following the same protocols (having the same consensus). Therefore, there exists a multitude of clients for specific blockchains.

Ethereum has more than 10 active projects listed[6] which are implemented in all sorts of programming languages including in Rust (Parity Ethereum).

The most obvious use case for public blockchains is providing means to supply and store digital value. Bitcoin, the most prominent player in the field, is doing exactly that by providing a digital currency: The Bitcoin (BTC). Due to the limited amount of coins, the immutability of the ledger and the decentralized nature, those values can neither be randomly appearing, disappearing nor can anybody embezzle anything.

They are unique and steady. Changes are not opaque anymore. And here is the important takeaway: there is no need for any participant to trust in anybody. Everyone is watching. It is almost impossible to manipulate the system since it doesn’t depend only on a single actor.

That also means, you cannot DDoS[7] the network. Those attack vectors just disappear. Security comes by design.

Surely enough, currency and coins are not the only thing that can be stored. Contracts and Wills saved forever immutably? Land Registry[8]? Proofing possession of digital goods? Your poem being stored forever? Publication of information that no one ever can destroy?

Blockchain got you covered. With features like the Ethereum Virtual Machine (EVM[9]) it is even possible to deploy logic (i.e., code that does state transitions) onto the chain. There possibilities and challenges seem promising for the future.

What might smart and creative people come up with using blockchain tech in the future?

Private Blockchains are basically not different from public blockchains. They use the same technology but come with additional features, which are not common in public chains. Private blockchains are not meant to be joined by anyone but a regulated audience. Those could be different organizations which are sharing some joint workflows, possessions or values.

To enable a collaboration like that, frameworks also needs features like authentication, authorization, permissioning (user roles), rights management, deployment, networking, monitoring and so forth. But why should anyone want to deal with such complicated technology just to share a chain of transactions and data between some limited companies?

Well, the audience seems way smaller. Remember the shared state which all the participants store and the consensus which the participants agree on? There is no central authority which you just have to trust. No one can just manipulate data. There are no two valid realities existing at a time. Also, since every party stores at least one copy of the ledger, there is no risk on loosing access to the data. If you share some interest with someone else but you cannot absolutely trust on the integrity of the company, private blockchains may have you covered.

I hope that introduction has helped you a bit on gaining a feeling of the technology itself and got us on the same page. If you feel like something is missing or are disagreeing on some statements, please do not hesitate to drop me a message. I really appreciate discussions.

Implementing The Blockchain - PART I

This section will start on implementing the blockchain itself. We will define the project layout including the basic data structures which constitute the blockchain. Those will serve as a skeleton for all the implementations that are necessary later. Please note that I am by no means a Rust (or blockchain) expert. In fact, this is my first Rust project.

Given that I am not a C(++) / Rust developer, the code may not be idiomatic at all. Please do not use the code I provide as a blueprint for anything you do which might be security relevant in any form. You have been warned. However, I hope it will make the underlying concepts and building blocks clear. Again, any feedback you drop me in will be gladly used to improve the code.

The structure of the project

We will go with a very simple folder structure, which basically consists of only three files. A main file (main.rs) which will contain the code that will invoke the blockchain and does some sample operations to demonstrate the functionality. Another file which will be inside a subdirectory will contain our actual blockchain data structures and implementations.

The last file will be the cargo.toml file which will hold the external dependencies of the project together with some project meta data. The structure will be as follows:

/rchain
 |-- src
 |---- rchain
 |------ mod.rs
 |-- main.rs
 |-- cargo.toml

The cargo.toml comes with only one dependency at the moment which will be a hashing-library (Blake2 [10]) and looks like that:

[package]
name = "rchain_v1"
version = "0.1.0"
authors = ["Richard Vogel <[email protected]>"]
edition = "2018"
 
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
[dependencies]
blake2 = "*"

The first time you run cargo build within the directory to build the project and create executables, cargo will download and compile the dependencies for your architecture. That is a great step from complicated dependency management in C(++): it is similar to what pip for Python is doing. At the time of writing there are over 44,000 so called crates in the index[11].

This is not the same as 255,683 releases in the Python package index but should contain something for most usage purposes. You can also call C-libraries from Rust.

