Skip to main content

provenant/parsers/
about.rs

1// SPDX-FileCopyrightText: Provenant contributors
2// SPDX-License-Identifier: Apache-2.0
3
4//! Parser for AboutCode .ABOUT metadata files.
5//!
6//! Extracts package metadata from AboutCode .ABOUT YAML files which describe
7//! software components, licenses, and related information.
8//!
9//! # Supported Formats
10//! - .ABOUT (case-sensitive uppercase extension)
11//!
12//! # Key Features
13//! - YAML-based metadata parsing
14//! - Package URL (purl) parsing for type/namespace extraction
15//! - Owner party information
16//! - File reference tracking (about_resource field)
17//! - License expression extraction
18//! - Flexible field mapping (home_url/homepage_url)
19//!
20//! # Implementation Notes
21//! - Uses yaml_serde for YAML parsing
22//! - Uses packageurl crate for purl parsing
23//! - Extension is case-sensitive and must be uppercase (.ABOUT not .about)
24//! - Type can be overridden by 'type' field or extracted from 'purl' field
25//! - Graceful error handling: logs warnings and returns default on parse failure
26
27use crate::models::{DatasourceId, FileReference, PackageData, PackageType, Party};
28use crate::parser_warn as warn;
29use crate::parsers::utils::{read_file_to_string, truncate_field};
30use packageurl::PackageUrl;
31use std::path::Path;
32use std::str::FromStr;
33use url::Url;
34use yaml_serde::Value;
35
36use super::PackageParser;
37use super::license_normalization::{
38    DeclaredLicenseMatchMetadata, build_declared_license_data, normalize_spdx_declared_license,
39    normalize_spdx_expression,
40};
41
42const FIELD_TYPE: &str = "type";
43const FIELD_PURL: &str = "purl";
44const FIELD_PACKAGE_URL: &str = "package_url";
45const FIELD_NAMESPACE: &str = "namespace";
46const FIELD_NAME: &str = "name";
47const FIELD_VERSION: &str = "version";
48const FIELD_HOME_URL: &str = "home_url";
49const FIELD_HOMEPAGE_URL: &str = "homepage_url";
50const FIELD_DOWNLOAD_URL: &str = "download_url";
51const FIELD_COPYRIGHT: &str = "copyright";
52const FIELD_LICENSE_EXPRESSION: &str = "license_expression";
53const FIELD_OWNER: &str = "owner";
54const FIELD_ABOUT_RESOURCE: &str = "about_resource";
55
56/// AboutCode .ABOUT file parser.
57///
58/// Parses AboutCode metadata files that contain package information,
59/// licensing, and file references in YAML format.
60pub struct AboutFileParser;
61
62#[derive(Clone)]
63struct InferredAboutIdentity {
64    package_type: PackageType,
65    namespace: Option<String>,
66    name: Option<String>,
67    version: Option<String>,
68}
69
70impl PackageParser for AboutFileParser {
71    const PACKAGE_TYPE: PackageType = PackageType::About;
72
73    fn extract_packages(path: &Path) -> Vec<PackageData> {
74        let yaml = match read_and_parse_yaml(path) {
75            Ok(yaml) => yaml,
76            Err(e) => {
77                warn!("Failed to read or parse .ABOUT file at {:?}: {}", path, e);
78                return vec![default_package_data()];
79            }
80        };
81
82        // Extract type and purl information
83        let about_type = yaml
84            .get(FIELD_TYPE)
85            .and_then(|v| v.as_str())
86            .map(String::from);
87
88        let about_namespace = yaml
89            .get(FIELD_NAMESPACE)
90            .and_then(|v| v.as_str())
91            .map(|v| truncate_field(v.to_string()));
92
93        let purl_string = yaml
94            .get(FIELD_PURL)
95            .and_then(|v| v.as_str())
96            .or_else(|| yaml.get(FIELD_PACKAGE_URL).and_then(|v| v.as_str()))
97            .map(|v| truncate_field(v.to_string()));
98
99        // Parse purl if present
100        let (purl_type, purl_namespace, purl_name, purl_version) =
101            if let Some(ref purl_str) = purl_string {
102                match PackageUrl::from_str(purl_str) {
103                    Ok(purl) => (
104                        Some(truncate_field(purl.ty().to_string())),
105                        purl.namespace().map(|v| truncate_field(v.to_string())),
106                        Some(truncate_field(purl.name().to_string())),
107                        purl.version().map(|v| truncate_field(v.to_string())),
108                    ),
109                    Err(e) => {
110                        warn!("Failed to parse purl '{}': {}", purl_str, e);
111                        (None, None, None, None)
112                    }
113                }
114            } else {
115                (None, None, None, None)
116            };
117
118        let inferred = infer_about_from_download_url(
119            yaml.get(FIELD_DOWNLOAD_URL).and_then(|v| v.as_str()),
120            yaml.get(FIELD_NAME)
