Parsers: adds "unchecked" option for faster parsing

Does not validate IRIs and language tags

Cargo.lock

@@ -1113,9 +1113,9 @@ checksum = "8d91edf4fbb970279443471345a4e8c491bf05bb283b3e6c88e4e606fd8c181b"
 
 [[package]]
 name = "oxiri"
-version = "0.2.2"
+version = "0.2.3-alpha.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bb175ec8981211357b7b379869c2f8d555881c55ea62311428ec0de46d89bd5c"
+checksum = "b225dad32cfaa43a960b93f01fa7f87528ac07e794b80f6d9a0153e0222557e2"
 
 [[package]]
 name = "oxrdf"

cli/Cargo.toml

@@ -32,7 +32,7 @@ clap = { version = "4.0", features = ["derive"] }
 oxigraph = { version = "0.4.0-alpha.1-dev", path = "../lib" }
 rand = "0.8"
 url = "2.4"
-oxiri = "0.2"
+oxiri = "0.2.3-alpha.1"
 flate2 = "1.0"
 rayon-core = "1.11"
 

cli/src/main.rs

@@ -121,6 +121,8 @@ enum Command {
         destination: PathBuf,
     },
     /// Load file(s) into the store.
+    ///
+    /// Feel free to enable the --lenient option if you know your input is valid to get better performances.
     Load {
         /// Directory in which Oxigraph data are persisted.
         #[arg(short, long, value_hint = ValueHint::DirPath)]
@@ -143,6 +145,8 @@ enum Command {
         #[arg(long, value_hint = ValueHint::Url)]
         base: Option<String>,
         /// Attempt to keep loading even if the data file is invalid.
+        ///
+        /// This disables most of validation on RDF content.
         #[arg(long)]
         lenient: bool,
         /// Name of the graph to load the data to.
@@ -391,6 +395,7 @@ pub fn main() -> anyhow::Result<()> {
                     format.context("The --format option must be set when loading from stdin")?,
                     base.as_deref(),
                     graph,
+                    lenient,
                 )
             } else {
                 ThreadPoolBuilder::new()
@@ -444,6 +449,7 @@ pub fn main() -> anyhow::Result<()> {
                                             }),
                                             base.as_deref(),
                                             graph,
+                                            lenient,
                                         )
                                     } else {
                                         bulk_load(
@@ -454,6 +460,7 @@ pub fn main() -> anyhow::Result<()> {
                                             }),
                                             base.as_deref(),
                                             graph,
+                                            lenient,
                                         )
                                     }
                                 } {
@@ -784,6 +791,7 @@ fn bulk_load(
     format: RdfFormat,
     base_iri: Option<&str>,
     to_graph_name: Option<NamedNode>,
+    lenient: bool,
 ) -> anyhow::Result<()> {
     let mut parser = RdfParser::from_format(format);
     if let Some(to_graph_name) = to_graph_name {
@@ -794,6 +802,9 @@ fn bulk_load(
             .with_base_iri(base_iri)
             .with_context(|| format!("Invalid base IRI {base_iri}"))?;
     }
+    if lenient {
+        parser = parser.unchecked();
+    }
     loader.load_from_read(parser, read)?;
     Ok(())
 }

fuzz/fuzz_targets/nquads.rs

@@ -4,43 +4,60 @@ use libfuzzer_sys::fuzz_target;
 use oxrdf::Quad;
 use oxttl::{NQuadsParser, NQuadsSerializer};
 
-fn parse<'a>(chunks: impl IntoIterator<Item = &'a [u8]>) -> (Vec<Quad>, Vec<String>) {
+fn parse<'a>(
+    chunks: impl IntoIterator<Item = &'a [u8]>,
+    unchecked: bool,
+) -> (Vec<Quad>, Vec<String>) {
     let mut quads = Vec::new();
     let mut errors = Vec::new();
-    let mut parser = NQuadsParser::new().with_quoted_triples().parse();
+    let mut parser = NQuadsParser::new().with_quoted_triples();
+    if unchecked {
+        parser = parser.unchecked();
+    }
+    let mut reader = parser.parse();
     for chunk in chunks {
-        parser.extend_from_slice(chunk);
-        while let Some(result) = parser.read_next() {
+        reader.extend_from_slice(chunk);
+        while let Some(result) = reader.read_next() {
             match result {
                 Ok(quad) => quads.push(quad),
                 Err(error) => errors.push(error.to_string()),
             }
         }
     }
-    parser.end();
-    while let Some(result) = parser.read_next() {
+    reader.end();
+    while let Some(result) = reader.read_next() {
         match result {
             Ok(quad) => quads.push(quad),
             Err(error) => errors.push(error.to_string()),
         }
     }
-    assert!(parser.is_end());
+    assert!(reader.is_end());
     (quads, errors)
 }
 
 fuzz_target!(|data: &[u8]| {
     // We parse with splitting
-    let (quads, errors) = parse(data.split(|c| *c == 0xFF));
+    let (quads, errors) = parse(data.split(|c| *c == 0xFF), false);
     // We parse without splitting
-    let (quads_without_split, errors_without_split) = parse([data
+    let (quads_without_split, errors_without_split) = parse(
+        [data
             .iter()
             .copied()
             .filter(|c| *c != 0xFF)
             .collect::<Vec<_>>()
-        .as_slice()]);
+            .as_slice()],
+        false,
+    );
     assert_eq!(quads, quads_without_split);
     assert_eq!(errors, errors_without_split);
 
+    // We test also unchecked if valid
+    if errors.is_empty() {
+        let (quads_unchecked, errors_unchecked) = parse(data.split(|c| *c == 0xFF), true);
+        assert!(errors_unchecked.is_empty());
+        assert_eq!(quads, quads_unchecked);
+    }
+
     // We serialize
     let mut writer = NQuadsSerializer::new().serialize_to_write(Vec::new());
     for quad in &quads {

fuzz/fuzz_targets/trig.rs

@@ -4,31 +4,37 @@ use libfuzzer_sys::fuzz_target;
 use oxrdf::{Dataset, GraphName, Quad, Subject, Term, Triple};
 use oxttl::{TriGParser, TriGSerializer};
 
-fn parse<'a>(chunks: impl IntoIterator<Item = &'a [u8]>) -> (Vec<Quad>, Vec<String>) {
+fn parse<'a>(
+    chunks: impl IntoIterator<Item = &'a [u8]>,
+    unchecked: bool,
+) -> (Vec<Quad>, Vec<String>) {
     let mut quads = Vec::new();
     let mut errors = Vec::new();
     let mut parser = TriGParser::new()
         .with_quoted_triples()
         .with_base_iri("http://example.com/")
-        .unwrap()
-        .parse();
+        .unwrap();
+    if unchecked {
+        parser = parser.unchecked();
+    }
+    let mut reader = parser.parse();
     for chunk in chunks {
-        parser.extend_from_slice(chunk);
-        while let Some(result) = parser.read_next() {
+        reader.extend_from_slice(chunk);
+        while let Some(result) = reader.read_next() {
             match result {
                 Ok(quad) => quads.push(quad),
                 Err(error) => errors.push(error.to_string()),
             }
         }
     }
-    parser.end();
-    while let Some(result) = parser.read_next() {
+    reader.end();
+    while let Some(result) = reader.read_next() {
         match result {
             Ok(quad) => quads.push(quad),
             Err(error) => errors.push(error.to_string()),
         }
     }
-    assert!(parser.is_end());
+    assert!(reader.is_end());
     (quads, errors)
 }
 
