Skip to main content

rust_docx/parse/
build.rs

1use crate::model::{DocModel, DocModelMetadata, DocumentNoteDefinition};
2use std::cell::RefCell;
3
4use crate::package::OoxmlPackage;
5
6use super::content_types::parse_content_types;
7use super::context::ParseContext;
8#[cfg(target_arch = "wasm32")]
9use super::document::parse_document_xml;
10use super::metadata::{
11    extract_document_open_tag, extract_section_properties_xml, parse_document_background_color,
12    parse_document_compatibility_settings, parse_document_default_tab_stop_twips,
13    parse_document_page_count_from_app_properties,
14};
15use super::numbering::parse_numbering_with_assets;
16use super::relationships::parse_part_relationships;
17use super::sections::{
18    parse_document_comments, parse_document_notes_from_part, parse_document_sections,
19    parse_footer_sections, parse_header_sections,
20};
21use super::styles::{acquire_document_style_sheet, release_document_style_metadata};
22
23pub fn build_doc_model_with_options(pkg: &OoxmlPackage, options: &crate::ParseOptions) -> DocModel {
24    let mut model = build_doc_model(pkg);
25    if options.include_3d_models {
26        super::model3d::attach_models(&mut model, pkg, options);
27    }
28    model
29}
30
31pub fn build_doc_model(pkg: &OoxmlPackage) -> DocModel {
32    #[cfg(not(target_arch = "wasm32"))]
33    {
34        let bytes = pkg.parts.values().map(|part| part.content.len())
35            .chain(pkg.binary_assets.values().map(Vec::len))
36            .fold(0usize, usize::saturating_add);
37        if let Some(pool) = (bytes >= 64 * 1024)
38            .then(super::namespaces::worker_pool).flatten()
39        {
40            let mut identity = None;
41            let mut model = pool.in_place_scope(|scope| {
42                scope.spawn(|_| {
43                    identity = Some(crate::structured::package_identity(pkg));
44                });
45                build_doc_model_contents(pkg)
46            });
47            crate::structured::assign_source_ids(&mut model, &identity.expect("joined identity task"));
48            return model;
49        }
50    }
51    let identity = crate::structured::package_identity(pkg);
52    let mut model = build_doc_model_contents(pkg);
53    crate::structured::assign_source_ids(&mut model, &identity);
54    model
55}
56
57fn build_doc_model_contents(pkg: &OoxmlPackage) -> DocModel {
58    let _style_cache = super::style::run_style_cache_scope();
59    let binary_assets = &pkg.binary_assets;
60    let (normalized, indexes) = super::namespaces::normalized_package(pkg);
61    let pkg = &normalized;
62    let _relationship_cache = crate::package::relationship_cache_scope(pkg);
63    let _xml_index = crate::xml_index::scope(
64        pkg.parts
65            .values()
66            .map(|part| (part.name.as_str(), part.content.as_str())),
67        indexes,
68    );
69    let mut warnings = pkg.warnings.clone();
70    warnings.extend(
71        crate::source_graph::preserved_features(pkg)
72            .into_iter()
73            .map(|feature| {
74                format!(
75                    "{}: {} occurrence(s) of {} have support level {}",
76                    feature.part_name, feature.occurrences, feature.feature, feature.support
77                )
78            }),
79    );
80    let main_part = crate::package::main_document_part_name(pkg);
81    let source_part_names = ["footnotes", "endnotes", "comments"]
82        .into_iter()
83        .map(|kind| {
84            (
85                kind.to_string(),
86                crate::package::related_document_part_name(pkg, kind),
87            )
88        })
89        .chain(std::iter::once(("document".into(), main_part.clone())))
90        .collect();
91    let document_xml = pkg.parts.get(&main_part).map(|part| part.content.as_str());
92    if document_xml.is_none() {
93        warnings.push("Missing word/document.xml".to_string());
94    }
95
96    let resolved_document_xml = document_xml.unwrap_or("");
97    let document_open_tag = extract_document_open_tag(resolved_document_xml);
98    let document_background_color = parse_document_background_color(resolved_document_xml);
99    let document_page_count = parse_document_page_count_from_app_properties(pkg);
100    let compatibility = parse_document_compatibility_settings(pkg);
101    let default_tab_stop_twips = parse_document_default_tab_stop_twips(pkg);
102    let section_properties_xml = extract_section_properties_xml(resolved_document_xml);
103    let content_types = parse_content_types(pkg);
104    let style_handle = acquire_document_style_sheet(pkg);