121                .and_then(yaml_value_to_string)
122                .as_deref(),
123            yaml.get(FIELD_VERSION)
124                .and_then(yaml_value_to_string)
125                .as_deref(),
126        );
127
128        let package_type = about_type
129            .clone()
130            .or(purl_type)
131            .and_then(|s| s.parse::<crate::models::PackageType>().ok())
132            .or_else(|| inferred.as_ref().map(|identity| identity.package_type))
133            .unwrap_or(Self::PACKAGE_TYPE);
134
135        // Priority: about_namespace > purl_namespace
136        let namespace = about_namespace
137            .clone()
138            .or(purl_namespace.clone())
139            .or_else(|| {
140                inferred
141                    .as_ref()
142                    .and_then(|identity| identity.namespace.clone())
143            })
144            .map(truncate_field);
145
146        // Name and version from YAML or purl
147        let name = yaml
148            .get(FIELD_NAME)
149            .and_then(yaml_value_to_string)
150            .or(purl_name.clone())
151            .or_else(|| inferred.as_ref().and_then(|identity| identity.name.clone()))
152            .map(truncate_field);
153
154        let version = yaml
155            .get(FIELD_VERSION)
156            .and_then(yaml_value_to_string)
157            .or(purl_version.clone())
158            .or_else(|| {
159                inferred
160                    .as_ref()
161                    .and_then(|identity| identity.version.clone())
162            })
163            .map(truncate_field);
164
165        // Homepage URL (two possible field names)
166        let homepage_url = yaml
167            .get(FIELD_HOME_URL)
168            .and_then(|v| v.as_str())
169            .or_else(|| yaml.get(FIELD_HOMEPAGE_URL).and_then(|v| v.as_str()))
170            .map(|v| truncate_field(v.to_string()));
171
172        let download_url = yaml
173            .get(FIELD_DOWNLOAD_URL)
174            .and_then(|v| v.as_str())
175            .map(|v| truncate_field(v.to_string()));
176
177        let copyright = yaml
178            .get(FIELD_COPYRIGHT)
179            .and_then(|v| v.as_str())
180            .map(|v| truncate_field(v.to_string()));
181
182        let extracted_license_statement = yaml
183            .get(FIELD_LICENSE_EXPRESSION)
184            .and_then(|v| v.as_str())
185            .map(|v| truncate_field(v.to_string()));
186        let file_references = extract_file_references(&yaml);
187        let (declared_license_expression, declared_license_expression_spdx, license_detections) =
188            extracted_license_statement
189                .as_deref()
190                .and_then(normalize_spdx_expression)
191                .map(|normalized| {
192                    build_declared_license_data(
193                        normalized,
194                        DeclaredLicenseMatchMetadata::single_line(
195                            extracted_license_statement.as_deref().unwrap_or_default(),
196                        ),
197                    )
198                })
199                .unwrap_or_else(|| {
200                    normalize_spdx_declared_license(extracted_license_statement.as_deref())
201                });
202
203        let vcs_url = yaml
204            .get(Value::String("vcs_url".to_string()))
205            .and_then(|v| v.as_str())
206            .map(|v| truncate_field(v.to_string()));
207
208        let extra_data = build_extra_data(&yaml);
209
210        let purl = purl_string
211            .or_else(|| {
212                let name = yaml
213                    .get(FIELD_NAME)
214                    .and_then(yaml_value_to_string)
215                    .or(purl_name.clone())
216                    .or_else(|| inferred.as_ref().and_then(|identity| identity.name.clone()));
217                let version = yaml
218                    .get(FIELD_VERSION)
219                    .and_then(yaml_value_to_string)
220                    .or(purl_version.clone())
221                    .or_else(|| {
222                        inferred
223                            .as_ref()
224                            .and_then(|identity| identity.version.clone())
225                    });
226                let namespace = about_namespace.clone().or_else(|| {
227                    inferred
228                        .as_ref()
229                        .and_then(|identity| identity.namespace.clone())
230                });
231                build_about_purl(
232                    package_type,
233                    namespace.as_deref(),
234                    name.as_deref(),