The

rchain
directory within the project is a module. This is Rusts way to organize your source code into multiple files and directories. Again, this is conceptionally similar to Python’s modules (without you having to include a
__init__.py
 
file).

You can happily put modules into modules, creating some means of hierarchy. Contrary to Python, Rust supports access modifiers, which can be used to allow and prevent other modules or submodules from accessing specific members.

The Blockchain module

Let’s go into the details of the implementation itself. We start with a UML class diagram to illustrate how everything falls together within the module (although Rust doesn’t have explicit classes). Refer to this graphic to this place when you feel you got lost in all the details.

Some details are a bit simplified or omitted in order to keep the diagram reasonable. We will walk through each of the components step by step. The code itself will be heavily commented and documented. Please read the code snippets which I provide here (also read the code in the repository to get the full picture), even if you do not understand every small detail of the syntax.

Rust can be a bit hard on newcomers, especially if you come from a high-level language like JavaScript or Python. I will explain some of the more peculiar Rusty syntax within the article. If you have little to no coding experience and just want to know how blockchains work: still read the code comments!

The first part will focus on the data structures itself (i.e., what purpose they serve and what data they contain). The second part will focus on some of the more important implementations (i.e., functions and methods). The third part will finally demonstrate a simple usage of the

Blockchain
-Module.

Data structures

This section of the implementation will explain all data structures and all their attributes / members in detail. Rust lets us split the declaration of the data and its implementation into two different parts. If you have ever written some C or C++ code, that might be similar to header (.h/.hpp-files) and (.c / .cpp) files which separate declaration and implementation.

However, in Rust only the data members are declared in this declarative part (except for traits, which are somewhat special to Rust; we will see an example of them here). This section will also provide many concepts of blockchain-relevant concepts.

Again, even if you do not understand all the code details, read the text which will make you understand the meaning and usage of each building block.

Blockchain

The definition of the

Blockchain
struct is as follows:

Rust does not have a class keyword. This doesn’t mean rust is not supporting class-like project structures. We will use structs instead, which feature a lot of what classes in other languages can do

Obviously, the struct stores a list of Blocks which we will define in detail later. Also, there is a Map / Dictionary of Accounts stored, which keeps track of the current final state of the blockchain. The mapping of that map is User Id to Account. For now, the User Id is arbitrary, in the next article those keys will be used to represent the public key of the account.

Our blockchains world state is completely represented by this Map. Finally. pending_transactions will be used for bookkeeping of ready but not yet executed transactions. We will not make use of this for now. This is a preparation for the next article.

The first thing you may notice is that I used “

///
” for comments here instead of the typical “
//
”. Those are doc comments. Rust comes shipped with its own documentation builder, which generates a nice web interface documenting whole project and all of its dependencies.

Everything that is marked with

///
will be included as a description within the web interface. You can invoke the builder by typing cargo doc within the main folder. You can find a version of the current documentation on here.

You may also wonder for the line

#[derive(Debug, Clone)]
. This is the Rust syntax for injecting functionality / behaviors into a struct. Technically, this provides default-implementations of traits, which the compiler provides for free (the next part will go into traits).

Here we use Debug and Clone. Debug is a helper for pretty-printing a struct using the print!-command, which otherwise would not work as easily; Clone is used to be able to create duplicates of objects.

Note the

pub
keyword in front of some of the struct’s members. If you do not provide that keyword, those would not be visible / accessible from outside of the struct itself (i.e., default private).

WorldState

As mentioned earlier, the blockchain is just a special type of ledger that stores information in a special way. As seen in the

Blockchain
structure above, the blocks and the resulting accounts are stored in two different members. If we would not care for the history of the transactions but just for the result, we could remove the blocks altogether.

Actually, we could just instead write transactions to a file or a database or really wherever you want them to be stored in whatever way you want them to be structured. Maybe take a break here and think a bit why this is true. Although it will introduce some additional difficulty to the article, I wanted to attribute that fact by modelling it into the code.

Also, this demonstrates an important part of rusts syntax: the traits, which looks like that:

Traits basically define a behavior or interface which can be shared across multiple objects. This particular interface really just defines some functionality which an object should provide in order to be a WorldState. Our blockchain will implement that trait later.

