Add Endor source reader before lexing with provenance and URI provider

crates/frontend/src/reader.rs

@@ -0,0 +1,147 @@
+use std::error::Error;
+use std::fmt;
+use std::fs::File;
+use std::io::{self, Read};
+use std::path::Path;
+use std::string::FromUtf8Error;
+
+use crate::{SourceOrigin, SourceUnit};
+
+/// The host resolves names, authorizes access and selects a revision before
+/// returning its source bytes. No URI path layout or hash algorithm is assumed.
+pub trait EndorSourceProvider {
+    fn load(&self, uri: &str) -> io::Result<ResolvedEndorSource>;
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ResolvedEndorSource {
+    pub revision: String,
+    pub bytes: Vec<u8>,
+}
+
+/// Stateless source adapters, before lexing. Streams are read to EOF; session
+/// prompting and incomplete REPL input detection belong to the interactive host.
+#[derive(Debug, Clone, Copy, Default)]
+pub struct EndorReader;
+
+impl EndorReader {
+    pub fn from_text(name: impl Into<String>, text: impl Into<String>) -> SourceUnit {
+        SourceUnit::new(SourceOrigin::Text { name: name.into() }, text.into())
+    }
+
+    pub fn from_stream(
+        name: impl Into<String>,
+        reader: impl Read,
+    ) -> Result<SourceUnit, SourceReadError> {
+        Self::read(SourceOrigin::Stream { name: name.into() }, reader)
+    }
+
+    pub fn from_file(path: impl AsRef<Path>) -> Result<SourceUnit, SourceReadError> {
+        let path = path.as_ref();
+        let origin = SourceOrigin::File {
+            path: path.to_owned(),
+        };
+        let file = File::open(path).map_err(|error| SourceReadError {
+            origin: origin.to_string(),
+            kind: SourceReadErrorKind::Io(error),
+        })?;
+        Self::read(origin, file)
+    }
+
+    pub fn from_endor<P: EndorSourceProvider + ?Sized>(
+        uri: &str,
+        provider: &P,
+    ) -> Result<SourceUnit, SourceReadError> {
+        // Only recognize the scheme and reject empty/control-filled locators.
+        // Namespace syntax, revision selectors and percent decoding are host policy.
+        let valid = uri.split_once(':').is_some_and(|(scheme, address)| {
+            scheme.eq_ignore_ascii_case("endor") && !address.is_empty()
+        }) && !uri
+            .chars()
+            .any(|character| character.is_whitespace() || character.is_control());
+        if !valid {
+            return Err(SourceReadError {
+                origin: uri.into(),
+                kind: SourceReadErrorKind::InvalidEndorUri,
+            });
+        }
+        let resolved = provider.load(uri).map_err(|error| SourceReadError {
+            origin: uri.into(),
+            kind: SourceReadErrorKind::Io(error),
+        })?;
+        if resolved.revision.trim().is_empty() {
+            return Err(SourceReadError {
+                origin: uri.into(),
+                kind: SourceReadErrorKind::MissingRevision,
+            });
+        }
+        let origin = SourceOrigin::Endor {
+            uri: uri.into(),
+            revision: resolved.revision,
+        };
+        Self::decode(origin, resolved.bytes)
+    }
+
+    fn read(origin: SourceOrigin, mut reader: impl Read) -> Result<SourceUnit, SourceReadError> {
+        let mut bytes = Vec::new();
+        reader
+            .read_to_end(&mut bytes)
+            .map_err(|error| SourceReadError {
+                origin: origin.to_string(),
+                kind: SourceReadErrorKind::Io(error),
+            })?;
+        Self::decode(origin, bytes)
+    }
+
+    fn decode(origin: SourceOrigin, bytes: Vec<u8>) -> Result<SourceUnit, SourceReadError> {
+        let text = String::from_utf8(bytes).map_err(|error| SourceReadError {
+            origin: origin.to_string(),
+            kind: SourceReadErrorKind::InvalidUtf8(error),
+        })?;
+        Ok(SourceUnit::new(origin, text))
+    }
+}
+
+#[derive(Debug)]
+pub enum SourceReadErrorKind {
+    Io(io::Error),
+    InvalidUtf8(FromUtf8Error),
+    InvalidEndorUri,
+    MissingRevision,
+}
+
+#[derive(Debug)]
+pub struct SourceReadError {
+    pub origin: String,
+    pub kind: SourceReadErrorKind,
+}
+
+impl fmt::Display for SourceReadError {
+    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(formatter, "{}: ", self.origin)?;
+        match &self.kind {
+            SourceReadErrorKind::Io(error) => write!(formatter, "{error}"),
+            SourceReadErrorKind::InvalidUtf8(error) => write!(
+                formatter,
+                "invalid UTF-8 at byte {}",
+                error.utf8_error().valid_up_to()
+            ),
+            SourceReadErrorKind::InvalidEndorUri => formatter.write_str(
+                "expected a nonempty endor URI without whitespace or control characters",
+            ),
+            SourceReadErrorKind::MissingRevision => {
+                formatter.write_str("source provider returned no resolved revision")
+            }
+        }
+    }
+}
+
+impl Error for SourceReadError {
+    fn source(&self) -> Option<&(dyn Error + 'static)> {
+        match &self.kind {
+            SourceReadErrorKind::Io(error) => Some(error),
+            SourceReadErrorKind::InvalidUtf8(error) => Some(error),
+            _ => None,
+        }
+    }
+}