ontocore_parser/
obo.rs

1//! OBO Format 1.4 parser via [`fastobo`] → OntoCore catalog model.
2
3use fastobo::ast::{HeaderClause, PropertyValue, TermClause};
4use ontocore_core::{
5    limits::{MAX_FILE_BYTES, MAX_TRIPLES_PER_FILE},
6    read_to_string_capped, Annotation, Axiom, Entity, EntityKind, SourceLocation,
7    AXIOM_KIND_SUB_CLASS_OF,
8};
9use std::collections::BTreeMap;
10use std::path::Path;
11
12use crate::rdf::{assemble_parsed_ontology, ParseError, ParsedOntology, Result};
13
14/// Default OBO PURL prefix (terms use `GO:0000001` → `…/obo/GO_0000001`).
15const DEFAULT_OBO_BASE: &str = "http://purl.obolibrary.org/obo/";
16const RDFS_SUBCLASS_OF: &str = "http://www.w3.org/2000/01/rdf-schema#subClassOf";
17const IAO_DEFINITION: &str = "http://purl.obolibrary.org/obo/IAO_0000115";
18const OBO_INOWL_NS: &str = "http://www.geneontology.org/formats/oboInOwl#";
19
20pub fn parse_obo_text(path: &Path, ontology_id: &str, source_text: &str) -> Result<ParsedOntology> {
21    let doc = fastobo::from_str(source_text)
22        .map_err(|e| ParseError::Rdf(format!("OBO parse error in {}: {e}", path.display())))?;
23
24    let mut namespaces = BTreeMap::new();
25    for clause in doc.header().iter() {
26        if let HeaderClause::Idspace(prefix, url, _) = clause {
27            namespaces.insert(prefix.to_string(), url.to_string());
28        }
29    }
30
31    let mut entities = Vec::new();
32    let mut annotations = Vec::new();
33    let mut axioms = Vec::new();
34    let mut axiom_counter = 0usize;
35
36    for entity in doc.entities() {
37        let Some(term) = entity.as_term() else {
38            continue;
39        };
40        let obo_id = ident_to_string(term.id().as_inner());
41        let iri = obo_id_to_iri(&obo_id, &namespaces);
42        let short_name = obo_id.split(':').next_back().unwrap_or(&obo_id).to_string();
43
44        let mut labels = Vec::new();
45        let mut comments = Vec::new();
46        let mut deprecated = false;
47
48        for clause in term.clauses() {
49            match clause.as_inner() {
50                TermClause::Name(name) => labels.push(name.to_string()),
51                TermClause::Comment(comment) => comments.push(comment.to_string()),
52                TermClause::Def(def) => {
53                    // Definition is IAO_0000115 only — do not also emit rdfs:comment (#308).
54                    annotations.push(Annotation {
55                        subject: iri.clone(),
56                        predicate: IAO_DEFINITION.to_string(),
57                        object: def.text().to_string(),
58                        ontology_id: ontology_id.to_string(),
59                        source_location: SourceLocation::default(),
60                    });
61                }
62                TermClause::IsObsolete(true) => deprecated = true,
63                TermClause::IsA(parent) => {
64                    axiom_counter += 1;
65                    let parent_id = ident_to_string(parent.as_ref());
66                    axioms.push(Axiom {
67                        id: format!("{ontology_id}#axiom-{axiom_counter}"),
68                        ontology_id: ontology_id.to_string(),
69                        subject: iri.clone(),
70                        // Absolute IRIs so catalog SQL / SPARQL / diff agree (#305).
