Change rbx_reflection_database to support loading from FS

rbx_binary/src/deserializer/mod.rs

@@ -54,7 +54,7 @@ impl<'db> Deserializer<'db> {
     /// Create a new `Deserializer` with the default settings.
     pub fn new() -> Self {
         Self {
-            database: rbx_reflection_database::get(),
+            database: rbx_reflection_database::get().unwrap(),
         }
     }
 

rbx_binary/src/serializer/mod.rs

@@ -53,7 +53,7 @@ impl<'db> Serializer<'db> {
     /// Create a new `Serializer` with the default settings.
     pub fn new() -> Self {
         Serializer {
-            database: rbx_reflection_database::get(),
+            database: rbx_reflection_database::get().unwrap(),
             compression: CompressionType::default(),
         }
     }

rbx_reflection_database/CHANGELOG.md

@@ -5,7 +5,13 @@
 * `Model.WorldPivotData`'s default value is now `null`. ([#450])
 * `SharedString` properties now have default values. ([#458])
 * `Instance.DefinesCapabilities` is now an alias for `Instance.Sandboxed` ([#459])
+- The database may now be loaded dynamically from the local file system. ([#376])
 
+  The location is OS-dependent but it will only be loaded if one exists. The location may also be manually specified using the `RBX_DATABASE` environment variable.
+
+  `get` is unchanged in use, but will return a locally stored database if it exists, and the bundled one if not. Two new methods were added: `get_bundled` will only fetch the local database and `get_local` will only fetch a locally stored one.
+
+[#376]: https://github.com/rojo-rbx/rbx-dom/pull/376
 [#458]: https://github.com/rojo-rbx/rbx-dom/pull/458
 [#450]: https://github.com/rojo-rbx/rbx-dom/pull/450
 [#459]: https://github.com/rojo-rbx/rbx-dom/pull/459

rbx_reflection_database/Cargo.toml

@@ -18,3 +18,5 @@ rbx_reflection = { version = "4.7.0", path = "../rbx_reflection" }
 lazy_static = "1.4.0"
 serde = "1.0.137"
 rmp-serde = "1.1.1"
+dirs = "5.0.1"
+log = "0.4.20"

rbx_reflection_database/README.md

@@ -4,4 +4,24 @@
 
 More details about this crate are available on [the rbx-dom GitHub](https://github.com/rojo-rbx/rbx-dom#readme).
 
-Contains a generated Roblox reflection database using the types from [rbx_reflection](https://crates.io/crates/rbx_reflection). This crate embeds a MessagePack-format database that is exposed through this crate's only method, `get`.
+Contains an API to get a Roblox reflection database using the types from [`rbx_reflection`](https://crates.io/crates/rbx_reflection). This crate embeds a database for this purpose, but also provides an API for dependents to get a reflection database from a consistent location.
+
+The general way this crate should be used is via `get`. This method will search for a locally stored reflection database and return it if it's found. If it isn't, it will instead return the bundled one. The details for where it searches are below.
+
+Additionally, this crate exposes `get_local` and `get_bundled` for only loading the locally stored database or only the bundled one respectively.
+
+## Local Details
+
+This crate will load a reflection database from the file system if one exists in the default location. This location varies upon the OS and is specified here:
+
+| OS      | Location                                                            |
+|:--------|:--------------------------------------------------------------------|
+| Windows | `%localappdata%/.rbxreflection/database.msgpack`                    |
+| MacOS   | `$HOME/Library/Application Support/.rbxreflection/database.msgpack` |
+| Linux   | `$HOME/.rbxreflection/database.msgpack`                             |
+
+Additionally, a location override may be specified via the `RBX_DATABASE` environment variable. The `RBX_DATABASE` variable points to the override `database.msgpack` file, _not_ to an override `.rbxreflection` directory.
+
+Both the default `database.msgpack` files and any files pointed to by `RBX_DATABASE` must be valid MessagePack serializations of a [`ReflectionDatabase`][ReflectionDatabase] if they're present.
+
+[ReflectionDatabase]: https://docs.rs/rbx_reflection/latest/rbx_reflection/struct.ReflectionDatabase.html

rbx_reflection_database/src/error.rs

@@ -0,0 +1,24 @@
+use std::{fmt, io};
+
+#[derive(Debug, Clone)]
+pub struct Error(String);
+
+impl std::error::Error for Error {}
+
+impl fmt::Display for Error {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "{}", self.0)
+    }
+}
+
+impl From<rmp_serde::decode::Error> for Error {
+    fn from(value: rmp_serde::decode::Error) -> Self {
+        Self(value.to_string())
+    }
+}
+
+impl From<io::Error> for Error {
+    fn from(value: io::Error) -> Self {
+        Self(value.to_string())
+    }
+}

