Compare commits

...

3 Commits

6 changed files with 400 additions and 113 deletions

1
.gitignore vendored
View File

@ -1,3 +1,4 @@
result/
result
target/
*.avanterules

View File

@ -21,7 +21,7 @@ there is one, and a list of any dependent resources.
{
"objectId": <merkle-hash>,
"content_address": <content-hash>,
"path": "/path/name0",
"name": "/path/name0",
"dependents": [
{
"path": "path/name1",

View File

@ -17,7 +17,7 @@ async fn ref_path(refs: Arc<HashMap<String, Arc<Reference>>>, Path(path): Path<S
}
}
async fn object_path(objects: Arc<HashMap<String, String>>, Path(addr): Path<String>) -> String {
async fn object_path(objects: Arc<HashMap<String, Vec<u8>>>, Path(addr): Path<String>) -> Vec<u8> {
dbg!(&addr);
match objects.get(&addr) {
Some(o) => o.clone(),

View File

@ -21,7 +21,7 @@ async fn get_client_js() -> impl IntoResponse {
#[derive(Debug, Serialize, Deserialize)]
enum ServerMsg {
Reference(Reference),
Object(String),
Object(Vec<u8>),
}
#[derive(Debug, Serialize, Deserialize)]

View File

@ -4,88 +4,158 @@ use blake2::{Blake2b512, Digest};
use rand::Rng;
use serde::{Deserialize, Serialize};
/// A `Reference` represents a node in a content-addressable graph structure.
///
/// Each reference contains:
/// - An id that uniquely identifies the referenced object
/// - A content address hash that represents the content
/// - A name that provides a human-readable name for the reference
/// - A list of dependent references that this reference depends on
///
/// References form a directed acyclic graph where each node can have multiple dependents.
/// When the content of a reference changes, the object id of all parent references
/// are automatically recalculated to maintain content addressability throughout the graph.
#[derive(Serialize, Deserialize, Debug, Clone)]
pub struct Reference {
pub object_id: String,
pub content_address: String,
pub path: String,
pub id: String,
pub content_address: Option<String>,
pub name: String,
#[serde(skip_serializing_if = "Vec::is_empty")]
pub dependents: Vec<Arc<Reference>>,
}
impl Reference {
pub fn new(object_id: String, content_address: String, path: String) -> Self {
/// Creates a new Reference with the specified object ID, content address, and path.
///
/// # Parameters
/// * `content_address` - The content-based hash address of the referenced object
/// * `name` - The name of the reference
///
/// # Returns
/// A new Reference instance with no dependents
pub fn new(content_address: Option<String>, name: String) -> Self {
// Calculate the reference_id from the content_address and path
let hasher = Self::initial_hash(&content_address, &name);
let calculated_id = format!("{:x}", hasher.finalize());
Self {
object_id,
id: calculated_id,
content_address,
path,
name,
dependents: Vec::new(),
}
}
pub fn add_dep(mut self, dep: Arc<Reference>) -> Self {
self.dependents.push(dep);
self
/// Adds a dependent reference to this reference.
///
/// This establishes a parent-child relationship where this reference depends on
/// the provided reference. When a dependent reference changes, the parent's
/// id will be recalculated.
///
/// # Parameters
/// * `dep` - An Arc-wrapped Reference to add as a dependent
///
/// # Returns
/// The modified Reference with the new dependent added
pub fn add_dep(&self, dep: Arc<Reference>) -> Self {
let mut cloned = self.clone();
cloned.dependents.push(dep);
// Recalculate the ID based on dependents, content_address, and path