A database backend could do also, or a JSON backend or something else. This functionality is not at all specific to block chains.

You may notice some rather unusual notation when not having seen the rust language. The first thing might be the

 -> ~
-notation. This is just the Rust way to define a return type. If you’re a Python developer those look just the same as type hints, except that they’re not optional and your program will not compile if you screw them. The next thing to know is that 

Rusts types cannot be set to null

(or None if you come from Python; do not confuse that with Rusts Option::None, which is not exactly the same) and does not support Exceptions. Therefore, you see return values like Optionand Result. The Option indicates that the function might return valid data (e.g., Option<&Account>indicates that the function might return a reference (&) to some Account-instance or not).

Similarly, for indicating a function which may fail or succeed (where you might use Exceptions in other languages) you can use the Error-type. That’s a bit more involved in this example as this is written as

-> Result<(), &'static str>
.

Consider the two cases: a function might succeed or fail (possibly for multiple reasons, just like there are different types of Exceptions). For this reason, you will have to provide a data type for each case: the success case and the fail case. For that function we provided

()
for the success- and
&’static str
for the fail case. Both need explanation.

The

()
just indicates that if we succeed, we really don’t want to return any additional information. It is just a type carrying no information at all (somewhat like
void
in C(++) or more generally a unit-type[12]). The
&’static str
type is a bit more special.

The

static
(
-notation) really is something unique to Rust. This is a so-called lifetime specifier and is one of the reasons Rust feels like being a managed language (like Java, JavaScript or Python) while not needing any Garbage Collector[13] that tidies up the mess we leave in the memory.

And as you can guess, things get a bit more involved there. We will not dig into this. Whether can this article provide enough insights (it would fill whole other articles) nor do I feel Rust-mature enough to explain that.

For now: this specific lifetime specifier just means that we want to return a pointer to a string which is baked into the binary (thus the value is static indicating the lifetime of the whole program, i.e., that value never dies). The last thing to note here is that

Result
and
Option
are technically nothing but common Rust Enumerations.

We will see soon that Rust Enumerations can carry variable data within their variants, which also enables this exact use case. Due to the frequent usage, those Enums are globally available in every module per default.

If you’ve read the previous chapters, you will recognize most of the syntax here, except the unique pub(crate). This is a special access modifier which allows all methods within that project access the variable, while third parties could not. We will make use of that at the usage example. Let’s focus on the members.

Typically, a block does not include only one action to be performed, therefore transactions member stores a vector of all the actions it includes. The hash will store a normalized representation of the whole block (including all transactions) inside.

For generating such a hash, there is a hash function needed. You might have heard of MD5 or SHA-hashes. Those get some arbitrary long input data and produce an (almost) unique and handy output while preserving some cryptographic properties[14].

Since those hashes are the ingredient which makes a blockchain being able to glue together, I want to graphically illustrate what a hash-function actually is doing. Note that even the smallest change in input will produce a significantly different output. This is one of the properties a good hash function should provide.

We will make use of that property when implementing adding the Proof of Work in the next article.

We will generate and store that very hash of the block within each transaction. If someone would change just anything in the block, those hashes would not match anymore, which easily can be checked by just hashing again. And here comes the glue of the blockchain: Blocks (except the first one) will also store the hash of their previous block (here:

prev_hash
).

So, if someone tampers with the blockchain, that person would not only invalidate that very same block but also would invalidate all pointers to previous hashes in the upcoming blocks. The whole blockchain will fall apart. 

A concrete blockchain with two connected blocks could look like that:

Normally, adding blocks to a distributed blockchain is computationally intensive and / or needs authorization. Again, we will see why when adding the Proof of Work in part II to our blockchain. For now, let’s just say, that this is practically infeasible in real scenarios unless quantum computers might change that.

The last variable, the nonce (number used once) will be (miss-)used later too for the proof of work. We come back to the

nonce
again, when talking about
Transaction
s below.

This finishes the description of the block itself. Please make sure you understand the topic of hashes, and how they are used to tie the blocks together, since it is the integral component of the blockchains’ tamper-resistance.

