more improvements for Rust bindgen (#558)

* add clone for configs

* add open_path

* report unused

* add stub

* add rust parser to extract cdk functions; didc to take external.rust section

* fix mode

* refactor to report span

* checkpoint, use codespan_reporting

* good

* fix

* move rust_check to cdk

* bump version

* add def_use analysis

* add custom target for didc

* readme

* generate only subset of bindings

* s/method/func/

* checkpoint

* checkpoint

* checkpoint

* use context

* fix

* add config_source

* check unused at the property level

* no scoped path

* specialize result type

* fix

* prompt color

* add use_type test

* update tests

* bump version

* fix

Changelog.md

@@ -3,30 +3,25 @@
 
 ## 2024-05-03
 
-### candid_parser 0.2.0-beta.0
+### candid_parser 0.2.0-beta
 
 * Breaking changes:
   + Rewrite `configs` and `random` modules, adapting TOML format as the parser. `configs` module is no longer under a feature flag, and no longer depend on dhall.
   + Rewrite Rust bindgen to use the new `configs` module. Use `emit_bindgen` to generate type bindings, and use `output_handlebar` to provide a handlebar template for the generated module. `compile` function provides a default template. The generated file without config is exactly the same as before.
-
-* TO-DOs
-  + Spec for path and matching semantics
-  + Warning for unused path
-  + Rust bindgen:
-    * Generate `use_type` tests. How to handle recursive types?
-    * Threading state through `nominalize`
-    * When the path starts with method, duplicate type definition when necessary.
-    * Number label renaming
+* Non-breaking changes:
+  + `utils::check_rust_type` function to check if a Rust type implements the provided candid type.
 
 ## 2024-04-11
 
-### Candid 0.10.5 -- 0.10.8
+### Candid 0.10.5 -- 0.10.9
 
 * Switch `HashMap` to `BTreeMap` in serialization and `T::ty()`. This leads to around 20% perf improvement for serializing complicated types.
 * Disable memoization for unrolled types in serialization to save cycle cost. In some cases, type table can get slightly larger, but it's worth the trade off.
 * Fix bug in `text_size`
 * Fix decoding cost calculation overflow
 * Fix length check in decoding principal type
+* Implement `CandidType` for `serde_bytes::ByteArray`
+* Add `pretty::candid::pp_init_args` function to pretty print init args
 
 ## 2024-02-27
 

rust/candid/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "candid"
-version = "0.10.8"
+version = "0.10.9"
 edition = "2021"
 rust-version.workspace = true
 authors = ["DFINITY Team"]
@@ -39,6 +39,7 @@ rand.workspace = true
 serde_cbor = "0.11.2"
 serde_json = "1.0.74"
 bincode = "1.3.3"
+candid_parser = { path = "../candid_parser" }
 
 [features]
 bignum = ["dep:num-bigint", "dep:num-traits"]

rust/candid/src/pretty/candid.rs

@@ -205,6 +205,9 @@ fn pp_actor(ty: &Type) -> RcDoc {
     }
 }
 
+pub fn pp_init_args<'a>(env: &'a TypeEnv, args: &'a [Type]) -> RcDoc<'a> {
+    pp_defs(env).append(pp_args(args))
+}
 pub fn compile(env: &TypeEnv, actor: &Option<Type>) -> String {
     match actor {
         None => pp_defs(env).pretty(LINE_WIDTH).to_string(),

rust/candid/tests/serde.rs

@@ -2,6 +2,7 @@ use candid::{
     decode_one_with_config, encode_one, CandidType, Decode, DecoderConfig, Deserialize, Encode,
     Int, Nat,
 };
+use candid_parser::utils::check_rust_type;
 
 #[test]
 fn test_error() {
@@ -263,6 +264,8 @@ fn test_struct() {
     // with memoization on the unrolled type, type table will have 2 entries.
     // all_check(list, "4449444c026e016c02a0d2aca8047c90eddae70400010000");
     all_check(list, "4449444c036e016c02a0d2aca8047c90eddae704026e01010000");
+    check_rust_type::<List>("type List = record {head: int; tail: opt List}; (List)").unwrap();
+    check_rust_type::<List>("type T = record {head: int; tail: opt T}; (T)").unwrap();
 }
 