let mut hasher = Self::initial_hash(&self.content_address, &self.name);
for dependent in &self.dependents {
hasher.update(&dependent.id);
}
cloned.id = format!("{:x}", hasher.finalize());
cloned
}
/// Converts this Reference into an Arc-wrapped Reference for shared ownership.
///
/// This is useful when a Reference needs to be shared across multiple owners
/// in the graph structure.
///
/// # Returns
/// An Arc-wrapped version of this Reference
pub fn to_arc(self) -> Arc<Self> {
Arc::new(self)
}
/// Determines if this Reference is a leaf node (has no dependents).
///
/// Leaf nodes directly represent content, while non-leaf nodes derive their
/// content address from their dependents.
///
/// # Returns
/// `true` if this Reference has no dependents, `false` otherwise
pub fn is_leaf(&self) -> bool {
return self.dependents.is_empty();
}
/// Calculates a content address hash based on dependent references
pub fn calculate_content_address(&self) -> String {
if self.is_leaf() {
return self.content_address.clone();
}
fn initial_hash(content_address: &Option<String>, path: &String) -> Blake2b512 {
let mut hasher = Blake2b512::new();
for dep in &self.dependents {
hasher.update(&dep.content_address);
if let Some(content_address) = content_address {
hasher.update(content_address);
}
format!("{:x}", hasher.finalize())
hasher.update(path);
hasher
}
}
pub struct Graph {
pub root: Arc<Reference>,
pub refs: Arc<HashMap<String, Arc<Reference>>>,
pub objects: Arc<HashMap<String, String>>,
pub objects: Arc<HashMap<String, Vec<u8>>>,
}
pub fn random_object() -> (String, Vec<u8>) {
let mut rng = rand::rng();
let random_size = rng.random_range(50..=4096);
let random_bytes: Vec<u8> = (0..random_size)
.map(|_| rng.random::<u8>())
.collect();
let mut hasher = Blake2b512::new();
hasher.update(&random_bytes);
let hash = format!("{:x}", hasher.finalize());
(hash, random_bytes)
}
impl Graph {
pub fn new(root: Arc<Reference>) -> Self {
let mut refs = HashMap::new();
refs.insert(root.name.clone(), root.clone());
let refs = Arc::new(refs);
let objects = Arc::new(HashMap::new());
Self {
root,
refs,
objects,
}
}
/// Gets a reference by its path
pub fn get_reference(&self, path: &str) -> Option<Arc<Reference>> {
self.refs.get(path).cloned()
}
/// Gets an object by its content address
pub fn get_object(&self, content_address: &str) -> Option<&String> {
pub fn get_object(&self, content_address: &str) -> Option<&Vec<u8>> {
self.objects.get(content_address)
}
}
pub fn random_object() -> (String, String) {
let mut rng = rand::rng();
let random_size = rng.random_range(50..=4096);
let random_string: String = (0..random_size)
.map(|_| rng.sample(rand::distr::Alphanumeric) as char)
.collect();
let mut hasher = Blake2b512::new();
hasher.update(&random_string);
let hash = format!("{:x}", hasher.finalize());
(hash, random_string)
}
impl Graph {
/// Updates a reference to point to a new object, recalculating content addresses
/// for all affected references in the graph
pub fn update_reference(&mut self, path: &String, new_object_id: String, new_content: String) -> Result<(), String> {
/// Updates a reference to point to a new object, recalculating content addresses and IDs
/// for all affected references in the graph.
///
/// The reference ID is calculated from the content address, name, and any dependents,
/// ensuring that it's truly content-addressable.
pub fn update_object_reference(&mut self, name: &String, new_content: Vec<u8>) -> Result<(), String> {