@@ -66,14 +72,22 @@ fn serialize_quads(quads: &[Quad]) -> Vec<u8> {
 
 fuzz_target!(|data: &[u8]| {
     // We parse with splitting
-    let (quads, errors) = parse(data.split(|c| *c == 0xFF));
+    let (quads, errors) = parse(data.split(|c| *c == 0xFF), false);
     // We parse without splitting
-    let (quads_without_split, errors_without_split) = parse([data
+    let (quads_without_split, errors_without_split) = parse(
+        [data
             .iter()
             .copied()
             .filter(|c| *c != 0xFF)
             .collect::<Vec<_>>()
-        .as_slice()]);
+            .as_slice()],
+        false,
+    );
+    let (quads_unchecked, errors_unchecked) = parse(data.split(|c| *c == 0xFF), true);
+    if errors.is_empty() {
+        assert!(errors_unchecked.is_empty());
+    }
+
     let bnodes_count = quads.iter().map(count_quad_blank_nodes).sum::<usize>();
     if bnodes_count == 0 {
         assert_eq!(
@@ -83,6 +97,15 @@ fuzz_target!(|data: &[u8]| {
             String::from_utf8_lossy(&serialize_quads(&quads)),
             String::from_utf8_lossy(&serialize_quads(&quads_without_split))
         );
+        if errors.is_empty() {
+            assert_eq!(
+                quads,
+                quads_unchecked,
+                "Validating:\n{}\nUnchecked:\n{}",
+                String::from_utf8_lossy(&serialize_quads(&quads)),
+                String::from_utf8_lossy(&serialize_quads(&quads_unchecked))
+            );
+        }
     } else if bnodes_count <= 4 {
         let mut dataset_with_split = quads.iter().collect::<Dataset>();
         let mut dataset_without_split = quads_without_split.iter().collect::<Dataset>();
@@ -95,6 +118,19 @@ fuzz_target!(|data: &[u8]| {
             String::from_utf8_lossy(&serialize_quads(&quads)),
             String::from_utf8_lossy(&serialize_quads(&quads_without_split))
         );
+        if errors.is_empty() {
+            if errors.is_empty() {
+                let mut dataset_unchecked = quads_unchecked.iter().collect::<Dataset>();
+                dataset_unchecked.canonicalize();
+                assert_eq!(
+                    dataset_with_split,
+                    dataset_unchecked,
+                    "Validating:\n{}\nUnchecked:\n{}",
+                    String::from_utf8_lossy(&serialize_quads(&quads)),
+                    String::from_utf8_lossy(&serialize_quads(&quads_unchecked))
+                );
+            }
+        }
     }
     assert_eq!(errors, errors_without_split);
 

lib/Cargo.toml

@@ -31,7 +31,7 @@ hex = "0.4"
 json-event-parser = "0.2.0-alpha.2"
 md-5 = "0.10"
 oxilangtag = "0.1"
-oxiri = "0.2"
+oxiri = "0.2.3-alpha.1"
 oxrdf = { version = "0.2.0-alpha.1-dev", path = "oxrdf", features = ["rdf-star", "oxsdatatypes"] }
 oxrdfio = { version = "0.1.0-alpha.1-dev", path = "oxrdfio", features = ["rdf-star"] }
 oxsdatatypes = { version = "0.2.0-alpha.1-dev", path="oxsdatatypes" }

lib/benches/store.rs

@@ -24,6 +24,16 @@ fn parse_nt(c: &mut Criterion) {
             }
         })
     });
+    group.bench_function("parse BSBM explore 1000 unchecked", |b| {
+        b.iter(|| {
+            for r in RdfParser::from_format(RdfFormat::NTriples)
+                .unchecked()
+                .parse_read(data.as_slice())
+            {
+                r.unwrap();
+            }
+        })
+    });
 }
 
 fn store_load(c: &mut Criterion) {

lib/oxrdf/Cargo.toml

@@ -21,7 +21,7 @@ rdf-star = []
 [dependencies]
 rand = "0.8"
 oxilangtag = "0.1"
-oxiri = "0.2"
+oxiri = "0.2.3-alpha.1"
 oxsdatatypes = { version = "0.2.0-alpha.1-dev", path="../oxsdatatypes", optional = true }
 
 [lints]

lib/oxrdfio/src/parser.rs

@@ -158,20 +158,16 @@ impl RdfParser {
     /// # Result::<_,Box<dyn std::error::Error>>::Ok(())
     /// ```
     #[inline]
-    pub fn with_base_iri(self, base_iri: impl Into<String>) -> Result<Self, IriParseError> {
-        Ok(Self {
-            inner: match self.inner {
+    pub fn with_base_iri(mut self, base_iri: impl Into<String>) -> Result<Self, IriParseError> {
+        self.inner = match self.inner {
             RdfParserKind::N3(p) => RdfParserKind::N3(p),
             RdfParserKind::NTriples(p) => RdfParserKind::NTriples(p),
             RdfParserKind::NQuads(p) => RdfParserKind::NQuads(p),
             RdfParserKind::RdfXml(p) => RdfParserKind::RdfXml(p.with_base_iri(base_iri)?),
             RdfParserKind::TriG(p) => RdfParserKind::TriG(p.with_base_iri(base_iri)?),
             RdfParserKind::Turtle(p) => RdfParserKind::Turtle(p.with_base_iri(base_iri)?),
-            },
-            default_graph: self.default_graph,
-            without_named_graphs: self.without_named_graphs,
-            rename_blank_nodes: self.rename_blank_nodes,
-        })
+        };
+        Ok(self)
     }
 
     /// Provides the name graph name that should replace the default graph in the returned quads.
@@ -190,13 +186,9 @@ impl RdfParser {
     /// # Result::<_,Box<dyn std::error::Error>>::Ok(())
     /// ```
     #[inline]
-    pub fn with_default_graph(self, default_graph: impl Into<GraphName>) -> Self {
-        Self {
-            inner: self.inner,
-            default_graph: default_graph.into(),
-            without_named_graphs: self.without_named_graphs,
-            rename_blank_nodes: self.rename_blank_nodes,
-        }
+    pub fn with_default_graph(mut self, default_graph: impl Into<GraphName>) -> Self {
+        self.default_graph = default_graph.into();
+        self
     }
 