71                        predicate: RDFS_SUBCLASS_OF.to_string(),
72                        object: obo_id_to_iri(&parent_id, &namespaces),
73                        axiom_kind: AXIOM_KIND_SUB_CLASS_OF.to_string(),
74                        source_location: SourceLocation::default(),
75                    });
76                }
77                TermClause::Synonym(syn) => {
78                    annotations.push(Annotation {
79                        subject: iri.clone(),
80                        predicate: format!("{OBO_INOWL_NS}has{}Synonym", scope_label(syn.scope())),
81                        object: syn.description().to_string(),
82                        ontology_id: ontology_id.to_string(),
83                        source_location: SourceLocation::default(),
84                    });
85                }
86                TermClause::Xref(xref) => {
87                    annotations.push(Annotation {
88                        subject: iri.clone(),
89                        predicate: format!("{OBO_INOWL_NS}hasDbXref"),
90                        object: xref.as_ref().to_string(),
91                        ontology_id: ontology_id.to_string(),
92                        source_location: SourceLocation::default(),
93                    });
94                }
95                TermClause::PropertyValue(pv) => {
96                    let (predicate, object) = property_value_parts(pv, &namespaces);
97                    annotations.push(Annotation {
98                        subject: iri.clone(),
99                        predicate,
100                        object,
101                        ontology_id: ontology_id.to_string(),
102                        source_location: SourceLocation::default(),
103                    });
104                }
105                _ => {}
106            }
107        }
108
109        entities.push(Entity {
110            iri,
111            short_name,
112            kind: EntityKind::Class,
113            ontology_id: ontology_id.to_string(),
114            source_location: SourceLocation::default(),
115            labels,
116            comments,
117            deprecated,
118            obo_id: Some(obo_id),
119            characteristics: Default::default(),
120        });
121
122        if entities.len() + annotations.len() + axioms.len() > MAX_TRIPLES_PER_FILE {
123            return Err(ParseError::LimitExceeded(format!(
124                "OBO file exceeds entity/axiom limit: {}",
125                path.display()
126            )));
127        }
128    }
129
130    if !namespaces.contains_key("") {
131        namespaces.insert(String::new(), DEFAULT_OBO_BASE.to_string());
132    }
133
134    Ok(assemble_parsed_ontology(
135        ontology_id,
136        Some(DEFAULT_OBO_BASE.to_string()),
137        namespaces,
138        entities,
139        annotations,
140        axioms,
141    ))
142}
143
144fn ident_to_string<T: std::fmt::Display>(ident: &T) -> String {
145    ident.to_string()
146}
147
148fn scope_label(scope: &fastobo::ast::SynonymScope) -> &'static str {
149    use fastobo::ast::SynonymScope;
150    match scope {
151        SynonymScope::Exact => "Exact",
152        SynonymScope::Broad => "Broad",
153        SynonymScope::Narrow => "Narrow",
154        SynonymScope::Related => "Related",
155    }
156}
157
158/// Split property_value into predicate IRI + object (not the full Display form) (#306).
159fn property_value_parts(
160    pv: &PropertyValue,
161    namespaces: &BTreeMap<String, String>,
162) -> (String, String) {
163    let predicate = obo_id_to_iri(&ident_to_string(pv.property()), namespaces);
164    let object = match pv {
165        PropertyValue::Resource(r) => obo_id_to_iri(&ident_to_string(r.target()), namespaces),
166        PropertyValue::Literal(l) => l.literal().as_str().to_string(),
167    };
168    (predicate, object)
169}
170
171/// Map an OBO ID to an IRI using `idspace:` expansions when present.
172fn obo_id_to_iri(obo_id: &str, namespaces: &BTreeMap<String, String>) -> String {
173    if obo_id.starts_with("http://") || obo_id.starts_with("https://") {
174        return obo_id.to_string();
175    }
176    if let Some((prefix, local)) = obo_id.split_once(':') {
177        if let Some(ns) = namespaces.get(prefix) {
178            // When the idspace URL already ends with `PREFIX_` (e.g. `…/obo/GO_`), append
179            // only the local segment. Otherwise append `PREFIX_LOCAL` per OBO PURL rules.