// Update the root reference if needed
if name == &self.root.name {
self.root = self.refs.get(name).unwrap().clone();
return Ok(());
}
// Create a mutable copy of our maps
let mut refs = HashMap::new();
for (k, v) in self.refs.as_ref() {
@ -98,34 +168,36 @@ impl Graph {
}
// Find the reference to update
let ref_to_update = refs.get(path).ok_or_else(|| format!("Reference with path {} not found", path))?;
let ref_to_update = refs.get(name).ok_or_else(|| format!("Reference with name {} not found", name))?;
// Calculate hash for the new content
let mut hasher = Blake2b512::new();
hasher.update(&new_content);
let new_address = format!("{:x}", hasher.finalize());
// Create updated reference
let updated_ref = Arc::new(Reference {
object_id: new_object_id,
content_address: new_address.clone(),
path: path.to_string(),
dependents: ref_to_update.dependents.clone(),
});
// Create a new reference with the updated content address
// The ID will be calculated based on the content address, name, and dependents
let mut updated_ref = Reference::new(
Some(new_address.clone()),
name.to_string()
);
// Add all dependents to the updated reference
for dep in &ref_to_update.dependents {
updated_ref = updated_ref.add_dep(dep.clone());
}
let updated_ref = updated_ref.to_arc();
// Update objects map with new content
objects.insert(new_address.clone(), new_content);
// Update references map with new reference
refs.insert(path.to_string(), updated_ref.clone());
refs.insert(name.to_string(), updated_ref.clone());
// Find and update all parent references that contain this reference
self.update_parent_references(&mut refs, path)?;
// Update the root reference if needed
if path == &self.root.path {
self.root = refs.get(path).unwrap().clone();
}
self.update_parent_references(&mut refs, name);
// Update the Arc maps
self.refs = Arc::new(refs);
@ -133,102 +205,114 @@ impl Graph {
Ok(())
}
/// TODO(jwall): Add new reference
/// Recursively updates parent references when a child reference changes
fn update_parent_references(&self, refs: &mut HashMap<String, Arc<Reference>>, updated_path: &str) -> Result<(), String> {
fn update_parent_references(&mut self, refs: &mut HashMap<String, Arc<Reference>>, updated_name: &str) {
// Find all references that have the updated reference as a dependent
let parent_paths: Vec<String> = refs
let parent_names: Vec<String> = refs
.iter()
.filter(|(_, r)| r.dependents.iter().any(|dep| dep.path == updated_path))
.map(|(path, _)| path.clone())
.filter(|(_, r)| r.dependents.iter().any(|dep| dep.name == updated_name))
.map(|(name, _)| name.clone())
.collect();
for parent_path in parent_paths {
if let Some(parent_ref) = refs.get(&parent_path) {
// Create a new list of dependents with the updated reference
let mut new_dependents = Vec::new();
for parent_name in parent_names {
if let Some(parent_ref) = refs.get(&parent_name) {
// Create a new reference with the same content address and name
let mut updated_parent = Reference::new(
parent_ref.content_address.clone(),
parent_ref.name.clone()
);
// Add all dependents, replacing the updated one
for dep in &parent_ref.dependents {
if dep.path == updated_path {
// Update the root reference if needed
if dep.name == updated_name {
// Use the updated reference
new_dependents.push(refs.get(updated_path).unwrap().clone());
updated_parent = updated_parent.add_dep(refs.get(updated_name).unwrap().clone());
} else {
