I have started out learning Rust and is currently trying to write a small neural network as personal exercise. I want to define a struct for my forthcoming Layers/Clusters/Groups of nodes. My initial definition looks like this:
struct Layer {
name: String, // Human readable name
id: String, // UUID in the future
order: u8, // int for sorting
width: u8, // Number of nodes
input: [&'Self], // References to other Layers that feed input into this
}
The thing I am struggling with is the input field which should contain a list of references to other Layer-instances. I will know at compile time how many each Layer will have in the list so it wont have to me mutable. Is it possible to do this? I cant find a solution on the Google machine or in "the book".
Please advise.
Is it possible to do this? I cant find a solution on the Google machine or in "the book".
Possible yes, though I would not recommend it.
Let's start with the possible: &Self would be a "layer reference" with an unnamed lifetime, a lifetime name is for the form '<symbol>, so when you write &'Self you're specifying a reference of lifetime 'Self, but you're never specifying the type being refered to, which is why rustc complains about "expected type".
If you add a "proper" lifetime name, and parametrize the structure, it compiles fine:
struct Layer<'sublayers> {
name: String, // Human readable name
id: String, // UUID in the future
order: u8, // int for sorting
width: u8, // Number of nodes
input: [&'sublayers Self], // References to other Layers that feed input into this
}
However I would not recommend it as the last member being a slice means it's a DST which are difficult to work with at the best of time -- as the nomicon specifically notes "custom DSTs are a largely half-baked feature for now".
Since Rust doesn't yet have const generics proper you can't use an array you'd parameterize through layer either (e.g. Layer<const Size> and input: [&Self;Size], maybe one day), so you probably want something like a vector or a slice reference e.g.
struct Layer<'slice, 'sublayers: 'slice> {
name: String, // Human readable name
id: String, // UUID in the future
order: u8, // int for sorting
width: u8, // Number of nodes
input: &'slice [&'sublayers Self], // References to other Layers that feed input into this
}
Related
We are setting up our rust services and using prost-build to bridge between proto <-> rust land. Our proto definitions are in proto3
Lets take the following proto message:
message Test {
string id = 1;
string body = 2;
string maybe_nullable_thing = 3;
}
This generates a struct like so:
pub struct Test {
#[prost(string, tag="1")]
pub id: ::prost::alloc::string::String,
#[prost(string, tag="2")]
pub body: ::prost::alloc::string::String,
#[prost(string, tag="3")]
pub maybe_nullable_thing: ::prost::alloc::string::String,
}
In other languagues where we have tried this, the fields of a proto message are optional by design and can be left out. In the example there can be cases where maybe_nullable_thing can not be set.
I can work around this issue by using the optional keyword. Altho I remember that it was not the best practice to do so(maybe I am mistaken?)
In terms of best practice with proto3 and rust in general, is it okay to use the optional keyword? If I use serde along with my Test struct I can see the default values of all the fields begin set to "".to_owned() (or empty string).
So I am not sure whats the best practice here? Would love to get some pointers on the best way forward here.
Looking at the readme for Tokio's PROST! tool, it appears their advice is to wrap any non-repeated and non-scalar fields, or any optional fields with Option<T>. This may or may not be different for prost-build, but it should give you a good reference to what's expected when using proto3 and Rust.
In general, however, you should wrap any value you want to be optional in Option<T>. This is not a bad practice, this is the default, standard way to represent "maybe nullable things" in Rust.
This is my data model:
pub struct RaffleDetails {
prize: Balance,
start: Timestamp,
end: Timestamp,
participants: UnorderedMap<AccountId, Balance>,
}
pub struct RaffleDapp {
raffles: UnorderedMap<AccountId, RaffleDetails>,
}
How can I insert a key-value pair in the 'participants' variable?
I tried self.raffles.get(&raffle_account_id).unwrap().participants.insert(&env::predecessor_account_id(), &confidence); but it's not persistent.
References:
UnorderedMap
NEAR Rust SDK
You need to make sure you are updating the RaffleDetails instance that is in the map, not a copy/clone of it.
I'm not familiar with UnorderedMap, but it seems to me the get() method returns a copy of the value that is in the map, so you are only updating the copied value. I don't know if UnorderedMap allows you to mutate a value in it directly (skimming through the docs, I don't see such a method). What you can do though is re-insert the modified RaffleDetails into the raffles map (so as to replace the old one with the modified one).
