Fuzzer: ensure that NQuad/TriG segmentation does not affect results

fuzz/Cargo.toml

@@ -1,7 +1,6 @@
 [package]
 name = "oxigraph-fuzz"
 version = "0.0.0"
-authors = ["Automatically generated"]
 publish = false
 edition = "2021"
 
@@ -12,6 +11,7 @@ cargo-fuzz = true
 anyhow = "1"
 lazy_static = "1"
 libfuzzer-sys = "0.4"
+oxrdf = { path = "../lib/oxrdf", features = ["rdf-star"] }
 oxttl = { path = "../lib/oxttl", features = ["rdf-star"] }
 oxrdfxml = { path = "../lib/oxrdfxml" }
 spargebra = { path = "../lib/spargebra", features = ["rdf-star", "sep-0006"] }

fuzz/fuzz_targets/nquads.rs

@@ -1,27 +1,45 @@
 #![no_main]
 
 use libfuzzer_sys::fuzz_target;
-use oxttl::{NQuadsParser, NQuadsSerializer};
+use oxrdf::Quad;
+use oxttl::{NQuadsParser, NQuadsSerializer, SyntaxError};
 
-fuzz_target!(|data: &[u8]| {
-    // We parse
+fn parse<'a>(chunks: impl IntoIterator<Item = &'a [u8]>) -> (Vec<Quad>, Vec<SyntaxError>) {
     let mut quads = Vec::new();
+    let mut errors = Vec::new();
     let mut parser = NQuadsParser::new().with_quoted_triples().parse();
-    for chunk in data.split(|c| *c == 0xFF) {
+    for chunk in chunks {
         parser.extend_from_slice(chunk);
         while let Some(result) = parser.read_next() {
-            if let Ok(quad) = result {
-                quads.push(quad);
+            match result {
+                Ok(quad) => quads.push(quad),
+                Err(error) => errors.push(error),
             }
         }
     }
     parser.end();
     while let Some(result) = parser.read_next() {
-        if let Ok(quad) = result {
-            quads.push(quad);
+        match result {
+            Ok(quad) => quads.push(quad),
+            Err(error) => errors.push(error),
         }
     }
     assert!(parser.is_end());
+    (quads, errors)
+}
+
+fuzz_target!(|data: &[u8]| {
+    // We parse with splitting
+    let (quads, errors) = parse(data.split(|c| *c == 0xFF));
+    // We parse without splitting
+    let (quads_without_split, errors_without_split) = parse([data
+        .iter()
+        .copied()
+        .filter(|c| *c != 0xFF)
+        .collect::<Vec<_>>()
+        .as_slice()]);
+    assert_eq!(quads, quads_without_split);
+    assert_eq!(errors.len(), errors_without_split.len());
 
     // We serialize
     let mut writer = NQuadsSerializer::new().serialize_to_write(Vec::new());

fuzz/fuzz_targets/sparql_query.rs

@@ -3,5 +3,5 @@ use libfuzzer_sys::fuzz_target;
 use spargebra::Query;
 
 fuzz_target!(|data: &str| {
-    Query::parse(data, None);
+    let _ = Query::parse(data, None);
 });

fuzz/fuzz_targets/sparql_results_json.rs

@@ -3,4 +3,4 @@ use libfuzzer_sys::fuzz_target;
 use oxigraph_fuzz::result_format::fuzz_result_format;
 use sparesults::QueryResultsFormat;
 
-fuzz_target!(|data: &[u8]| { fuzz_result_format(QueryResultsFormat::Json, data) });
+fuzz_target!(|data: &[u8]| fuzz_result_format(QueryResultsFormat::Json, data));

fuzz/fuzz_targets/sparql_results_tsv.rs

@@ -3,4 +3,4 @@ use libfuzzer_sys::fuzz_target;
 use oxigraph_fuzz::result_format::fuzz_result_format;
 use sparesults::QueryResultsFormat;
 
-fuzz_target!(|data: &[u8]| { fuzz_result_format(QueryResultsFormat::Tsv, data) });
+fuzz_target!(|data: &[u8]| fuzz_result_format(QueryResultsFormat::Tsv, data));

fuzz/fuzz_targets/sparql_results_xml.rs

@@ -3,4 +3,4 @@ use libfuzzer_sys::fuzz_target;
 use oxigraph_fuzz::result_format::fuzz_result_format;
 use sparesults::QueryResultsFormat;
 