// Keep the existing dependent
new_dependents.push(dep.clone());
updated_parent = updated_parent.add_dep(dep.clone());
}
}
// Calculate new content address based on updated dependents
let mut hasher = Blake2b512::new();
for dep in &new_dependents {
hasher.update(&dep.content_address);
// The ID is automatically calculated in the add_dep method
let updated_parent = Arc::new(updated_parent);
if updated_parent.name == self.root.name {
self.root = updated_parent.clone();
}
let new_address = format!("{:x}", hasher.finalize());
// Create updated parent reference
let updated_parent = Arc::new(Reference {
object_id: parent_ref.object_id.clone(),
content_address: new_address,
path: parent_ref.path.clone(),
dependents: new_dependents,
});
// Update the references map
refs.insert(parent_path.clone(), updated_parent);
refs.insert(parent_name.clone(), updated_parent);
// Recursively update parents of this parent
self.update_parent_references(refs, &parent_path)?;
self.update_parent_references(refs, &parent_name);
}
}
Ok(())
}
pub fn random_graph() -> Graph {
let path_root = String::from("ref/0");
let root_name = String::from("ref/0");
let mut objects = HashMap::new();
let mut refs = HashMap::new();
// Create the root reference
let mut root_ref = Reference::new(
"username:0".to_string(),
String::from("0"),
path_root.clone(),
Some(String::from("root_content")),
root_name.clone(),
);
let mut root_hasher = Blake2b512::new();
// Create 10 item references
for i in 1..=10 {
let item_name = format!("/item/{}", i);
let mut item_ref = Reference::new(
format!("item:{}", i),
format!("0:{}", i),
format!("/item/{}", i),
Some(format!("item_content_{}", i)),
item_name.clone(),
);
let mut hasher = Blake2b512::new();
// Create 10 subitems for each item
for j in 1..=10 {
let (address, content) = random_object();
hasher.update(&content);
let subitem_name = format!("/item/{}/subitem/{}", i, j);
// Create a leaf reference
let leaf_ref = Reference::new(
format!("item:{}:subitem:{}", i, j),
format!("{}", address),
format!("/item/{}/subitem/{}", i, j),
)
.to_arc();
Some(address.clone()),
subitem_name,
).to_arc();
// Add the leaf reference as a dependent to the item reference
item_ref = item_ref.add_dep(leaf_ref.clone());
// Store the content in the objects map
objects.insert(address.clone(), content);
hasher.update(&leaf_ref.content_address);
refs.insert(leaf_ref.path.clone(), leaf_ref);
// Store the leaf reference in the refs map
refs.insert(leaf_ref.name.clone(), leaf_ref);
}
let hash = format!("{:x}", hasher.finalize());
item_ref.content_address = hash;
root_hasher.update(&item_ref.content_address);
let rc_ref = item_ref.to_arc();
root_ref = root_ref.add_dep(rc_ref.clone());
refs.insert(rc_ref.path.clone(), rc_ref);
// Convert the item reference to Arc and add it to the root reference
let arc_item_ref = item_ref.to_arc();
root_ref = root_ref.add_dep(arc_item_ref.clone());
// Store the item reference in the refs map
refs.insert(arc_item_ref.name.clone(), arc_item_ref);
}
root_ref.content_address = format!("{:x}", root_hasher.finalize());
let rc_root = root_ref.to_arc();
refs.insert(rc_root.path.clone(), rc_root.clone());
dbg!(&objects);
// Convert the root reference to Arc
let arc_root_ref = root_ref.to_arc();
// Store the root reference in the refs map
refs.insert(arc_root_ref.name.clone(), arc_root_ref.clone());
Graph {
root: rc_root,
root: arc_root_ref,
refs: Arc::new(refs),
objects: Arc::new(objects),
}
}
}
#[cfg(test)]
mod test;