235                    version.as_deref(),
236                )
237            })
238            .map(truncate_field);
239
240        // Owner party
241        let parties = extract_owner_party(&yaml);
242
243        // File references
244        vec![PackageData {
245            package_type: Some(package_type),
246            namespace,
247            name,
248            version,
249            qualifiers: None,
250            subpath: None,
251            primary_language: None,
252            description: None,
253            release_date: None,
254            parties,
255            keywords: Vec::new(),
256            homepage_url,
257            download_url,
258            size: None,
259            sha1: None,
260            md5: None,
261            sha256: None,
262            sha512: None,
263            bug_tracking_url: None,
264            code_view_url: None,
265            vcs_url,
266            copyright,
267            holder: None,
268            declared_license_expression,
269            declared_license_expression_spdx,
270            license_detections,
271            other_license_expression: None,
272            other_license_expression_spdx: None,
273            other_license_detections: Vec::new(),
274            extracted_license_statement,
275            notice_text: None,
276            source_packages: Vec::new(),
277            file_references,
278            is_private: false,
279            is_virtual: false,
280            extra_data,
281            dependencies: Vec::new(),
282            repository_homepage_url: None,
283            repository_download_url: None,
284            api_data_url: None,
285            datasource_id: Some(DatasourceId::AboutFile),
286            purl,
287        }]
288    }
289
290    fn is_match(path: &Path) -> bool {
291        path.extension()
292            .and_then(|ext| ext.to_str())
293            .is_some_and(|ext| ext == "ABOUT")
294    }
295}
296
297/// Reads and parses a YAML file.
298fn read_and_parse_yaml(path: &Path) -> Result<yaml_serde::Mapping, String> {
299    let content =
300        read_file_to_string(path, None).map_err(|e| format!("Failed to read file: {}", e))?;
301
302    parse_yaml_mapping(&content)
303        .or_else(|yaml_error| parse_shallow_scalar_mapping(&content).ok_or(yaml_error))
304}
305
306fn parse_yaml_mapping(content: &str) -> Result<yaml_serde::Mapping, String> {
307    let value: Value =
308        yaml_serde::from_str(content).map_err(|e| format!("Failed to parse YAML: {}", e))?;
309
310    match value {
311        Value::Mapping(map) => Ok(map),
312        _ => Err("Expected YAML mapping at root".to_string()),
313    }
314}
315
316fn parse_shallow_scalar_mapping(content: &str) -> Option<yaml_serde::Mapping> {
317    let mut map = yaml_serde::Mapping::new();
318    let mut saw_mapping_entry = false;
319
320    for line in content.lines() {
321        let trimmed = line.trim();
322        if trimmed.is_empty() || trimmed.starts_with('#') {
323            continue;
324        }
325        if line.starts_with(char::is_whitespace) {
326            return None;
327        }
328
329        let (raw_key, raw_value) = trimmed.split_once(':')?;
330        let key = raw_key.trim();
331        if key.is_empty()
332            || !key.chars().all(|character| {
333                character.is_ascii_alphanumeric() || matches!(character, '_' | '-')
334            })
335        {
336            return None;
337        }
338
339        let value = raw_value.trim();
340        if value.is_empty() {
341            return None;
342        }
343
344        saw_mapping_entry = true;
345        map.insert(
346            Value::String(key.to_string()),
347            Value::String(unquote_yaml_scalar(value)),
348        );
349    }
350
351    saw_mapping_entry.then_some(map)
352}
353
354fn unquote_yaml_scalar(value: &str) -> String {
355    if value.len() >= 2 {
356        let mut characters = value.chars();
357        let first = characters.next();
358        let last = value.chars().last();
359        if matches!(
360            (first, last),
361            (Some('"'), Some('"')) | (Some('\''), Some('\''))
362        ) {
363            return value[1..value.len() - 1].to_string();
364        }
365    }
366    value.to_string()
367}
368
369/// Converts a YAML value to a string, handling strings, numbers, and booleans.
370fn yaml_value_to_string(value: &Value) -> Option<String> {
371    match value {
372        Value::String(s) => Some(s.clone()),
373        Value::Number(n) => Some(n.to_string()),
374        Value::Bool(b) => Some(b.to_string()),
375        _ => None,
376    }
377}
378
379/// Extracts owner party information from YAML.