     /// Sets that the parser must fail if parsing a named graph.
@@ -212,13 +204,9 @@ impl RdfParser {
     /// assert!(parser.parse_read(file.as_bytes()).next().unwrap().is_err());
     /// ```
     #[inline]
-    pub fn without_named_graphs(self) -> Self {
-        Self {
-            inner: self.inner,
-            default_graph: self.default_graph,
-            without_named_graphs: true,
-            rename_blank_nodes: self.rename_blank_nodes,
-        }
+    pub fn without_named_graphs(mut self) -> Self {
+        self.without_named_graphs = true;
+        self
     }
 
     /// Renames the blank nodes ids from the ones set in the serialization to random ids.
@@ -240,13 +228,27 @@ impl RdfParser {
     /// # Result::<_,Box<dyn std::error::Error>>::Ok(())
     /// ```
     #[inline]
-    pub fn rename_blank_nodes(self) -> Self {
-        Self {
-            inner: self.inner,
-            default_graph: self.default_graph,
-            without_named_graphs: self.without_named_graphs,
-            rename_blank_nodes: true,
+    pub fn rename_blank_nodes(mut self) -> Self {
+        self.rename_blank_nodes = true;
+        self
     }
+
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.inner = match self.inner {
+            RdfParserKind::N3(p) => RdfParserKind::N3(p.unchecked()),
+            RdfParserKind::NTriples(p) => RdfParserKind::NTriples(p.unchecked()),
+            RdfParserKind::NQuads(p) => RdfParserKind::NQuads(p.unchecked()),
+            RdfParserKind::RdfXml(p) => RdfParserKind::RdfXml(p.unchecked()),
+            RdfParserKind::TriG(p) => RdfParserKind::TriG(p.unchecked()),
+            RdfParserKind::Turtle(p) => RdfParserKind::Turtle(p.unchecked()),
+        };
+        self
     }
 
     /// Parses from a [`Read`] implementation and returns an iterator of quads.

lib/oxrdfxml/Cargo.toml

@@ -21,7 +21,7 @@ async-tokio = ["dep:tokio", "quick-xml/async-tokio"]
 [dependencies]
 oxrdf = { version = "0.2.0-alpha.1-dev", path = "../oxrdf" }
 oxilangtag = "0.1"
-oxiri = "0.2"
+oxiri = "0.2.3-alpha.1"
 quick-xml = ">=0.29, <0.32"
 tokio = { version = "1.29", optional = true, features = ["io-util"] }
 

lib/oxrdfxml/src/parser.rs

@@ -52,6 +52,7 @@ use tokio::io::{AsyncRead, BufReader as AsyncBufReader};
 #[derive(Default)]
 #[must_use]
 pub struct RdfXmlParser {
+    unchecked: bool,
     base: Option<Iri<String>>,
 }
 
@@ -62,6 +63,17 @@ impl RdfXmlParser {
         Self::default()
     }
 
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.unchecked = true;
+        self
+    }
+
     #[inline]
     pub fn with_base_iri(mut self, base_iri: impl Into<String>) -> Result<Self, IriParseError> {
         self.base = Some(Iri::parse(base_iri.into())?);
@@ -158,6 +170,7 @@ impl RdfXmlParser {
             in_literal_depth: 0,
             known_rdf_id: HashSet::default(),
             is_end: false,
+            unchecked: self.unchecked,
         }
     }
 }
@@ -414,6 +427,7 @@ struct RdfXmlReader<R> {
     in_literal_depth: usize,
     known_rdf_id: HashSet<String>,
     is_end: bool,
+    unchecked: bool,
 }
 
 impl<R> RdfXmlReader<R> {
@@ -551,19 +565,28 @@ impl<R> RdfXmlReader<R> {
             let attribute = attribute.map_err(Error::InvalidAttr)?;
             if attribute.key.as_ref().starts_with(b"xml") {
                 if attribute.key.as_ref() == b"xml:lang" {
-                    let tag = self.convert_attribute(&attribute)?;
-                    language = Some(
+                    let tag = self.convert_attribute(&attribute)?.to_ascii_lowercase();
+                    language = Some(if self.unchecked {
+                        tag
+                    } else {
                         LanguageTag::parse(tag.to_ascii_lowercase())
                             .map_err(|error| SyntaxError {
                                 inner: SyntaxErrorKind::InvalidLanguageTag { tag, error },
                             })?
-                            .into_inner(),
-                    );
+                            .into_inner()
+                    });
                 } else if attribute.key.as_ref() == b"xml:base" {
                     let iri = self.convert_attribute(&attribute)?;
-                    base_iri = Some(Iri::parse(iri.clone()).map_err(|error| SyntaxError {
+                    base_iri = Some(
+                        if self.unchecked {
+                            Iri::parse_unchecked(iri.clone())
+                        } else {
+                            Iri::parse(iri.clone())
+                        }
+                        .map_err(|error| SyntaxError {
                             inner: SyntaxErrorKind::InvalidIri { iri, error },
-                    })?)
+                        })?,
+                    )
                 } else {
                     // We ignore other xml attributes
                 }
@@ -622,12 +645,7 @@ impl<R> RdfXmlReader<R> {
                     .into());
                 } else {
                     property_attrs.push((
-                        NamedNode::new(attribute_url.clone()).map_err(|error| SyntaxError {
-                            inner: SyntaxErrorKind::InvalidIri {
-                                iri: attribute_url,
-                                error,
-                            },
-                        })?,
+                        self.parse_iri(attribute_url)?,
                         self.convert_attribute(&attribute)?,
                     ));
                 }
@@ -637,7 +655,7 @@ impl<R> RdfXmlReader<R> {
         //Parsing with the base URI
         let id_attr = match id_attr {
             Some(iri) => {
-                let iri = resolve(&base_iri, iri)?;
+                let iri = self.resolve_iri(&base_iri, iri)?;
                 if self.known_rdf_id.contains(iri.as_str()) {
                     return Err(SyntaxError::msg(format!(
                         "{} has already been used as rdf:ID value",
@@ -701,12 +719,7 @@ impl<R> RdfXmlReader<R> {
                     .into());
                 } else {
                     Self::build_node_elt(
-                        NamedNode::new(tag_name.clone()).map_err(|error| SyntaxError {
-                            inner: SyntaxErrorKind::InvalidIri {
-                                iri: tag_name,
-                                error,
-                            },
-                        })?,
+                        self.parse_iri(tag_name)?,
                         base_iri,
                         language,
                         id_attr,
@@ -727,12 +740,7 @@ impl<R> RdfXmlReader<R> {
                     .into());
                 }
                 Self::build_node_elt(
-                    NamedNode::new(tag_name.clone()).map_err(|error| SyntaxError {
-                        inner: SyntaxErrorKind::InvalidIri {
-                            iri: tag_name,
-                            error,
-                        },
-                    })?,
+                    self.parse_iri(tag_name)?,
                     base_iri,
                     language,
                     id_attr,
@@ -766,12 +774,7 @@ impl<R> RdfXmlReader<R> {
                     ))
                     .into());
                 } else {
-                    NamedNode::new(tag_name.clone()).map_err(|error| SyntaxError {
-                        inner: SyntaxErrorKind::InvalidIri {
-                            iri: tag_name,
-                            error,
-                        },
-                    })?
+                    self.parse_iri(tag_name)?
                 };
                 match parse_type {
                     RdfXmlParseType::Default => {
@@ -1156,32 +1159,51 @@ impl<R> RdfXmlReader<R> {
         base_iri: &Option<Iri<String>>,
         attribute: &Attribute<'_>,
     ) -> Result<NamedNode, ParseError> {
-        Ok(resolve(base_iri, self.convert_attribute(attribute)?)?)
+        Ok(self.resolve_iri(base_iri, self.convert_attribute(attribute)?)?)
     }
 