180            if ns.ends_with(&format!("{prefix}_")) {
181                return format!("{ns}{local}");
182            }
183            return format!("{ns}{prefix}_{local}");
184        }
185    }
186    let normalized = obo_id.replace(':', "_");
187    format!("{DEFAULT_OBO_BASE}{normalized}")
188}
189
190pub fn parse_obo_file(
191    path: &Path,
192    ontology_id: &str,
193    _content_hash: &str,
194    _modified_time: u64,
195) -> Result<ParsedOntology> {
196    let content = read_to_string_capped(path, MAX_FILE_BYTES)
197        .map_err(|e| ParseError::LimitExceeded(format!("{}: {e}", path.display())))?;
198    parse_obo_text(path, ontology_id, &content)
199}
200
201#[cfg(test)]
202mod tests {
203    use super::*;
204    use std::io::Write;
205    use tempfile::NamedTempFile;
206
207    #[test]
208    fn parses_minimal_obo() {
209        let mut file = NamedTempFile::new().unwrap();
210        writeln!(
211            file,
212            "format-version: 1.2\nontology: test\n\n[Term]\nid: TEST:0000001\nname: example term\nis_a: TEST:0000002 ! parent\n"
213        )
214        .unwrap();
215        let parsed = parse_obo_file(file.path(), "doc-1", "hash", 0).unwrap();
216        assert_eq!(parsed.entities.len(), 1);
217        assert_eq!(parsed.entities[0].obo_id.as_deref(), Some("TEST:0000001"));
218        assert_eq!(parsed.entities[0].iri, "http://purl.obolibrary.org/obo/TEST_0000001");
219        assert_eq!(parsed.entities[0].labels, vec!["example term"]);
220        assert_eq!(parsed.axioms.len(), 1);
221        assert_eq!(parsed.axioms[0].predicate, "http://www.w3.org/2000/01/rdf-schema#subClassOf");
222        assert_eq!(parsed.axioms[0].object, "http://purl.obolibrary.org/obo/TEST_0000002");
223        assert!(parsed.triple_count > 0);
224        assert!(!parsed.quads().is_empty(), "OBO must materialize RDF quads");
225    }
226
227    #[test]
228    fn materializes_all_synonym_scopes_and_definition_in_sparql_quads() {
229        let text = "format-version: 1.2\nontology: test\n\n[Term]\n\
230id: TEST:0000001\n\
231name: test term\n\
232synonym: \"exact syn\" EXACT []\n\
233synonym: \"broad syn\" BROAD []\n\
234synonym: \"narrow syn\" NARROW []\n\
235synonym: \"related syn\" RELATED []\n\
236def: \"A definition.\" []\n";
237        let parsed = parse_obo_text(Path::new("syn.obo"), "doc-1", text).unwrap();
238        let predicates: std::collections::BTreeSet<_> =
239            parsed.quads().iter().map(|q| q.predicate.as_str().to_string()).collect();
240        assert!(predicates.contains("http://www.geneontology.org/formats/oboInOwl#hasExactSynonym"));
241        assert!(predicates.contains("http://www.geneontology.org/formats/oboInOwl#hasBroadSynonym"));
242        assert!(
243            predicates.contains("http://www.geneontology.org/formats/oboInOwl#hasNarrowSynonym")
244        );
245        assert!(
246            predicates.contains("http://www.geneontology.org/formats/oboInOwl#hasRelatedSynonym")
247        );
248        assert!(predicates.contains("http://purl.obolibrary.org/obo/IAO_0000115"));
249        assert!(
250            !predicates.contains("http://www.w3.org/2000/01/rdf-schema#comment"),
251            "def: must not also emit rdfs:comment (#308)"
252        );
253        assert!(parsed.annotations.iter().all(|a| a.predicate.contains("://")));
254        assert!(parsed.axioms.iter().all(|a| a.predicate.contains("://")));
255    }
256
257    #[test]
258    fn parses_def_and_synonym_via_fastobo() {
259        let text = "format-version: 1.2\nontology: test\n\n[Term]\n\
260id: TEST:0000001\n\
261name: example\n\
262def: \"A definition.\" []\n\
263synonym: \"alt label\" EXACT []\n\
264comment: \"A true comment.\"\n";
265        let parsed = parse_obo_text(Path::new("t.obo"), "doc-1", text).unwrap();
266        assert!(