Transaction

The Transaction structure represents some request to the blockchain which on execution will lead to a change of the WorldState (a state transition) and looks as follows:

Every transaction must be linked to an existing user at a specific time, whom will be stored in the

from
and the
created_at
field, respectively. Those are probably the easiest to understand fields.

The

record
holds the actual information on the change that we want to commit. It represents a command together with its accompanied data (
TransactionData
), which we as the developer allow to be transmitted to the blockchain. We will watch them later.

The

signature
will hold your digital sign of the message. This field will make sure that no one except the person which is stored in the
from
field can be the one who sent the transaction, i.e., no one will be able to submit transactions on behalf of someone else. Therefore, it represents the base layer for authentication. This will be achieved using public and private keys which we will create directly from code. The second part of the implementation will dive into the details.

We already have seen the

nonce
in the block structure. A nonce represents a number, which will be used exactly once for a specific purpose. This is what its acronym number used once is representing. The system should make sure, that no two transactions having the same nonce should be accepted in the system.

Although it is not completely implemented, I marked the code position where you would have to add some code in order to enforce that rule. But why should anybody care for this at all? Using a nonce will help preventing at least three problems: networks problems, user errors and malicious intends. All due to one single number.

The basic thought is as follows: when not accepting a nonce twice, a request which will be transmitted more than once to the system will be rejected. This fixes (i) faulty network stacks, which for some reason transmit a network package twice, (ii) users or faulty user interfaces which for whatever reason accidentally transmit the same message multiple times (you don’t want to send the very same payment accidentally twice), and (iii) some attacker might capture the request as is and — without any change — may submit it again.

This attack can also make you spend your hard-earned coins multiple times, too. The attack (which is not unique to distributed blockchains) is known as replay attack for obvious reasons. If we would not have a nonce, even a signed message cannot prevent replay attacks.

TransactionData

The

TransactionData
structure is part of the transaction itself. Here we define, which actions our blockchain should be able to process. If you want to extend the prototype and add features to your personal blockchain, this data structure is the place to start. The current definition looks like that:

You may have noticed that this time, we are not defining a

struct
but rather an enumeration (
enum
). Basically, an enumeration represents a data type that only holds a defined set of values (variants) and only exactly one at a time. Most programming languages support enumerations. This specific example defines the four variants
CreateUserAccount
,
ChangeStoreValue TransferTokens
,
CreateTokens
.

Somewhat specific to Rust is that every variant may hold additional variable data. The

CreateUserAccount
holds exactly one unnamed value of type String which represents the account id of the user account which should be created.

Those arguments can also be named for better clarity (as in

TransferTokens {to: String, amount: u128}
). The upcoming implementations will provide an example on how to use and destructure those values.

Account and AccountType

Finally, we will focus on the definition of the Accounts which represents the result of the execution of all blocks within the blockchain (i.e., the world state):

An account of our blockchain hold

tokens
which represent the accounts credit (i.e., this carries the value like ETH or BTC). I did not decide for name of the coins, so if you feel like, just refactor that member to your liking. The meaning of the other two members may not be obvious. Let’s start with the
store
. Again, that member holds a dictionary of values. As you may have noticed, there was a variant
ChangeStoreValue
defined for the
TransactionData
.

This transaction is meant to write arbitrary values into the user account and store it into the store. We could use it for example to store different kind of digital values (e.g., purchases of digital music) or physical values (e.g., owned properties) or just any other information we like into the account.

The remaining member

acc_type
is intended to represent the type of account and is modelled as enumeration just like
TransactionData
:

Different accounts can be used as the base building block for realizing your own realization of permissioning. Maybe you want only specific accounts to be able to submit specific requests? Or only some accounts have permissions to add or verify blocks, create tokens or hold business logic (like smart contracts or chain code)? Here is the starting point for those implementations.

This finishes the declarations of the data structures. If you made it until here: you have covered almost anything which is needed to understand the nuts and bolts of this blockchain, well done 😊. The rest will be needy greedy details. But that is what you came for, right?

Structure Implementations

While the last section was very detailed and covered almost every line of code, this section will only focus on the important parts of the implementations. However, if you don’t come here to learn some Rust: most of the details here are actually treated already informatively in the sections before. You may just skim read that section.