-    fn resolve_entity(&self, e: &str) -> Option<&str> {
-        self.custom_entities.get(e).map(String::as_str)
-    }
-}
-
-fn resolve(base_iri: &Option<Iri<String>>, relative_iri: String) -> Result<NamedNode, SyntaxError> {
+    fn resolve_iri(
+        &self,
+        base_iri: &Option<Iri<String>>,
+        relative_iri: String,
+    ) -> Result<NamedNode, SyntaxError> {
         if let Some(base_iri) = base_iri {
-        Ok(base_iri
-            .resolve(&relative_iri)
+            Ok(NamedNode::new_unchecked(
+                if self.unchecked {
+                    base_iri.resolve_unchecked(&relative_iri)
+                } else {
+                    base_iri.resolve(&relative_iri)
+                }
                 .map_err(|error| SyntaxError {
                     inner: SyntaxErrorKind::InvalidIri {
                         iri: relative_iri,
                         error,
                     },
                 })?
-            .into())
+                .into_inner(),
+            ))
+        } else {
+            self.parse_iri(relative_iri)
+        }
+    }
+
+    fn parse_iri(&self, relative_iri: String) -> Result<NamedNode, SyntaxError> {
+        Ok(NamedNode::new_unchecked(if self.unchecked {
+            relative_iri
         } else {
-        NamedNode::new(relative_iri.clone()).map_err(|error| SyntaxError {
+            Iri::parse(relative_iri.clone())
+                .map_err(|error| SyntaxError {
                     inner: SyntaxErrorKind::InvalidIri {
                         iri: relative_iri,
                         error,
                     },
-        })
+                })?
+                .into_inner()
+        }))
+    }
+
+    fn resolve_entity(&self, e: &str) -> Option<&str> {
+        self.custom_entities.get(e).map(String::as_str)
     }
 }
 

lib/oxttl/Cargo.toml

@@ -22,7 +22,7 @@ async-tokio = ["dep:tokio"]
 [dependencies]
 memchr = "2.5"
 oxrdf = { version = "0.2.0-alpha.1-dev", path = "../oxrdf" }
-oxiri = "0.2"
+oxiri = "0.2.3-alpha.1"
 oxilangtag = "0.1"
 tokio = { version = "1.29", optional = true, features = ["io-util"] }
 

lib/oxttl/src/lexer.rs

@@ -6,12 +6,12 @@ use oxrdf::NamedNode;
 use std::borrow::Cow;
 use std::cmp::min;
 use std::collections::HashMap;
-use std::ops::{Range, RangeInclusive};
+use std::ops::Range;
 use std::str;
 
 #[derive(Debug, PartialEq, Eq)]
 pub enum N3Token<'a> {
-    IriRef(Iri<String>),
+    IriRef(String),
     PrefixedName {
         prefix: &'a str,
         local: Cow<'a, str>,
@@ -42,6 +42,7 @@ pub struct N3LexerOptions {
 
 pub struct N3Lexer {
     mode: N3LexerMode,
+    unchecked: bool,
 }
 
 // TODO: there are a lot of 'None' (missing data) returned even if the stream is ending!!!
@@ -61,7 +62,7 @@ impl TokenRecognizer for N3Lexer {
             b'<' => match *data.get(1)? {
                 b'<' => Some((2, Ok(N3Token::Punctuation("<<")))),
                 b'=' if self.mode == N3LexerMode::N3 => {
-                    if let Some((consumed, result)) = Self::recognize_iri(data, options) {
+                    if let Some((consumed, result)) = self.recognize_iri(data, options) {
                         Some(if let Ok(result) = result {
                             (consumed, Ok(result))
                         } else {
@@ -74,7 +75,7 @@ impl TokenRecognizer for N3Lexer {
                     }
                 }
                 b'-' if self.mode == N3LexerMode::N3 => {
-                    if let Some((consumed, result)) = Self::recognize_iri(data, options) {
+                    if let Some((consumed, result)) = self.recognize_iri(data, options) {
                         Some(if let Ok(result) = result {
                             (consumed, Ok(result))
                         } else {
@@ -86,7 +87,7 @@ impl TokenRecognizer for N3Lexer {
                         None
                     }
                 }
-                _ => Self::recognize_iri(data, options),
+                _ => self.recognize_iri(data, options),
             },
             b'>' => {
                 if *data.get(1)? == b'>' {
@@ -119,7 +120,7 @@ impl TokenRecognizer for N3Lexer {
                     Self::recognize_string(data, b'\'')
                 }
             }
-            b'@' => Self::recognize_lang_tag(data),
+            b'@' => self.recognize_lang_tag(data),
             b'.' => match data.get(1) {
                 Some(b'0'..=b'9') => Self::recognize_number(data),
                 Some(_) => Some((1, Ok(N3Token::Punctuation(".")))),
@@ -162,18 +163,19 @@ impl TokenRecognizer for N3Lexer {
                 }
             }
             b'0'..=b'9' | b'+' | b'-' => Self::recognize_number(data),
-            b'?' => Self::recognize_variable(data, is_ending),
-            _ => Self::recognize_pname_or_keyword(data, is_ending),
+            b'?' => self.recognize_variable(data, is_ending),
+            _ => self.recognize_pname_or_keyword(data, is_ending),
         }
     }
 }
 
 impl N3Lexer {
-    pub fn new(mode: N3LexerMode) -> Self {
-        Self { mode }
+    pub fn new(mode: N3LexerMode, unchecked: bool) -> Self {
+        Self { mode, unchecked }
     }
 
     fn recognize_iri(
+        &self,
         data: &[u8],
         options: &N3LexerOptions,
     ) -> Option<(usize, Result<N3Token<'static>, TokenRecognizerError>)> {
@@ -186,7 +188,8 @@ impl N3Lexer {
             i += end;
             match data[i] {
                 b'>' => {
-                    return Some((i + 1, Self::parse_iri(string, 0..=i, options)));
+                    #[allow(clippy::range_plus_one)]
+                    return Some((i + 1, self.parse_iri(string, 0..i + 1, options)));
                 }
                 b'\\' => {
                     let (additional, c) = Self::recognize_escape(&data[i..], i, false)?;
@@ -205,29 +208,36 @@ impl N3Lexer {
     }
 
     fn parse_iri(
+        &self,
         iri: Vec<u8>,
-        position: RangeInclusive<usize>,
+        position: Range<usize>,
         options: &N3LexerOptions,
     ) -> Result<N3Token<'static>, TokenRecognizerError> {
-        let iri = String::from_utf8(iri).map_err(|e| {
-            (
-                position.clone(),
-                format!("The IRI contains invalid UTF-8 characters: {e}"),
-            )
-        })?;
-        let iri = if let Some(base_iri) = options.base_iri.as_ref() {
+        let iri = string_from_utf8(iri, position.clone())?;
+        Ok(N3Token::IriRef(
+            if let Some(base_iri) = options.base_iri.as_ref() {
+                if self.unchecked {
+                    base_iri.resolve_unchecked(&iri)
+                } else {
                     base_iri.resolve(&iri)
+                }
+                .map_err(|e| (position, e.to_string()))?
+                .into_inner()
+            } else if self.unchecked {
+                iri
             } else {
                 Iri::parse(iri)
-        }
-        .map_err(|e| (position, e.to_string()))?;
-        Ok(N3Token::IriRef(iri))
+                    .map_err(|e| (position, e.to_string()))?
+                    .into_inner()
+            },
+        ))
     }
 