267            !parsed.entities[0].comments.iter().any(|c| c.contains("definition")),
268            "def text must not land in entity.comments (#308)"
269        );
270        assert!(parsed.entities[0].comments.iter().any(|c| c.contains("true comment")));
271        assert!(parsed.annotations.iter().any(|a| {
272            a.predicate == "http://purl.obolibrary.org/obo/IAO_0000115"
273                && a.object.contains("definition")
274        }));
275        assert!(parsed.annotations.iter().any(|a| {
276            a.predicate == "http://www.geneontology.org/formats/oboInOwl#hasExactSynonym"
277        }));
278    }
279
280    #[test]
281    fn property_value_stores_target_not_display() {
282        let text = "format-version: 1.2\n\
283idspace: dc http://purl.org/dc/elements/1.1/\n\
284idspace: ORCID http://orcid.org/\n\n\
285[Term]\n\
286id: TEST:0000001\n\
287name: example\n\
288property_value: dc:creator ORCID:0000-0002-3947-4444\n\
289property_value: shoe_size \"8\" xsd:positiveInteger\n";
290        let parsed = parse_obo_text(Path::new("pv.obo"), "doc-1", text).unwrap();
291        let resource = parsed
292            .annotations
293            .iter()
294            .find(|a| {
295                a.predicate == "http://purl.org/dc/elements/1.1/dc_creator"
296                    || a.predicate == "http://purl.org/dc/elements/1.1/creator"
297            })
298            .expect("dc:creator annotation");
299        assert!(
300            !resource.object.contains("dc:creator"),
301            "object must not duplicate the property Display: {}",
302            resource.object
303        );
304        assert!(
305            resource.object.contains("0000-0002-3947-4444"),
306            "object should be the ORCID target: {}",
307            resource.object
308        );
309        let literal =
310            parsed.annotations.iter().find(|a| a.object == "8").expect("shoe_size literal");
311        assert!(!literal.object.contains("shoe_size"));
312        let predicates: std::collections::BTreeSet<_> =
313            parsed.quads().iter().map(|q| q.predicate.as_str().to_string()).collect();
314        assert!(
315            predicates.iter().any(|p| p.contains("creator") || p.contains("shoe_size")),
316            "property_value must appear in SPARQL quads"
317        );
318    }
319
320    #[test]
321    fn idspace_overrides_default_base() {
322        let text = "format-version: 1.2\n\
323idspace: GO http://purl.obolibrary.org/obo/GO_\n\n\
324[Term]\n\
325id: GO:0000001\n\
326name: mitochondrion\n";
327        let parsed = parse_obo_text(Path::new("go.obo"), "doc-1", text).unwrap();
328        assert_eq!(parsed.entities[0].iri, "http://purl.obolibrary.org/obo/GO_0000001");
329    }
330
331    #[test]
332    fn idspace_standard_obo_foundry_base_normalizes_colon_to_underscore() {
333        let text = "format-version: 1.2\n\
334idspace: GO http://purl.obolibrary.org/obo/\n\n\
335[Term]\n\
336id: GO:0000001\n\
337name: mitochondrion\n\
338is_a: GO:0000002 ! parent\n";
339        let parsed = parse_obo_text(Path::new("go.obo"), "doc-1", text).unwrap();
340        assert_eq!(parsed.entities[0].iri, "http://purl.obolibrary.org/obo/GO_0000001");
341        assert_eq!(parsed.axioms[0].object, "http://purl.obolibrary.org/obo/GO_0000002");
342    }
343
344    #[test]
345    fn rejects_oversized_obo_source_text() {
346        let huge = "x".repeat((ontocore_core::MAX_FILE_BYTES as usize) + 1);
347        let err = crate::rdf::parse_ontology_text(
348            Path::new("big.obo"),
349            ontocore_core::OntologyFormat::Obo,
350            "doc-1",
351            &huge,
352            huge.as_bytes(),
353        )
354        .unwrap_err();
355        assert!(err.to_string().contains("exceeds"));
356    }
357}
ontocore_parser/obo.rs

ontocore_parser/
obo.rs