 #[test]
@@ -289,6 +292,10 @@ fn optional_fields() {
         bar: bool,
         baz: Option<New>,
     }
+    check_rust_type::<NewStruct>(
+        "(record { foo: opt nat8; bar: bool; baz: opt variant { Foo; Bar: bool; Baz: bool }})",
+    )
+    .unwrap();
     let bytes = encode(&OldStruct {
         bar: true,
         baz: Some(Old::Foo),
@@ -334,6 +341,13 @@ fn test_equivalent_types() {
     struct TypeB {
         typea: Option<TypeA>,
     }
+    check_rust_type::<RootType>(
+        r#"
+type A = record { typeb: opt B };
+type B = record { typea: opt A };
+(record { typeas: vec A })"#,
+    )
+    .unwrap();
     // Encode to the following types leads to equivalent but different representations of TypeA
     all_check(
         RootType { typeas: Vec::new() },
@@ -569,6 +583,7 @@ fn test_tuple() {
         (Int::from(42), "💩".to_string()),
         "4449444c016c02007c017101002a04f09f92a9",
     );
+    check_rust_type::<(Int, String)>("(record {int; text})").unwrap();
     let none: Option<String> = None;
     let bytes =
         hex("4449444c046c04007c017e020103026d7c6e036c02a0d2aca8047c90eddae7040201002b010302030400");
@@ -677,6 +692,7 @@ fn test_field_rename() {
         #[serde(rename = "a-b")]
         B(u8),
     }
+    check_rust_type::<E2>(r#"(variant { "1-2 + 3": int8; "a-b": nat8 })"#).unwrap();
     all_check(E2::A(42), "4449444c016b02b684a7027bb493ee970d770100012a");
     #[derive(CandidType, Deserialize, PartialEq, Debug)]
     struct S2 {
@@ -698,7 +714,7 @@ fn test_generics() {
         g1: T,
         g2: E,
     }
-
+    check_rust_type::<G<i32, bool>>("(record {g1: int32; g2: bool})").unwrap();
     let res = G {
         g1: Int::from(42),
         g2: true,

rust/candid_parser/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "candid_parser"
-version = "0.2.0-beta.1"
+version = "0.2.0-beta.2"
 edition = "2021"
 rust-version.workspace = true
 authors = ["DFINITY Team"]
@@ -33,13 +33,13 @@ logos = "0.14"
 convert_case = "0.6"
 handlebars = "5.1"
 toml = { version = "0.8", default-features = false, features = ["parse"] }
+console = "0.15"
 
 arbitrary = { workspace = true, optional = true }
 fake = { version = "2.4", optional = true }
 rand = { version = "0.8", optional = true }
 num-traits = { workspace = true, optional = true }
 dialoguer = { version = "0.11", default-features = false, features = ["editor", "completion"], optional = true }
-console = { version = "0.15", optional = true }
 ctrlc = { version = "3.4", optional = true }
 
 [dev-dependencies]
@@ -49,7 +49,7 @@ rand.workspace = true
 
 [features]
 random = ["dep:arbitrary", "dep:fake", "dep:rand", "dep:num-traits"]
-assist = ["dep:dialoguer", "dep:console", "dep:ctrlc"]
+assist = ["dep:dialoguer", "dep:ctrlc"]
 all = ["random", "assist"]
 
 # docs.rs-specific configuration

rust/candid_parser/src/bindings/analysis.rs

@@ -1,6 +1,17 @@
 use crate::Result;
 use candid::types::{Type, TypeEnv, TypeInner};
-use std::collections::BTreeSet;
+use std::collections::{BTreeMap, BTreeSet};
+
+/// Select a subset of methods from an actor.
+pub fn project_methods(env: &TypeEnv, actor: &Option<Type>, methods: &[String]) -> Option<Type> {
+    let service = env.as_service(actor.as_ref()?).ok()?;
+    let filtered = service
+        .iter()
+        .filter(|(name, _)| methods.contains(name))
+        .cloned()
+        .collect();
+    Some(TypeInner::Service(filtered).into())
+}
 
 /// Same as chase_actor, with seen set as part of the type. Used for chasing type names from type definitions.
 pub fn chase_type<'a>(
@@ -53,6 +64,47 @@ pub fn chase_actor<'a>(env: &'a TypeEnv, actor: &'a Type) -> Result<Vec<&'a str>
     chase_type(&mut seen, &mut res, env, actor)?;
     Ok(res)
 }
+/// Given an actor, return a map from variable names to the (methods, arg) that use them.
+pub fn chase_def_use<'a>(
+    env: &'a TypeEnv,
+    actor: &'a Type,
+) -> Result<BTreeMap<String, Vec<String>>> {
+    let mut res = BTreeMap::new();
+    let actor = env.trace_type(actor)?;
+    if let TypeInner::Class(args, _) = actor.as_ref() {
+        for (i, arg) in args.iter().enumerate() {
+            let mut used = Vec::new();
+            chase_type(&mut BTreeSet::new(), &mut used, env, arg)?;
+            for var in used {
+                res.entry(var.to_string())
+                    .or_insert_with(Vec::new)
+                    .push(format!("init.arg{}", i));
+            }
+        }
+    }
+    for (id, ty) in env.as_service(&actor)? {
+        let func = env.as_func(ty)?;
+        for (i, arg) in func.args.iter().enumerate() {
+            let mut used = Vec::new();
+            chase_type(&mut BTreeSet::new(), &mut used, env, arg)?;
+            for var in used {
+                res.entry(var.to_string())
+                    .or_insert_with(Vec::new)
+                    .push(format!("{}.arg{}", id, i));
+            }
+        }
+        for (i, arg) in func.rets.iter().enumerate() {
+            let mut used = Vec::new();
+            chase_type(&mut BTreeSet::new(), &mut used, env, arg)?;
+            for var in used {
+                res.entry(var.to_string())
+                    .or_insert_with(Vec::new)
+                    .push(format!("{}.ret{}", id, i));
+            }
+        }
+    }
+    Ok(res)
+}
 
 pub fn chase_types<'a>(env: &'a TypeEnv, tys: &'a [Type]) -> Result<Vec<&'a str>> {
     let mut seen = BTreeSet::new();