-    fn recognize_pname_or_keyword(
-        data: &[u8],
+    fn recognize_pname_or_keyword<'a>(
+        &self,
+        data: &'a [u8],
         is_ending: bool,
-    ) -> Option<(usize, Result<N3Token<'_>, TokenRecognizerError>)> {
+    ) -> Option<(usize, Result<N3Token<'a>, TokenRecognizerError>)> {
         // [139s] 	PNAME_NS 	::= 	PN_PREFIX? ':'
         // [140s] 	PNAME_LN 	::= 	PNAME_NS PN_LOCAL
 
@@ -303,7 +313,8 @@ impl N3Lexer {
             ));
         }
 
-        let (consumed, pn_local_result) = Self::recognize_optional_pn_local(&data[i..], is_ending)?;
+        let (consumed, pn_local_result) =
+            self.recognize_optional_pn_local(&data[i..], is_ending)?;
         Some((
             consumed + i,
             pn_local_result.map(|(local, might_be_invalid_iri)| N3Token::PrefixedName {
@@ -314,12 +325,13 @@ impl N3Lexer {
         ))
     }
 
-    fn recognize_variable(
-        data: &[u8],
+    fn recognize_variable<'a>(
+        &self,
+        data: &'a [u8],
         is_ending: bool,
-    ) -> Option<(usize, Result<N3Token<'_>, TokenRecognizerError>)> {
+    ) -> Option<(usize, Result<N3Token<'a>, TokenRecognizerError>)> {
         // [36] 	QUICK_VAR_NAME 	::= 	"?" PN_LOCAL
-        let (consumed, result) = Self::recognize_optional_pn_local(&data[1..], is_ending)?;
+        let (consumed, result) = self.recognize_optional_pn_local(&data[1..], is_ending)?;
         Some((
             consumed + 1,
             result.and_then(|(name, _)| {
@@ -332,10 +344,11 @@ impl N3Lexer {
         ))
     }
 
-    fn recognize_optional_pn_local(
-        data: &[u8],
+    fn recognize_optional_pn_local<'a>(
+        &self,
+        data: &'a [u8],
         is_ending: bool,
-    ) -> Option<(usize, Result<(Cow<'_, str>, bool), TokenRecognizerError>)> {
+    ) -> Option<(usize, Result<(Cow<'a, str>, bool), TokenRecognizerError>)> {
         // [168s] 	PN_LOCAL 	::= 	(PN_CHARS_U | ':' | [0-9] | PLX) ((PN_CHARS | '.' | ':' | PLX)* (PN_CHARS | ':' | PLX))?
         let mut i = 0;
         let mut buffer = None; // Buffer if there are some escaped characters
@@ -359,7 +372,8 @@ impl N3Lexer {
                         } else if c == '\\' {
                             i += 1;
                             let a = char::from(*data.get(i)?);
-                            if matches!(
+                            if self.unchecked
+                                || matches!(
                                     a,
                                     '_' | '~'
                                         | '.'
@@ -375,7 +389,8 @@ impl N3Lexer {
                                         | ','
                                         | ';'
                                         | '='
-                            ) {
+                                )
+                            {
                                 // ok to escape
                             } else if matches!(a, '/' | '?' | '#' | '@' | '%') {
                                 // ok to escape but requires IRI validation
@@ -406,12 +421,18 @@ impl N3Lexer {
                             {
                                 return Some((0, Ok((Cow::Borrowed(""), false))));
                             }
+                            if !self.unchecked {
                                 might_be_invalid_iri |=
-                                Self::is_possible_pn_chars_base_but_not_valid_iri(c) || c == ':';
+                                    Self::is_possible_pn_chars_base_but_not_valid_iri(c)
+                                        || c == ':';
+                            }
                             i += consumed;
                         } else if Self::is_possible_pn_chars(c) || c == ':' || c == '.' {
+                            if !self.unchecked {
                                 might_be_invalid_iri |=
-                                Self::is_possible_pn_chars_base_but_not_valid_iri(c) || c == ':';
+                                    Self::is_possible_pn_chars_base_but_not_valid_iri(c)
+                                        || c == ':';
+                            }
                             i += consumed;
                         } else {
                             let buffer = if let Some(mut buffer) = buffer {
@@ -518,9 +539,10 @@ impl N3Lexer {
         }
     }
 
-    fn recognize_lang_tag(
-        data: &[u8],
-    ) -> Option<(usize, Result<N3Token<'_>, TokenRecognizerError>)> {
+    fn recognize_lang_tag<'a>(
+        &self,
+        data: &'a [u8],
+    ) -> Option<(usize, Result<N3Token<'a>, TokenRecognizerError>)> {
         // [144s] 	LANGTAG 	::= 	'@' [a-zA-Z]+ ('-' [a-zA-Z0-9]+)*
         let mut is_last_block_empty = true;
         for (i, c) in data[1..].iter().enumerate() {
@@ -532,25 +554,29 @@ impl N3Lexer {
                     Err((1..2, "A language code should always start with a letter").into()),
                 ));
             } else if is_last_block_empty {
-                return Some((i, Self::parse_lang_tag(&data[1..i], 1..i - 1)));
+                return Some((i, self.parse_lang_tag(&data[1..i], 1..i - 1)));
             } else if *c == b'-' {
                 is_last_block_empty = true;
             } else {
-                return Some((i + 1, Self::parse_lang_tag(&data[1..=i], 1..i)));
+                return Some((i + 1, self.parse_lang_tag(&data[1..=i], 1..i)));
             }
         }
         None
     }
 