Account 👤

We start with implementation of the Account structure. I chose that structure first because it really contains nothing interesting and no functionality at all so we can focus on Rust. So here it is:

Structure implementations start with the

impl
keyword followed by the structure we want to implement. The only function we want to add is the
new
function, which acts as a constructor for the Account (a function which is used to construct the object in a defined way). Rust has no strict definition for constructor functions (like
__init__()
in Python or a function that’s named like the class itself in Java and C++).

In fact, naming the function

new
is completely arbitrary. But since constructors are so common in object-oriented thinking, defining a function named
new
is the de-facto standard in Rust for implementing them. The action that function performs is the construction of the Account struct with some default values set.

You may have noticed the return type

Self
(notice the uppercase S) which is used to construct the structure. This is a special type that just represents the currents structure’s type (NOT its instance, which is written with lower case s). If you have Python background: that’s conceptionally similar to the
cls
(the first) parameter in a
@classmethod
). There is not much interesting going on except for the fact, that the
store
is initialized with an empty
HashMap
by calling the
new
implementation of the
HashMap
struct which is an example for that de-facto standard within the Rust standard library.

The rest of the chapter will not list constructors anymore. You can expect them to be available where reasonable, we will focus on the relevant code.

Append a block to the blockchain

The functionality to add a block to the blockchain is a core ingredient. Here you will have to enforce all policies a block has to comply with in order to be accepted. Due to the function is very lengthy and doing multiple things the code will be snippet and explained in pieces.

This snippet defines the function

append_block
. This function should add the Block and return a
Result
indicating the success or failure of the operation (in joint with an appropriate error text). The parameter
&mut self
indicates a reference to the instance itself (instance function) while also announcing that we are going to change the structures state (we possibly add a block) using the
mut
modifier.

This is similar to the (implicit)

this
pointer in C++, Java or JavaScript or to the
self
parameter in Python. Now the only thing we do is creating an indicator variable
genesis
which will be true if the length of the blockchain is zero (i.e., we are adding the first block which is denoted as genesis block).

Note that even if Rust is a strongly typed language you don’t always have to provide the data type explicitly. In this case the compiler infers that we are defining a Boolean variable (bool). In modern C++ and C# you would use the

auto-
keyword or the
var
-keyword in recent Java versions.

where

verify_own_hash()
is defined as a function on Block as:

This code snippet checks first if the hash is set and if it is set it will verify if its valid by comparing it to the correct hash. This is done by calling 

.is_some()
on the block's hash member. Remember that it is declared as
hash: Option<String>
, a variable which may or may not hold data. The indicator 
.is_some()
will indicate if the variable does actually contain data.

Knowing this fact, we can safely use 

.unwrap()
to get the value. This call would panic (exiting the program with error) if no value was set. Since we do not want to take ownership of the variable inside, we use 
.as_ref()
to tell Rust that we just want its reference. The call to 
.eq
on that String will compare it to the calculated hash ( 
.calculate_hash()
) defined for the Block. We will inspect that function later.

If

verify_own_hash()
evaluates to
false
we will create the
Err
holding the appropriate error message (previous snippet). Notice the 
.into()
call after the error message? This is the way for implicitly convert between compatible types. When creating a string using
“Some String”
we actually create a
&str
type (i.e., a pointer to fixed length string somewhere in memory) but the result wants to store a
String
(i.e., an owned possibly mutable String, which is not exactly the same), so we need to convert those.This is what
.into()
is doing.

Next, we check if the block is meant to be appended to the current last block. We do this by checking if prev_hash stores the actual the hash of the last block in the blockchain:

Even though both

prev_hash
and
hash
are of type
Option<String>
we can actually compare them via
==
since Rust provides a standard implementation which evaluates to
true
if both
Option
s contain no value or both contain a value and both values are the same (that is done by implementing the appropriate equals trait).

This is exactly what we want, since it deals also with the genesis block correctly which will not have a previous hash (yes, since it is the first).

Now, we check if the block contains transactions. This is just a policy we enforce, you might remove it if you want to support empty blocks.