I'm talking about something like this (I haven't tested compiling it):
let o = self.raffles.get(&raffle_account_id);
if let copied_rd = Some(o) {
copied_rd.participants.insert(&env::predecessor_account_id(), &confidence);
self.raffles.insert(&raffle_account_id, &copied_rd);
}
Recently I started to learn Rust and one of my main struggles is converting years of Object Oriented thinking into procedural code.
I'm trying to parse a XML that have tags that are processed by an specific handler that can deal with the data it gets from the children.
Further more I have some field members that are common between them and I would prefer not to have to write the same fields to all the handlers.
I tried my hand on it and my code came out like this:
use roxmltree::Node; // roxmltree = "0.14.0"
fn get_data_from(node: &Node) -> String {
let tag_name = get_node_name(node);
let tag_handler: dyn XMLTagHandler = match tag_name {
"name" => NameHandler::new(),
"phone" => PhoneHandler::new(),
_ => DefaultHandler::new()
}
if tag_handler.is_recursive() {
for child in node.children() {
let child_value = get_data_from(&child);
// do something with child value
}
}
let value: String = tag_handler.value()
value
}
// consider that handlers are on my project and can be adapted to my needs, and that XMLTagHandler is the trait that they share in common.
My main issues with this are:
This feels like a Object oriented approach to it;
is_recursive needs to be reimplemented to each struct because they traits cannot have field members, and I will have to add more fields later, which means more boilerplate for each new field;
I could use one type for a Handler and pass to it a function pointer, but this approach seems dirty. e.g.:=> Handler::new(my_other_params, phone_handler_func)
This feels like a Object oriented approach to it
Actually, I don't think so. This code is in clear violation of the Tell-Don't-Ask principle, which falls out from the central idea of object-oriented programming: the encapsulation of data and related behavior into objects. The objects (NameHandler, PhoneHandler, etc.) don't have enough knowledge about what they are to do things on their own, so get_data_from has to query them for information and decide what to do, rather than simply sending a message and letting the object figure out how to deal with it.
So let's start by moving the knowledge about what to do with each kind of tag into the handler itself:
trait XmlTagHandler {
fn foreach_child<F: FnMut(&Node)>(&self, node: &Node, callback: F);
}
impl XmlTagHandler for NameHandler {
fn foreach_child<F: FnMut(&Node)>(&self, _node: &Node, _callback: F) {
// "name" is not a recursive tag, so do nothing
}
}
impl XmlTagHandler for DefaultHandler {
fn foreach_child<F: FnMut(&Node)>(&self, node: &Node, callback: F) {
// all other tags may be recursive
for child in node.children() {
callback(child);
}
}
}
This way you call foreach_child on every kind of Handler, and let the handler itself decide whether the right action is to recurse or not. After all, that's why they have different types -- right?
To get rid of the dyn part, which is unnecessary, let's write a little generic helper function that uses XmlTagHandler to handle one specific kind of tag, and modify get_data_from so it just dispatches to the correct parameterized version of it. (I'll suppose that XmlTagHandler also has a new function so that you can create one generically.)
fn handle_tag<H: XmlTagHandler>(node: &Node) -> String {
let handler = H::new();
handler.foreach_child(node, |child| {
// do something with child value
});
handler.value()
}
fn get_data_from(node: &Node) -> String {
let tag_name = get_node_name(node);
match tag_name {
"name" => handle_tag::<NameHandler>(node),
"phone" => handle_tag::<PhoneHandler>(node),
_ => handle_tag::<DefaultHandler>(node),
}
}
If you don't like handle_tag::<SomeHandler>(node), also consider making handle_tag a provided method of XmlTagHandler, so you can instead write SomeHandler::handle(node).
Note that I have not really changed any of the data structures. Your presumption of an XmlTagHandler trait and various Handler implementors is a pretty normal way to organize code. However, in this case, it doesn't offer any real improvement over just writing three separate functions:
fn get_data_from(node: &Node) -> String {
let tag_name = get_node_name(node);
match tag_name {
"name" => get_name_from(node),
"phone" => get_phone_from(node),
_ => get_other_from(node),
}
}
In some languages, such as Java, all code has to be part of some class – so you can find yourself writing classes that don't exist for any other reason than to group related things together. In Rust you don't need to do this, so make sure that any added complication such as XmlTagHandler is actually pulling its weight.
is_recursive needs to be reimplemented to each struct because they traits cannot have field members, and I will have to add more fields later, which means more boilerplate for each new field
Without more information about the fields, it's impossible to really understand what problem you're facing here; however, in general, if there is a family of structs that have some data in common, you may want to make a generic struct instead of a trait. See the answers to How to reuse codes for Binary Search Tree, Red-Black Tree, and AVL Tree? for more suggestions.