-    fn parse_lang_tag(
-        lang_tag: &[u8],
+    fn parse_lang_tag<'a>(
+        &self,
+        lang_tag: &'a [u8],
         position: Range<usize>,
-    ) -> Result<N3Token<'_>, TokenRecognizerError> {
-        Ok(N3Token::LangTag(
-            LanguageTag::parse(str_from_utf8(lang_tag, position.clone())?)
+    ) -> Result<N3Token<'a>, TokenRecognizerError> {
+        let lang_tag = str_from_utf8(lang_tag, position.clone())?;
+        Ok(N3Token::LangTag(if self.unchecked {
+            lang_tag
+        } else {
+            LanguageTag::parse(lang_tag)
                 .map_err(|e| (position.clone(), e.to_string()))?
-                .into_inner(),
-        ))
+                .into_inner()
+        }))
     }
 
     fn recognize_string(
@@ -933,3 +959,14 @@ fn str_from_utf8(data: &[u8], range: Range<usize>) -> Result<&str, TokenRecogniz
             .into()
     })
 }
+
+fn string_from_utf8(data: Vec<u8>, range: Range<usize>) -> Result<String, TokenRecognizerError> {
+    String::from_utf8(data).map_err(|e| {
+        (
+            range.start + e.utf8_error().valid_up_to()
+                ..min(range.end, range.start + e.utf8_error().valid_up_to() + 4),
+            format!("Invalid UTF-8: {e}"),
+        )
+            .into()
+    })
+}

lib/oxttl/src/line_formats.rs

@@ -63,7 +63,7 @@ impl RuleRecognizer for NQuadsRecognizer {
                 NQuadsState::ExpectSubject => match token {
                     N3Token::IriRef(s) => {
                         self.subjects
-                            .push(NamedNode::from(s).into());
+                            .push(NamedNode::new_unchecked(s).into());
                         self.stack.push(NQuadsState::ExpectPredicate);
                         self
                     }
@@ -86,7 +86,7 @@ impl RuleRecognizer for NQuadsRecognizer {
                 NQuadsState::ExpectPredicate => match token {
                     N3Token::IriRef(p) => {
                         self.predicates
-                            .push(p.into());
+                            .push(NamedNode::new_unchecked(p));
                         self.stack.push(NQuadsState::ExpectedObject);
                         self
                     }
@@ -98,7 +98,7 @@ impl RuleRecognizer for NQuadsRecognizer {
                 NQuadsState::ExpectedObject => match token {
                     N3Token::IriRef(o) => {
                         self.objects
-                            .push(NamedNode::from(o).into());
+                            .push(NamedNode::new_unchecked(o).into());
                         self.stack
                             .push(NQuadsState::ExpectPossibleGraphOrEndOfQuotedTriple);
                         self
@@ -155,7 +155,7 @@ impl RuleRecognizer for NQuadsRecognizer {
                         self.objects.push(
                             Literal::new_typed_literal(
                                 value,
-                                d
+                                NamedNode::new_unchecked(d)
                             )
                             .into(),
                         );
@@ -171,7 +171,7 @@ impl RuleRecognizer for NQuadsRecognizer {
                             N3Token::IriRef(g) if context.with_graph_name => {
                                 self.emit_quad(
                                     results,
-                                    NamedNode::from(g).into(),
+                                    NamedNode::new_unchecked(g).into(),
                                 );
                                 self.stack.push(NQuadsState::ExpectDot);
                                 self
@@ -264,10 +264,11 @@ impl NQuadsRecognizer {
     pub fn new_parser(
         with_graph_name: bool,
         #[cfg(feature = "rdf-star")] with_quoted_triples: bool,
+        unchecked: bool,
     ) -> Parser<Self> {
         Parser::new(
             Lexer::new(
-                N3Lexer::new(N3LexerMode::NTriples),
+                N3Lexer::new(N3LexerMode::NTriples, unchecked),
                 MIN_BUFFER_SIZE,
                 MAX_BUFFER_SIZE,
                 true,

lib/oxttl/src/n3.rs

@@ -206,6 +206,7 @@ impl From<Quad> for N3Quad {
 #[derive(Default)]
 #[must_use]
 pub struct N3Parser {
+    unchecked: bool,
     base: Option<Iri<String>>,
     prefixes: HashMap<String, Iri<String>>,
 }
@@ -217,6 +218,17 @@ impl N3Parser {
         Self::default()
     }
 
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.unchecked = true;
+        self
+    }
+
     #[inline]
     pub fn with_base_iri(mut self, base_iri: impl Into<String>) -> Result<Self, IriParseError> {
         self.base = Some(Iri::parse(base_iri.into())?);
@@ -345,7 +357,7 @@ impl N3Parser {
     /// ```
     pub fn parse(self) -> LowLevelN3Reader {
         LowLevelN3Reader {
-            parser: N3Recognizer::new_parser(self.base, self.prefixes),
+            parser: N3Recognizer::new_parser(self.unchecked, self.base, self.prefixes),
         }
     }
 }
@@ -665,9 +677,14 @@ impl RuleRecognizer for N3Recognizer {
                 }
                 N3State::BaseExpectIri => return match token {
                     N3Token::IriRef(iri) => {
+                        match Iri::parse_unchecked(iri) {
+                            Ok(iri) => {
                                 context.lexer_options.base_iri = Some(iri);
                                 self
                             }
+                            Err(e) => self.error(errors, format!("Invalid base IRI: {e}"))
+                        }
+                    }
                     _ => self.error(errors, "The BASE keyword should be followed by an IRI"),
                 },
                 N3State::PrefixExpectPrefix => return match token {
@@ -681,9 +698,14 @@ impl RuleRecognizer for N3Recognizer {
                 },
                 N3State::PrefixExpectIri { name } => return match token {
                     N3Token::IriRef(iri) => {
+                        match Iri::parse_unchecked(iri) {
+                            Ok(iri) => {
                                 context.prefixes.insert(name, iri);
                                 self
                             }
+                            Err(e) => self.error(errors, format!("Invalid prefix IRI: {e}"))
+                        }
+                    }
                     _ => self.error(errors, "The PREFIX declaration should be followed by a prefix and its value as an IRI"),
                 },
                 // [9] 	triples 	::= 	subject predicateObjectList?
@@ -843,7 +865,7 @@ impl RuleRecognizer for N3Recognizer {
                 N3State::PathItem => {
                     return match token {
                         N3Token::IriRef(iri) => {
-                            self.terms.push(NamedNode::from(iri).into());
+                            self.terms.push(NamedNode::new_unchecked(iri).into());
                             self
                         }
                         N3Token::PrefixedName { prefix, local, might_be_invalid_iri } => match resolve_local_name(prefix, &local, might_be_invalid_iri, &context.prefixes) {
@@ -925,7 +947,7 @@ impl RuleRecognizer for N3Recognizer {
                 }
                 N3State::IriPropertyList => return match token {
                     N3Token::IriRef(id) => {
-                        self.terms.push(NamedNode::from(id).into());
+                        self.terms.push(NamedNode::new_unchecked(id).into());
                         self.stack.push(N3State::PropertyListEnd);
                         self.stack.push(N3State::PredicateObjectList);
                         self
@@ -999,7 +1021,7 @@ impl RuleRecognizer for N3Recognizer {
                 N3State::LiteralExpectDatatype { value } => {
                     match token {
                         N3Token::IriRef(datatype) => {
-                            self.terms.push(Literal::new_typed_literal(value, datatype).into());
+                            self.terms.push(Literal::new_typed_literal(value, NamedNode::new_unchecked(datatype)).into());
                             return self;
                         }
                         N3Token::PrefixedName { prefix, local, might_be_invalid_iri } => match resolve_local_name(prefix, &local, might_be_invalid_iri, &context.prefixes) {
@@ -1096,12 +1118,13 @@ impl RuleRecognizer for N3Recognizer {
 