There is nothing new to that code. Let’s move on to something more involved: verifying each transaction in the block. We want the transactions to be authentic (not tampered with) and we want them to be valid, i.e., the execution of the transaction must be possible. We do not want a transaction to spent tokens, which the account does not supply. Because a block contains multiple transactions, we will need some kind of rollback-functionality.

Otherwise, if a transaction is executed correctly and later transaction is not (forcing us to reject the block) we might end up with an inconsistence World State. That’s something we do not want. Since I don’t want to implement some complicated rollback algorithm, we just clone the whole state even though it really is not the best idea, since this could eat very much memory if the state is big:

Remember the

#[derive(Clone)]
on top of each struct? Cloning an object actually is and rather complicated operations since it has to clone all members and all structs which might contain structs which, … you get the point.

Rust lets us get away with a standard implementation by deriving from the clone trait for many standard data types.

Now we try to execute each operation on the chain to verify their validity:

Firstly, we iterate over all the blocks' transactions this is done by a

for
-loop that iterates over (
index, transaction)
-pairs that are emitted by
block.transactions.iter().enumerate()
. That’s equivalent to
enumerate()
in Python.

Basically, 

.iter()
creates an enumerable that emits a reference to each transaction of the list of transaction while 
.enumerate()
consumes that list and emits a pair that not only contains the transaction but also its position in the list.

Inside the loop we try to execute each transaction by calling

transaction.execute()
, which we will inspect later. The
execute
function itself indicates its success or failure by returning a
-> Result
. The statement
if let Err(err) 
is a pattern matching syntax which checks if the transaction returns an Error variant (
Err
) and if that’s the case it assigns the Errors data (the error String) to the variable
err
before executing the block within the curly brackets.

If the transaction executes correctly (i.e., if it returns the Results’ an

Ok
variant) it just goes on with the loop instead. If it fails to execute, we write the old world state (
old_state
) and return an Error ourselves. The
format!
syntax builds our actual String using placeholders.

Again, this is comparable to Pythons’ string formatting

(

"{} World"
.format("Hello")
).

Well, if everything went fine, we just append the block and indicate everything went fine by returning the

Ok
-variant:

In Part II of this series, we will come back to this function when adding Proof of Work and authentication (signatures).

Hashing 🝖

In the last function we already used hashes and called the Block’s calculate_hash() function. If you do not know what those hashes actually are, please refer back to the Block data structure description.

So that’s how we squash a complete block into a single hash value:

This is the place where we make use of the external crate Blake2. After creating an instance of Blake2 (

hasher
) we loop over all transactions inside the current block instance (
&self
). For each transaction we call the hasher’s 
.update()
function. Each update will feed some new data into the hasher which are some arbitrary bytes (datatype
u8
), which in this code is actually a hash itself: the hash of the transaction.

Yes: we actually hash the hashes of the transactions. Since the block itself also contains some data which is unique to the block, we will have to take those into account also when creating the whole block’s hash. We do this by just formatting the relevant information (previous hash and nonce) to a string and feeding it to the update function. This is somewhat lazy.

We could feed both values separately (which would make the representation more stable) but since those values are Option types we would have to account for them being values or none. This blows up the code unnecessarily.

To tell the hasher we are done adding data and start the hashing, we use its

finalize()
function, which will output the hash as an array of bytes (
[u8]
). We transform that array to a Vector of bytes and return it.

The transactions'

calculate_hash()
function looks very similar to this function, that’s why skip its details here.

Verify the blockchain 🗹

To check if the blockchain that is presented to you is valid (i.e., no one tampered with), we need to basically do the same checks (making sure the blocks are not tampered and correctly connected) as when adding the block to the blockchain, while just omitting some different messages. Refer to the Block’s function

check_validity()
for details.

Executing a transaction

Normally, the execution of a transaction comes with some change in the world state. Specifically, for this implementation this actually is a change in the

accounts
dictionary of the
Blockchain
struct. Remember, when we dealt with the world state we modeled those specifics away and said our blockchain should implement a specific interface (the
trait
WorldState
) instead, which just describes the functionality the backend (in that case the blockchain struct) should support.

Our

Blockchain
struct implements that interface like that:

Here we define the implementation of trait

WorldState
for the
Blockchain
. Specifically, that snippet shows the implementation of the function
get_user_ids()
, which just returns all keys of our accounts directory.