rbx_reflection_database/src/lib.rs

@@ -1,15 +1,144 @@
+//! Contains an API to get a Roblox reflection database using the types
+//! from [`rbx_reflection`]. This crate embeds a database for this purpose,
+//! but also provides an API for dependents to get a reflection database
+//! from a consistent location.
+//!
+//! The general way this crate should be used is via [`get`]. This method will
+//! search for a locally stored reflection database and return it if it's
+//! found. If it isn't, it will instead return the bundled one.
+//!
+//! Additionally, this crate exposes [`get_local`] and [`get_bundled`] for
+//! only loading the locally stored database or only the bundled one
+//! respectively.
+//!
+//! ## Local Details
+//!
+//! This crate will load a reflection database from the file system if one
+//! exists in the default location. This location varies upon the OS and is
+//! specified here:
+//!
+//! | OS      | Location                                                            |
+//! |:--------|:--------------------------------------------------------------------|
+//! | Windows | `%localappdata%/.rbxreflection/database.msgpack`                    |
+//! | MacOS   | `$HOME/Library/Application Support/.rbxreflection/database.msgpack` |
+//! | Linux   | `$HOME/.rbxreflection/database.msgpack`                             |
+//!
+//! Additionally, a location override may be specified via the `RBX_DATABASE`
+//! environment variable. The `RBX_DATABASE` variable points to the override
+//! `database.msgpack` file, _not_ to an override `.rbxreflection` directory.
+//!
+//! Both the default `database.msgpack` files and any files pointed to by
+//! `RBX_DATABASE` must be valid MessagePack serializations of a
+//! [`ReflectionDatabase`] if they're present.
+mod error;
+
 use rbx_reflection::ReflectionDatabase;
 
+use std::{env, fs, path::PathBuf, sync::OnceLock};
+
+pub use error::Error;
+
+/// An alias to avoid overly verbose types.
+type ResultOption<T> = Result<Option<T>, Error>;
+
 static ENCODED_DATABASE: &[u8] = include_bytes!("../database.msgpack");
 
+/// The name of an environment variable that may be used to specify
+/// the location of a reflection database to use. The expected format of
+/// a file at this point is MessagePack.
+pub const OVERRIDE_PATH_VAR: &str = "RBX_DATABASE";
+
+/// The name of the directory used for the local location for a reflection
+/// database. The directory will be placed inside the current user's
+/// local data folder on MacOS and Windows and inside
+/// the home directory on Linux.
+pub const LOCAL_DIR_NAME: &str = ".rbxreflection";
+
 lazy_static::lazy_static! {
-    static ref DATABASE: ReflectionDatabase<'static> = {
+    static ref BUNDLED_DATABASE: ReflectionDatabase<'static> = {
+        log::debug!("Loading bundled reflection database");
         rmp_serde::decode::from_slice(ENCODED_DATABASE).unwrap_or_else(|e| panic!("could not decode reflection database because: {}", e))
     };
 }
 