 impl N3Recognizer {
     pub fn new_parser(
+        unchecked: bool,
         base_iri: Option<Iri<String>>,
         prefixes: HashMap<String, Iri<String>>,
     ) -> Parser<Self> {
         Parser::new(
             Lexer::new(
-                N3Lexer::new(N3LexerMode::N3),
+                N3Lexer::new(N3LexerMode::N3, unchecked),
                 MIN_BUFFER_SIZE,
                 MAX_BUFFER_SIZE,
                 true,

lib/oxttl/src/nquads.rs

@@ -37,6 +37,7 @@ use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
 #[derive(Default)]
 #[must_use]
 pub struct NQuadsParser {
+    unchecked: bool,
     #[cfg(feature = "rdf-star")]
     with_quoted_triples: bool,
 }
@@ -48,6 +49,17 @@ impl NQuadsParser {
         Self::default()
     }
 
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.unchecked = true;
+        self
+    }
+
     /// Enables [N-Quads-star](https://w3c.github.io/rdf-star/cg-spec/2021-12-17.html#n-quads-star).
     #[cfg(feature = "rdf-star")]
     #[inline]
@@ -165,6 +177,7 @@ impl NQuadsParser {
                 true,
                 #[cfg(feature = "rdf-star")]
                 self.with_quoted_triples,
+                self.unchecked,
             ),
         }
     }

lib/oxttl/src/ntriples.rs

@@ -38,6 +38,7 @@ use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
 #[derive(Default)]
 #[must_use]
 pub struct NTriplesParser {
+    unchecked: bool,
     #[cfg(feature = "rdf-star")]
     with_quoted_triples: bool,
 }
@@ -49,6 +50,17 @@ impl NTriplesParser {
         Self::default()
     }
 
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.    ///
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.unchecked = true;
+        self
+    }
+
     /// Enables [N-Triples-star](https://w3c.github.io/rdf-star/cg-spec/2021-12-17.html#n-triples-star).
     #[cfg(feature = "rdf-star")]
     #[inline]
@@ -166,6 +178,7 @@ impl NTriplesParser {
                 false,
                 #[cfg(feature = "rdf-star")]
                 self.with_quoted_triples,
+                self.unchecked,
             ),
         }
     }
@@ -542,3 +555,26 @@ impl LowLevelNTriplesWriter {
         writeln!(write, "{} .", t.into())
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use oxrdf::{Literal, NamedNode};
+
+    #[test]
+    fn unchecked_parsing() {
+        let triples = NTriplesParser::new()
+            .unchecked()
+            .parse_read("<foo> <bar> \"baz\"@toolonglangtag .".as_bytes())
+            .collect::<Result<Vec<_>, _>>()
+            .unwrap();
+        assert_eq!(
+            triples,
+            [Triple::new(
+                NamedNode::new_unchecked("foo"),
+                NamedNode::new_unchecked("bar"),
+                Literal::new_language_tagged_literal_unchecked("baz", "toolonglangtag"),
+            )]
+        )
+    }
+}

lib/oxttl/src/terse.rs

@@ -107,9 +107,14 @@ impl RuleRecognizer for TriGRecognizer {
                 },
                 TriGState::BaseExpectIri => match token {
                     N3Token::IriRef(iri) => {
+                        match Iri::parse_unchecked(iri) {
+                            Ok(iri) => {
                                 context.lexer_options.base_iri = Some(iri);
                                 self
                             }
+                            Err(e) => self.error(errors, format!("Invalid base IRI: {e}"))
+                        }
+                    }
                     _ => self.error(errors, "The BASE keyword should be followed by an IRI"),
                 },
                 TriGState::PrefixExpectPrefix => match token {
@@ -123,9 +128,13 @@ impl RuleRecognizer for TriGRecognizer {
                 },
                 TriGState::PrefixExpectIri { name } => match token {
                     N3Token::IriRef(iri) => {
+                        match Iri::parse_unchecked(iri) {
+                            Ok(iri) => {
                                 context.prefixes.insert(name, iri);
                                 self
                             }
+                            Err(e) => self.error(errors, format!("Invalid prefix IRI: {e}"))
+                        }                    }
                     _ => self.error(errors, "The PREFIX declaration should be followed by a prefix and its value as an IRI"),
                 },
                 // [3g] 	triplesOrGraph 	::= 	labelOrSubject ( wrappedGraph | predicateObjectList '.' ) | quotedTriple predicateObjectList '.'
@@ -133,7 +142,7 @@ impl RuleRecognizer for TriGRecognizer {
                 TriGState::TriplesOrGraph => match token {
                     N3Token::IriRef(iri) => {
                         self.stack.push(TriGState::WrappedGraphOrPredicateObjectList {
-                            term: NamedNode::from(iri).into()
+                            term: NamedNode::new_unchecked(iri).into()
                         });
                         self
                     }
@@ -291,7 +300,7 @@ impl RuleRecognizer for TriGRecognizer {
                         self
                     }
                     N3Token::IriRef(iri) => {
-                        self.cur_subject.push(NamedNode::from(iri).into());
+                        self.cur_subject.push(NamedNode::new_unchecked(iri).into());
                         self.stack.push(TriGState::PredicateObjectList);
                         self
                     }
@@ -337,7 +346,7 @@ impl RuleRecognizer for TriGRecognizer {
                 // [7g] 	labelOrSubject 	::= 	iri | BlankNode
                 TriGState::GraphName => match token {
                     N3Token::IriRef(iri) => {
-                        self.cur_graph = NamedNode::from(iri).into();
+                        self.cur_graph = NamedNode::new_unchecked(iri).into();
                         self
                     }
                     N3Token::PrefixedName { prefix, local, might_be_invalid_iri } => match resolve_local_name(prefix, &local, might_be_invalid_iri, &context.prefixes) {
@@ -451,7 +460,7 @@ impl RuleRecognizer for TriGRecognizer {
                         self
                     }
                     N3Token::IriRef(iri) => {
-                        self.cur_predicate.push(NamedNode::from(iri));
+                        self.cur_predicate.push(NamedNode::new_unchecked(iri));
                         self
                     }
                     N3Token::PrefixedName { prefix, local, might_be_invalid_iri } => match resolve_local_name(prefix, &local, might_be_invalid_iri, &context.prefixes) {
@@ -479,7 +488,7 @@ impl RuleRecognizer for TriGRecognizer {
                 // [137s] 	BlankNode 	::= 	BLANK_NODE_LABEL | ANON
                 TriGState::Object => match token {
                     N3Token::IriRef(iri) => {
-                        self.cur_object.push(NamedNode::from(iri).into());
+                        self.cur_object.push(NamedNode::new_unchecked(iri).into());
                         self.emit_quad(results);
                         self
                     }
@@ -626,7 +635,7 @@ impl RuleRecognizer for TriGRecognizer {
                 TriGState::LiteralExpectDatatype { value, emit } => {
                     match token {
                         N3Token::IriRef(datatype) => {
-                            self.cur_object.push(Literal::new_typed_literal(value, datatype).into());
+                            self.cur_object.push(Literal::new_typed_literal(value, NamedNode::new_unchecked(datatype)).into());
                             if emit {
                                 self.emit_quad(results);
                             }
@@ -688,7 +697,7 @@ impl RuleRecognizer for TriGRecognizer {
                         self
                     }
                     N3Token::IriRef(iri) => {
-                        self.cur_subject.push(NamedNode::from(iri).into());
+                        self.cur_subject.push(NamedNode::new_unchecked(iri).into());
                         self
                     }
                     N3Token::PrefixedName { prefix, local, might_be_invalid_iri } => match resolve_local_name(prefix, &local, might_be_invalid_iri, &context.prefixes) {
@@ -720,7 +729,7 @@ impl RuleRecognizer for TriGRecognizer {
                         self
                     }
                     N3Token::IriRef(iri) => {
-                        self.cur_object.push(NamedNode::from(iri).into());
+                        self.cur_object.push(NamedNode::new_unchecked(iri).into());
                         self
                     }
                     N3Token::PrefixedName { prefix, local, might_be_invalid_iri } => match resolve_local_name(prefix, &local, might_be_invalid_iri, &context.prefixes) {
@@ -823,12 +832,13 @@ impl TriGRecognizer {
     pub fn new_parser(
         with_graph_name: bool,
         #[cfg(feature = "rdf-star")] with_quoted_triples: bool,
+        unchecked: bool,
         base_iri: Option<Iri<String>>,
         prefixes: HashMap<String, Iri<String>>,
     ) -> Parser<Self> {
         Parser::new(
             Lexer::new(
-                N3Lexer::new(N3LexerMode::Turtle),
+                N3Lexer::new(N3LexerMode::Turtle, unchecked),
                 MIN_BUFFER_SIZE,
                 MAX_BUFFER_SIZE,
                 true,