Because we did not come around that kind of syntax for now, let me explain this expression. The 

.keys()
function returns all indices of the
accounts
(which represent the unique ids of the accounts). We could use the keys as iterable in a f
or id in …
loop.

But instead we make use of functional programming paradigms here. If you coded in R or Python or JavaScript you probably came around the

map()
function. This is basically just a function which takes every element in an iterable (here: the account ids) and feeds them into another function, which can transform them. This idiom is technically referred to as function composition[15].

Here, the inline function

|s| s.clone()
really just takes each value into
s
(which is actually of type
&String
) and clones that value into a new position in memory. The final 
.collect()
just turns the resulting iterable into the return type
Vec<String>
.

Now having that settled, we can now come to the actual implementation of the

execute()
-function in the transaction which starts like that:

This looks a bit weird if you’re not used to template parameters. The interesting part might be the

<T: WorldState>
syntax right after the functions’ name
execute
. You can read the whole line like that: declare a public function named
execute
which takes a parameter world_state
 
whose type conforms to the
trait
WorldState
.

You basically declare a new type and set the constraint that this very type should implement the

WorldState
. You do not fix it to any specific concrete type (we could easily just remove the
T
and replace it with the declaration
world_state: &mut Blockchain
), which is arguably easier to read. However, this implementation allows us to use any transaction we define on any backend that implements the WorldState trait.

This backend could easily be a PostreSQL or MySQL handler. The actions we perform are for the most parts not specific to blockchains.

Having this covered, lets take a look in the implementation of the

execute
function which goes on like that:

Before each execution we check if the user who wants to execute it really exists. Here we use the

trait
’s function
get_account_by_id()
which is defined as
fn get_account_by_id(&self, id: &String) -> Option<&Account>
(i.e., as a function which might return an
account
or not).

Again, we destructure that

Option
by using the i
f let
pattern matching syntax which will execute the block if the function returns “some account”. If it doesn’t, we drop an error message.

After the pre-checks follows the implementation of the transaction’s specific logic pieces:

The statement match

&self.record
is just another way of pattern matching. Recall that the record is of type
TransactionData
which is an enumeration holding the action to execute and optionally additional data needed for execution. Here we test for each existing variant of the enumeration
TransactionData
and execute the appropriate actions on the
WorldState
if possible.

If we fail to do so for any reason, we will return an

Err
which would force the blockchain implementation to rollback the state. If you are interested in the concrete realizations of the omitted commands, I encourage you to watch the source 😊.

Those are all the implementations for the first part. If you made it until here and where no Rust developer, I am aware this was a lot to cover. Do not worry if some syntax still feels alien.

Some of it does for me still. However, the great error messages spilled out by the compiler are often quite helpful in spotting the problems. If you write code, its sometimes even fun to fix them and make the compiler happy (I fixed like a hundred of them while writing this little piece of code…).

If you compile the code and see some warning indicating unused variables of code and not constructed variants: you can safely ignore them. Most of them are just some preparations for Part II of this writeup or examples.

Usage example

This example will guide you through

  1. Initializing a valid blockchain adding some transactions (adding users, transferring tokens)
  2. Tampering with a transaction and detecting it
  3. Smarter tampering with a transaction, still detecting it

The contents of the following file are found in the main.rs file. If you read the implementations above, this code will be fairly simple to understand.

We start by declaring our module (

rchain
), import the dependencies, define the main function and create our blockchain object:

Like in C(++) or Java, a Rust program have a standard entry point declared by a function named

main()
, which we defined in that snippet.

Now we create a genesis block and define the users we want to create:

After doing so we loop over the

initial_users
vector in order to create transactions that create those users in the blockchain and assign them 100,000,000 tokens each:

The last lines add all the transactions to the genesis block. The value 0 in the transaction’s constructor should be the nonce. However, since we do not enforce this nonce at the blockchain right now we just fill it with a dummy zero.

It will be a good starting point for your own coding adventures to implement the policy to enforce of a correct nonce.

The created and filled block is now added to the blockchain. The result of the append function will be printed after which the whole blockchain also will be printed as a string representation:

The

{:#?}
formatter is used for pretty-printing. The output of the object will look almost like a valid JSON-object with spacings nicely laid out, whereas
{:?}
will not print out those spacings.