-fuzz_target!(|data: &[u8]| { fuzz_result_format(QueryResultsFormat::Xml, data) });
+fuzz_target!(|data: &[u8]| fuzz_result_format(QueryResultsFormat::Xml, data));

fuzz/fuzz_targets/sparql_update.rs

@@ -4,5 +4,5 @@ use spargebra::Update;
 use std::str;
 
 fuzz_target!(|data: &str| {
-    Update::parse(data, None);
+    let _ = Update::parse(data, None);
 });

fuzz/fuzz_targets/trig.rs

@@ -1,38 +1,96 @@
 #![no_main]
 
 use libfuzzer_sys::fuzz_target;
-use oxttl::{TriGParser, TriGSerializer};
+use oxrdf::{Dataset, GraphName, Quad, Subject, Term, Triple};
+use oxttl::{SyntaxError, TriGParser, TriGSerializer};
 
-fuzz_target!(|data: &[u8]| {
-    // We parse
+fn parse<'a>(chunks: impl IntoIterator<Item = &'a [u8]>) -> (Vec<Quad>, Vec<SyntaxError>) {
     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();
-    for chunk in data.split(|c| *c == 0xFF) {
+    for chunk in chunks {
         parser.extend_from_slice(chunk);
         while let Some(result) = parser.read_next() {
-            if let Ok(quad) = result {
-                quads.push(quad);
+            match result {
+                Ok(quad) => quads.push(quad),
+                Err(error) => errors.push(error),
             }
         }
     }
     parser.end();
     while let Some(result) = parser.read_next() {
-        if let Ok(quad) = result {
-            quads.push(quad);
+        match result {
+            Ok(quad) => quads.push(quad),
+            Err(error) => errors.push(error),
         }
     }
     assert!(parser.is_end());
+    (quads, errors)
+}
 
-    // We serialize
+fn count_triple_blank_nodes(triple: &Triple) -> usize {
+    (match &triple.subject {
+        Subject::BlankNode(_) => 1,
+        Subject::Triple(t) => count_triple_blank_nodes(t),
+        _ => 0,
+    }) + (match &triple.object {
+        Term::BlankNode(_) => 1,
+        Term::Triple(t) => count_triple_blank_nodes(t),
+        _ => 0,
+    })
+}
+
+fn count_quad_blank_nodes(quad: &Quad) -> usize {
+    (match &quad.subject {
+        Subject::BlankNode(_) => 1,
+        Subject::Triple(t) => count_triple_blank_nodes(t),
+        _ => 0,
+    }) + (match &quad.object {
+        Term::BlankNode(_) => 1,
+        Term::Triple(t) => count_triple_blank_nodes(t),
+        _ => 0,
+    }) + usize::from(matches!(quad.graph_name, GraphName::BlankNode(_)))
+}
+
+fn serialize_quads(quads: &[Quad]) -> Vec<u8> {
     let mut writer = TriGSerializer::new().serialize_to_write(Vec::new());
-    for quad in &quads {
+    for quad in quads {
         writer.write_quad(quad).unwrap();
     }
-    let new_serialization = writer.finish().unwrap();
+    writer.finish().unwrap()
+}
+
+fuzz_target!(|data: &[u8]| {
+    // We parse with splitting
+    let (quads, errors) = parse(data.split(|c| *c == 0xFF));
+    // We parse without splitting
+    let (quads_without_split, errors_without_split) = parse([data
+        .iter()
+        .copied()
+        .filter(|c| *c != 0xFF)
+        .collect::<Vec<_>>()
+        .as_slice()]);
+    if quads.iter().map(count_quad_blank_nodes).sum::<usize>() < 2 {
+        let mut dataset_with_split = quads.iter().collect::<Dataset>();
+        let mut dataset_without_split = quads_without_split.iter().collect::<Dataset>();
+        dataset_with_split.canonicalize();
+        dataset_without_split.canonicalize();
+        assert_eq!(
+            dataset_with_split,
+            dataset_without_split,
+            "With split:\n{}\nWithout split:\n{}",
+            String::from_utf8_lossy(&serialize_quads(&quads)),
+            String::from_utf8_lossy(&serialize_quads(&quads_without_split))
+        );
+    }
+    assert_eq!(errors.len(), errors_without_split.len());
+
+    // We serialize
+    let new_serialization = serialize_quads(&quads);
 
     // We parse the serialization
     let new_quads = TriGParser::new()