I could use one type for a Handler and pass to it a function pointer, but this approach seems dirty
Elegance is sometimes a useful thing, but it is subjective. I would recommend closures rather than function pointers, but this suggestion doesn't seem "dirty" to me. Making closures and putting them in data structures is a very normal way to write Rust code. If you can elaborate on what you don't like about it, perhaps someone could point out ways to improve it.
I want to write a function to be called like this:
send("message","address");
Where some other thread that is doing
let k = recv("address");
println!("{}",k);
sees message.
In particular, the message may be large, and so I'd like "move" or "zero-copy" semantics for sending the message.
In C, the solution is something like:
Allocate messages on the heap
Have a global, threadsafe hashmap that maps "address" to some memory location
Write pointers into the memory location on send, and wake up the receiver using a semaphore
Read pointers out of the memory location on receive, and wait on a semaphore to process new messages
But according to another SO question, step #2 "sounds like a bad idea". So I'd like to see a more Rust-idiomatic way to approach this problem.
You get these sort of move semantics automatically, and get achieve light-weight moves by placing large values into a Box (i.e. allocate them on the heap). Using type ConcurrentHashMap<K, V> = Mutex<HashMap<K, V>>; as the threadsafe hashmap (there's various ways this could be improved), one might have:
use std::collections::{HashMap, RingBuf};
use std::sync::Mutex;
type ConcurrentHashMap<K, V> = Mutex<HashMap<K, V>>;
lazy_static! {
pub static ref MAP: ConcurrentHashMap<String, RingBuf<String>> = {
Mutex::new(HashMap::new())
}
}
fn send(message: String, address: String) {
MAP.lock()
// find the place this message goes
.entry(address)
.get()
// create a new RingBuf if this address was empty
.unwrap_or_else(|v| v.insert(RingBuf::new()))
// add the message on the back
.push_back(message)
}
fn recv(address: &str) -> Option<String> {
MAP.lock()
.get_mut(address)
// pull the message off the front
.and_then(|buf| buf.pop_front())
}
That code is using the lazy_static! macro to achieve a global hashmap (it may be better to use a local object that wraps an Arc<ConcurrentHashMap<...>, fwiw, since global state can make reasoning about program behaviour hard). It also uses RingBuf as a queue, so that messages bank up for a given address. If you only wish to support one message at a time, the type could be ConcurrentHashMap<String, String>, send could become MAP.lock().insert(address, message) and recv just MAP.lock().remove(address).
(NB. I haven't compiled this, so the types may not match up precisely.)
Lets say I want to write a little client for an HTTP API. It has a resource that returns a list of cars:
GET /cars
It also accepts the two optional query parameters color and manufacturer, so I could query specific cars like:
GET /cars?color=black
GET /cars?manufacturer=BMW
GET /cars?color=green&manufacturer=VW
How would I expose these resources properly in Rust? Since Rust doesn't support overloading, defining multiple functions seems to be the usual approach, like:
fn get_cars() -> Cars
fn get_cars_by_color(color: Color) -> Cars
fn get_cars_by_manufacturer(manufacturer: Manufacturer) -> Cars
fn get_cars_by_manufacturer_and_color(manufacturer: Manufacturer, color: Color) -> Cars
But this will obviously not scale when you have more than a few parameters.
Another way would be to use a struct:
struct Parameters {
color: Option<Color>,
manufacturer: Option<Manufacturer>
}
fn get_cars(params: Parameters) -> Cars
This has the same scaling issue, every struct field must be set on creation (even if its value is just None).
I guess I could just accept a HashMap<String, String>, but that doesn't sound very good either.
So my question is, what is the proper/best way to do this in Rust?
You could use the Builder pattern, as mentioned here. For your particular API, it could look like this:
Cars::new_get()
.by_color("black")
.by_manufacturer("BMW")
.exec();
I would like to point out that no matter the solution, if you wish for a compile-time checked solution the "url parsing -> compile-time checkable" translation is necessarily hard-wired. You can generate that with an external script, with macros, etc... but in any case for the compiler to check it, it must exist at compile-time. There just is no short-cut.
Therefore, no matter which API you go for, at some point you will have something akin to:
fn parse_url(url: &str) -> Parameters {
let mut p: Parameters = { None, None };
if let Some(manufacturer) = extract("manufacturer", url) {
p.manufacturer = Some(Manufacturer::new(manufacturer));
}
if let Some(color) = extract("color", url) {
p.color = Some(Color::new(color));
}
p
}
And although you can try and sugarcoat it, the fundamentals won't change.