You should see the line Genesis block successfully added: Ok(()) in your terminal, after which the blockchain output starts. Notice the printout of the “accounts” which should contain alice and bob both having 100,000,000 tokens.

Now we add another block containing an action to transfer one token from alice to bob. The blockchain will again be printed. You should notice the changes in the balance of the accounts.

The last line should report a valid blockchain as Blockchain valid: Ok(()).

Now that everything seems fine, the first attack will start. Bob was not satisfied with this one coin Alice transferred. So, he decides to directly change the transaction in the blockchain.

Since we do not support signed messages for now, this can be easily done:

We first clone our working blockchain to have a new one to operate on which has the same state. Then we get ourselves a mutable reference to the second transaction in the first block using

bc_attack_1.blocks[1].transactions[0].borrow_mut()
. Now we destructure the record which is stored in the transaction (remember: this is a
TransactionData
struct) using the match-syntax.

This is the

TransferTokens
action we wanted to manipulate. Since the values should be changed in-place (directly in the memory where they are stored), we have to account for that fact by using the appropriate patterns for the match construct (
&mut
,
 ref mut amount
).

The binding syntax

{to: _, ...}
just indicates that we do not care for that
to
variable and silence a compiler warning indicating the declaration without usage.

Now we set the value at the memory position where amount refers to, to 100, which is Bob’s desired amount of coins, using

*amount=100
. After doing so we check the blockchain again:

Which should print: Is the Blockchain still valid? Err(“Stored hash for Block #2 does not match calculated hash (Code: 665234234)”,)

→ The blockchain is invalid.

While Bob failed, Alice also wants to try her luck. Since she didn’t want to steal coins from Bob, she changed the

CreateTokens
action in the genesis block to create her account having 100,000,000,000 tokens to start with.

Also, she noticed Bob’s mistake and didn’t want to run into the same trap, so she decided to also update the manipulated blocks hash. The basic attack looks conceptually the same as Bob’s attack.

However, she added the following line after changing the transaction:

will print something like (your output will slightly differ):

Is the Blockchain still valid? Err("Block #1 is not connected to previous block (Hashes do not match. Should be `ÜQ\u{14}²Jüø9É7x;\u{86}æÇÐ\u{8d}À¤þMy<\u{19}óí\u{5}Qoó;ÞNc\u{12}u×1qÕ@Öáø\u{87}áHÿÞ²Àpª£\u{4}·\u{1e}íµ«Þû±¬` but is ``óç\u{93},\nbÿÁ=\u{9b}º\u{98}&¼-8$Ö~ü}òB)F°yXÙ#ãL\u{84}dÜcû/\u{9a}½I\u{16}É\u{98}¥çC;ðJè\u{9c}ý\u{9f}\u{9a}/¼\u{1}þ<òÕû`)",)

 → The blockchain broke again 🎉🥳

Conclusions and outlook

This writeup covered a lot of blockchain basics and went down the rabbit hole by implementing the blockchain ledger as a simple prototype using the rather new programming language Rust. The usage of the implementation was demonstrated showcasing two possible attacks to the blockchain, which could successfully be noticed.

The current state of the implementation covers a blockchain prototype, which supports multiple types of transactions which can easily be extended. Also, it supports a simple rollback functionality in case transactions cannot be executed for any reason.

The upcoming Part II will add the following functionality:

  • Sign transactions using public/private keys to make them tamper proof
  • Add a transaction queue to the blockchain which holds requested transactions that are not yet executed (candidates for blocks)
  • Add Proof of Work to secure the blockchain against tampering and uncontrollable growth
  • Some more project structuring
  • Unit testing in Rust

I hope you had a nice time reading through this rather big document. Feel free to go on in implementing some of the missing functionality and add your own ideas. There are some

@TODO
markers which are a good starting point.

I am happy to hear any suggestions for improvements. If there is something that can be done better or easier, if there is something you’d like to add, if you see mistakes or if you need some more explanation somewhere: feel free to write me any suggestions. Just drop me a message. I hope to see you in Part II.

References 📚