View File

@ -0,0 +1,202 @@
use std::collections::HashMap;
use std::collections::HashSet;
use std::sync::Arc;
use rand::random_range;
use crate::{Graph, Reference, random_object};
fn get_random_candidate(graph: &Graph) -> Arc<Reference> {
// Pick a random leaf node to update
let refs: Vec<Arc<Reference>> =graph.refs.values().filter(|r| r.name != graph.root.name).map(|r| r.clone()).collect();
let random_index = random_range(0..refs.len());
refs[random_index].clone()
}
/// Tests that all dependencies are kept updated when new nodes are added
#[test]
fn test_dependencies_updated_when_nodes_added() {
// Create a simple graph
let mut graph = create_test_graph();
// Get the initial content address of the root
let initial_root_id = graph.root.id.clone();
let candidate = get_random_candidate(&graph);
// Update the leaf node
graph.update_object_reference(&candidate.name, random_object().1).unwrap();
// Verify that the leaf node's ID has changed
let updated_leaf = dbg!(graph.get_reference(&candidate.name).unwrap());
assert_ne!(updated_leaf.id, candidate.id,
"Leaf node ID should change when content is updated");
// Verify that the root's ID has changed
assert_ne!(graph.root.id, initial_root_id,
"Root ID should change when a dependent node is updated");
}
/// Tests that the root of the graph is not itself a dependency of any other node
#[test]
fn test_root_not_a_dependency() {
let graph = create_test_graph();
let root_name = graph.root.name.clone();
// Check all references to ensure none have the root as a dependent
for (_, reference) in graph.refs.as_ref() {
for dep in &reference.dependents {
assert_ne!(dep.name, root_name,
"Root should not be a dependency of any other node");
}
}
}
/// Tests that all nodes are dependents or transitive dependents of the root
#[test]
fn test_all_nodes_connected_to_root() {
let graph = create_test_graph();
// Collect all nodes reachable from the root
let mut reachable = HashSet::new();
fn collect_reachable(node: &Arc<Reference>, reachable: &mut HashSet<String>) {
reachable.insert(node.name.clone());
for dep in &node.dependents {
if !reachable.contains(&dep.name) {
collect_reachable(dep, reachable);
}
}
}
collect_reachable(&graph.root, &mut reachable);
// Check that all nodes in the graph are reachable from the root
for (name, _) in graph.refs.as_ref() {
assert!(reachable.contains(name),
"All nodes should be reachable from the root: {}", name);
}
}
/// Helper function to create a test graph with a known structure
fn create_test_graph() -> Graph {
let root_name = String::from("/root");
let mut objects = HashMap::new();
let mut refs = HashMap::new();
// Create the root reference
let mut root_ref = Reference::new(
Some(String::from("root_content")),
root_name.clone(),
);
// Create 3 item references
for i in 1..=3 {
let item_name = format!("/item/{}", i);
let mut item_ref = Reference::new(
Some(format!("item_content_{}", i)),
item_name.clone(),
);
// Create 3 subitems for each item
for j in 1..=3 {
let (address, content) = random_object();
let subitem_name = format!("/item/{}/subitem/{}", i, j);
// Create a leaf reference
let leaf_ref = Reference::new(
Some(address.clone()),
subitem_name,
).to_arc();
// Add the leaf reference as a dependent to the item reference
item_ref = item_ref.add_dep(leaf_ref.clone());
// Store the content in the objects map
objects.insert(address.clone(), content);
// Store the leaf reference in the refs map
refs.insert(leaf_ref.name.clone(), leaf_ref);
}
// Convert the item reference to Arc and add it to the root reference
let arc_item_ref = item_ref.to_arc();
root_ref = root_ref.add_dep(arc_item_ref.clone());
// Store the item reference in the refs map
refs.insert(arc_item_ref.name.clone(), arc_item_ref);
}
// Convert the root reference to Arc
let arc_root_ref = root_ref.to_arc();
// Store the root reference in the refs map
refs.insert(arc_root_ref.name.clone(), arc_root_ref.clone());
Graph {
root: arc_root_ref,
refs: Arc::new(refs),
objects: Arc::new(objects),
}
}
/// Tests that the graph correctly handles content-addressable properties
#[test]
fn test_content_addressable_properties() {
let mut graph = create_test_graph();
// Update a leaf node with the same content
let leaf_path = "/item/1/subitem/1".to_string();
let initial_leaf = graph.get_reference(&leaf_path).unwrap();
if let Some(content_address) = initial_leaf.content_address.clone() {
// Get the content for this address
let content = graph.get_object(&content_address).unwrap().clone();
// Update with the same content
graph.update_object_reference(&leaf_path, content).unwrap();
}
// Verify that nothing changed since the content is the same
let updated_leaf = graph.get_reference(&leaf_path).unwrap();
assert_eq!(updated_leaf.content_address, initial_leaf.content_address,
"Content address should not change when content remains the same");
}
/// Tests that the graph correctly handles ID calculation
#[test]
fn test_id_calculation() {
let mut graph = create_test_graph();
// Update a leaf node
let leaf_path = "/item/1/subitem/1".to_string();
let initial_leaf = graph.get_reference(&leaf_path).unwrap();
graph.update_object_reference(&leaf_path, "new content".as_bytes().to_vec()).unwrap();
// Verify that the ID changed
let updated_leaf = graph.get_reference(&leaf_path).unwrap();
assert_ne!(updated_leaf.id, initial_leaf.id,
"Reference ID should change when content changes");
// Verify that parent ID changed
let parent_path = "/item/1".to_string();
let parent = graph.get_reference(&parent_path).unwrap();
// Create a reference with the same properties to calculate expected ID
let mut test_ref = Reference::new(
parent.content_address.clone(),
parent.name.clone(),
);
// Add the same dependents
for dep in &parent.dependents {
test_ref = test_ref.add_dep(dep.clone());
}
// Verify the ID calculation is consistent
assert_eq!(parent.id, test_ref.id,
"ID calculation should be consistent for the same reference properties");
}