-pub fn get() -> &'static ReflectionDatabase<'static> {
-    &DATABASE
+static LOCAL_DATABASE: OnceLock<ResultOption<ReflectionDatabase<'static>>> = OnceLock::new();
+
+/// Returns a populated [`ReflectionDatabase`]. This will attempt to load one locally and
+/// if one can't be found, it will return one that is bundled with this crate.
+///
+/// ## Errors
+///
+/// Errors if a locally stored [`ReflectionDatabase`] could not be read
+/// or is invalid MessagePack.
+pub fn get() -> Result<&'static ReflectionDatabase<'static>, Error> {
+    Ok(get_local()?.unwrap_or(&BUNDLED_DATABASE))
+}
+
+/// Returns a reflection database from the file system, if one can be found.
+/// This is loaded from a location set by the `RBX_DATABASE` environment
+/// variable if it's set. Otherwise, the default location is checked.
+///
+/// The default location varies depending upon OS:
+///
+/// | OS      | Location                                                            |
+/// |:--------|:--------------------------------------------------------------------|
+/// | Windows | `%localappdata%/.rbxreflection/database.msgpack`                    |
+/// | MacOS   | `$HOME/Library/Application Support/.rbxreflection/database.msgpack` |
+/// | Linux   | `$HOME/.rbxreflection/database.msgpack`                             |
+///
+/// The file at the above location (or the one pointed to by `RBX_DATABASE`)
+/// must be valid MessagePack.
+///
+/// ## Errors
+///
+/// Errors if the file specified by `RBX_DATABASE` or in the default location
+/// exists but is invalid MessagePack.
+pub fn get_local() -> ResultOption<&'static ReflectionDatabase<'static>> {
+    let inner = LOCAL_DATABASE.get_or_init(|| {
+        if let Some(path) = get_local_location() {
+            if path.exists() {
+                let database: ReflectionDatabase<'static> =
+                    rmp_serde::from_slice(&fs::read(path)?)?;
+                Ok(Some(database))
+            } else {
+                Ok(None)
+            }
+        } else {
+            Ok(None)
+        }
+    });
+    match inner {
+        Ok(opt) => Ok(opt.as_ref()),
+        // This clone could be avoided because these references are static,
+        // but it'd involve some indirection and these errors are rare anyway.
+        Err(e) => Err(e.clone()),
+    }
+}
+
+/// Returns the locally bundled [`ReflectionDatabase`]. This database may or may
+/// not be up to date, but it will always exist.
+pub fn get_bundled() -> &'static ReflectionDatabase<'static> {
+    &BUNDLED_DATABASE
+}
+
+/// Fetches the location a [`ReflectionDatabase`] is expected to be loaded from.
+/// This may return [`None`] if the local data directory cannot be found.
+pub fn get_local_location() -> Option<PathBuf> {
+    if let Ok(location) = env::var(OVERRIDE_PATH_VAR) {
+        log::debug!("Using environment variable {OVERRIDE_PATH_VAR} to fetch reflection database");
+        Some(PathBuf::from(location))
+    } else {
+        // Due to concerns about the local data directory existing
+        // on Linux, we use the home directory instead.
+        #[cfg(target_os = "linux")]
+        let mut home = dirs::home_dir()?;
+        #[cfg(not(target_os = "linux"))]
+        let mut home = dirs::data_local_dir()?;
+
+        home.push(LOCAL_DIR_NAME);
+        home.push("database.msgpack");
+        Some(home)
+    }
 }
 
 #[cfg(test)]
@@ -19,13 +148,24 @@ mod test {
     use super::*;
 
     #[test]
-    fn smoke_test() {
-        let _database = get();
+    fn bundled() {
+        let _database = get_bundled();
+    }
+
+    #[test]
+    fn env_var() {
+        let mut test_path = PathBuf::from(env!("CARGO_MANIFEST_DIR"));
+        test_path.push("empty.msgpack");
+
+        env::set_var(OVERRIDE_PATH_VAR, &test_path);
+        let empty_db = get().unwrap();
+        println!("{:?}", empty_db.version);
+        assert!(empty_db.version == [0, 0, 0, 0]);
     }
 
     #[test]
     fn superclasses_iter_test() {
-        let database = get();
+        let database = get_bundled();
         let part_class_descriptor = database.classes.get("Part");
         let mut iter = database.superclasses_iter(part_class_descriptor.unwrap());
         fn class_descriptor_eq(lhs: Option<&ClassDescriptor>, rhs: Option<&ClassDescriptor>) {
@@ -51,7 +191,7 @@ mod test {
 
     #[test]
     fn has_superclass_test() {
-        let database = get();
+        let database = get_bundled();
         let part_class_descriptor = database.classes.get("Part").unwrap();
         let instance_class_descriptor = database.classes.get("Instance").unwrap();
         assert!(database.has_superclass(part_class_descriptor, instance_class_descriptor));

rbx_xml/src/deserializer.rs

@@ -78,7 +78,7 @@ impl<'db> DecodeOptions<'db> {
     pub fn new() -> Self {
         DecodeOptions {
             property_behavior: DecodePropertyBehavior::IgnoreUnknown,
-            database: rbx_reflection_database::get(),
+            database: rbx_reflection_database::get().unwrap(),
         }
     }
 

rbx_xml/src/serializer.rs

@@ -82,7 +82,7 @@ impl<'db> EncodeOptions<'db> {
     pub fn new() -> Self {
         EncodeOptions {
             property_behavior: EncodePropertyBehavior::IgnoreUnknown,
-            database: rbx_reflection_database::get(),
+            database: rbx_reflection_database::get().unwrap(),
         }
     }