lib/oxttl/src/trig.rs

@@ -42,6 +42,7 @@ use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
 #[derive(Default)]
 #[must_use]
 pub struct TriGParser {
+    unchecked: bool,
     base: Option<Iri<String>>,
     prefixes: HashMap<String, Iri<String>>,
     #[cfg(feature = "rdf-star")]
@@ -55,6 +56,17 @@ impl TriGParser {
         Self::default()
     }
 
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.unchecked = true;
+        self
+    }
+
     #[inline]
     pub fn with_base_iri(mut self, base_iri: impl Into<String>) -> Result<Self, IriParseError> {
         self.base = Some(Iri::parse(base_iri.into())?);
@@ -192,6 +204,7 @@ impl TriGParser {
                 true,
                 #[cfg(feature = "rdf-star")]
                 self.with_quoted_triples,
+                self.unchecked,
                 self.base,
                 self.prefixes,
             ),

lib/oxttl/src/turtle.rs

@@ -44,6 +44,7 @@ use tokio::io::{AsyncRead, AsyncWrite};
 #[derive(Default)]
 #[must_use]
 pub struct TurtleParser {
+    unchecked: bool,
     base: Option<Iri<String>>,
     prefixes: HashMap<String, Iri<String>>,
     #[cfg(feature = "rdf-star")]
@@ -57,6 +58,17 @@ impl TurtleParser {
         Self::default()
     }
 
+    /// Assumes the file is valid to make parsing faster.
+    ///
+    /// It will skip some validations.
+    ///
+    /// Note that if the file is actually not valid, then broken RDF might be emitted by the parser.
+    #[inline]
+    pub fn unchecked(mut self) -> Self {
+        self.unchecked = true;
+        self
+    }
+
     #[inline]
     pub fn with_base_iri(mut self, base_iri: impl Into<String>) -> Result<Self, IriParseError> {
         self.base = Some(Iri::parse(base_iri.into())?);
@@ -194,6 +206,7 @@ impl TurtleParser {
                 false,
                 #[cfg(feature = "rdf-star")]
                 self.with_quoted_triples,
+                self.unchecked,
                 self.base,
                 self.prefixes,
             ),

lib/spargebra/Cargo.toml

@@ -23,7 +23,7 @@ sep-0006 = []
 [dependencies]
 peg = "0.8"
 rand = "0.8"
-oxiri = "0.2"
+oxiri = "0.2.3-alpha.1"
 oxilangtag = "0.1"
 oxrdf = { version = "0.2.0-alpha.1-dev", path="../oxrdf" }
 

lib/src/store.rs

@@ -1053,7 +1053,6 @@ impl<'a> Transaction<'a> {
     /// Retrieves quads with a filter on each quad component.
     ///
     /// Usage example:
-    /// Usage example:
     /// ```
     /// use oxigraph::store::{StorageError, Store};
     /// use oxigraph::model::*;
@@ -1601,19 +1600,22 @@ impl BulkLoader {
     ///
     /// <div class="warning">This method is optimized for speed. See [the struct](BulkLoader) documentation for more details.</div>
     ///
-    /// Usage example:
+    /// To get better speed on valid datasets, consider enabling [`RdfParser::unchecked`] option to skip some validations.
+    ///
     /// Usage example:
     /// ```
     /// use oxigraph::store::Store;
-    /// use oxigraph::io::RdfFormat;
+    /// use oxigraph::io::{RdfParser, RdfFormat};
     /// use oxigraph::model::*;
-    /// use oxrdfio::RdfParser;
     ///
     /// let store = Store::new()?;
     ///
     /// // insert a dataset file (former load_dataset method)
     /// let file = b"<http://example.com> <http://example.com> <http://example.com> <http://example.com/g> .";
-    /// store.bulk_loader().load_from_read(RdfFormat::NQuads, file.as_ref())?;
+    /// store.bulk_loader().load_from_read(
+    ///     RdfParser::from_format(RdfFormat::NQuads).unchecked(), // we inject a custom parser with options
+    ///     file.as_ref()
+    /// )?;
     ///
     /// // insert a graph file (former load_graph method)
     /// let file = b"<> <> <> .";

lints/test_debian_compatibility.py

@@ -5,7 +5,7 @@ from urllib.request import urlopen
 
 TARGET_DEBIAN_VERSIONS = ["sid"]
 IGNORE_PACKAGES = {"oxigraph-js", "oxigraph-testsuite", "pyoxigraph", "sparql-smith"}
-ALLOWED_MISSING_PACKAGES = {"codspeed-criterion-compat", "escargot", "json-event-parser", "oxhttp", "quick-xml"}
+ALLOWED_MISSING_PACKAGES = {"codspeed-criterion-compat", "escargot", "json-event-parser", "oxhttp", "oxiri", "quick-xml"}
 
 base_path = Path(__file__).parent.parent