380fn extract_owner_party(yaml: &yaml_serde::Mapping) -> Vec<Party> {
381    let owner = yaml
382        .get(Value::String(FIELD_OWNER.to_string()))
383        .map(|v| match v {
384            Value::String(s) => truncate_field(s.clone()),
385            _ => truncate_field(format!("{:?}", v)),
386        });
387
388    if let Some(owner_name) = owner {
389        if !owner_name.is_empty() {
390            vec![Party {
391                r#type: Some("person".to_string()),
392                role: Some("owner".to_string()),
393                name: Some(owner_name),
394                email: None,
395                url: None,
396                organization: None,
397                organization_url: None,
398                timezone: None,
399            }]
400        } else {
401            Vec::new()
402        }
403    } else {
404        Vec::new()
405    }
406}
407
408/// Extracts file references from YAML.
409fn extract_file_references(yaml: &yaml_serde::Mapping) -> Vec<FileReference> {
410    let about_resource = yaml
411        .get(Value::String(FIELD_ABOUT_RESOURCE.to_string()))
412        .and_then(|v| v.as_str());
413    let license_file = yaml
414        .get(Value::String("license_file".to_string()))
415        .and_then(|v| v.as_str());
416    let notice_file = yaml
417        .get(Value::String("notice_file".to_string()))
418        .and_then(|v| v.as_str());
419
420    let mut refs = Vec::new();
421
422    if let Some(path) = about_resource {
423        refs.push(FileReference {
424            path: truncate_field(path.to_string()),
425            size: None,
426            sha1: None,
427            md5: None,
428            sha256: None,
429            sha512: None,
430            extra_data: None,
431        });
432    }
433
434    for path in [license_file, notice_file].into_iter().flatten() {
435        refs.push(FileReference {
436            path: truncate_field(path.to_string()),
437            size: None,
438            sha1: None,
439            md5: None,
440            sha256: None,
441            sha512: None,
442            extra_data: None,
443        });
444    }
445
446    refs
447}
448
449/// Returns a default (empty) PackageData structure.
450fn default_package_data() -> PackageData {
451    PackageData {
452        package_type: Some(PackageType::About),
453        datasource_id: Some(DatasourceId::AboutFile),
454        ..Default::default()
455    }
456}
457
458fn infer_about_from_download_url(
459    download_url: Option<&str>,
460    about_name: Option<&str>,
461    about_version: Option<&str>,
462) -> Option<InferredAboutIdentity> {
463    let url = Url::parse(download_url?).ok()?;
464    let host = url.host_str()?;
465
466    if matches!(host, "pypi.python.org" | "files.pythonhosted.org") {
467        let name = about_name.map(str::to_string)?;
468        let version = about_version.map(str::to_string);
469        return Some(InferredAboutIdentity {
470            package_type: PackageType::Pypi,
471            namespace: None,
472            name: Some(name),
473            version,
474        });
475    }
476
477    if matches!(host, "raw.githubusercontent.com" | "github.com") {
478        let mut segments = url.path_segments()?;
479        let owner = segments.next()?.to_string();
480        let repo = segments.next()?.to_string();
481        return Some(InferredAboutIdentity {
482            package_type: PackageType::Github,
483            namespace: Some(owner),
484            name: Some(repo),
485            version: None,
486        });
487    }
488
489    None
490}
491
492fn build_about_purl(
493    package_type: PackageType,
494    namespace: Option<&str>,
495    name: Option<&str>,
496    version: Option<&str>,
497) -> Option<String> {
498    if package_type == PackageType::About {
499        return None;
500    }
501
502    let name = name?;
503    let mut purl = PackageUrl::new(package_type.as_str(), name).ok()?;
504    if let Some(namespace) = namespace {
505        purl.with_namespace(namespace).ok()?;
506    }
507    if let Some(version) = version {
508        purl.with_version(version).ok()?;
509    }
510    Some(purl.to_string())
511}
512
513fn build_extra_data(
514    yaml: &yaml_serde::Mapping,
515) -> Option<std::collections::HashMap<String, serde_json::Value>> {
516    let mut extra_data = std::collections::HashMap::new();
517    for key in ["license_file", "notice_file", "notes"] {
518        if let Some(value) = yaml.get(Value::String(key.to_string()))
519            && let Some(value) = yaml_value_to_string(value)
520        {
521            extra_data.insert(
522                key.to_string(),
523                serde_json::Value::String(truncate_field(value)),
524            );
525        }
526    }
527    (!extra_data.is_empty()).then_some(extra_data)
528}
529
530crate::register_parser!(
531    "AboutCode .ABOUT metadata file",
532    &["**/*.ABOUT"],
533    "about",
534    "",
535    Some("https://aboutcode-toolkit.readthedocs.io/en/latest/specification.html"),
536);