105    let style_sheet: &super::context::ParsedStyleSheet = std::borrow::Borrow::borrow(&style_handle);
106    super::style::bind_style_cache(style_sheet);
107    let numbering_definitions = parse_numbering_with_assets(pkg, &content_types, binary_assets);
108    let document_relationships = parse_part_relationships(pkg, &main_part);
109
110    let mut part_cache =
111        super::sections::PartNodeCache::new(pkg, &content_types, style_sheet, binary_assets);
112    let header_sections = parse_header_sections(
113        pkg,
114        resolved_document_xml,
115        &document_relationships,
116        &mut warnings,
117        &mut part_cache,
118    );
119    let footer_sections = parse_footer_sections(
120        pkg,
121        resolved_document_xml,
122        &document_relationships,
123        &mut warnings,
124        &mut part_cache,
125    );
126    let sections = parse_document_sections(
127        resolved_document_xml,
128        &document_relationships,
129        &mut warnings,
130        &mut part_cache,
131    );
132    drop(part_cache);
133    let footnotes = parse_document_notes_from_part(
134        pkg,
135        &crate::package::related_document_part_name(pkg, "footnotes"),
136        "w:footnote",
137        &content_types,
138        style_sheet,
139        &mut warnings,
140        binary_assets,
141    );
142    let endnotes = parse_document_notes_from_part(
143        pkg,
144        &crate::package::related_document_part_name(pkg, "endnotes"),
145        "w:endnote",
146        &content_types,
147        style_sheet,
148        &mut warnings,
149        binary_assets,
150    );
151    let comments = parse_document_comments(pkg, &content_types, style_sheet, binary_assets);
152
153    let context = ParseContext {
154        relationships: document_relationships.clone(),
155        content_types: content_types.clone(),
156        parts: &pkg.parts,
157        binary_assets,
158        style_sheet: style_sheet,
159        warnings: RefCell::new(Vec::new()),
160    };
161
162    #[cfg(not(target_arch = "wasm32"))]
163    let nodes =
164        super::document::parse_document_xml_parallel(resolved_document_xml, &context, &_xml_index);
165    #[cfg(target_arch = "wasm32")]
166    let nodes = parse_document_xml(resolved_document_xml, &context);
167    warnings.extend(context.warnings.borrow().iter().cloned());
168
169    let (paragraph_styles, default_paragraph_style_id) =
170        release_document_style_metadata(style_handle);
171    let model = DocModel {
172        nodes,
173        metadata: DocModelMetadata {
174            document_id: None,
175            source_part_names: Some(source_part_names),
176            source_parts: pkg.parts.len() as i64,
177            warnings,
178            document_page_count,
179            document_open_tag,
180            document_background_color,
181            section_properties_xml,
182            sections: Some(sections),
183            header_sections,
184            footer_sections,
185            paragraph_styles,
186            default_paragraph_style_id,
187            default_tab_stop_twips,
188            document_settings_imported: crate::package::related_document_part(pkg, "settings")
189                .is_some()
190                .then_some(true),
191            line_spacing_compatibility_imported: crate::package::related_document_part(
192                pkg, "settings",
193            )
194            .is_some()
195            .then_some(true),
196            numbering_definitions,
197            compatibility,
198            footnotes: clone_notes_if_non_empty(footnotes),
199            endnotes: clone_notes_if_non_empty(endnotes),
200            comments: if comments.is_empty() {
201                None
202            } else {
203                Some(comments)
204            },
205        },
206    };
207    model
208}
209
210fn clone_notes_if_non_empty(
211    notes: Vec<DocumentNoteDefinition>,
212) -> Option<Vec<DocumentNoteDefinition>> {
213    if notes.is_empty() {
214        return None;
215    }
216
217    Some(notes)
218}
219
220#[cfg(test)]
221mod tests {
222    use std::collections::HashMap;
223
224    use super::*;
225    use crate::package::OoxmlPart;
226
227    #[test]
228    fn build_doc_model_warns_when_document_missing() {
229        let pkg = OoxmlPackage {
230            parts: HashMap::new(),
231            binary_assets: HashMap::new(),
232            warnings: Vec::new(),
233        };
234        let model = build_doc_model(&pkg);
235        assert_eq!(model.nodes.len(), 1);
236        assert!(model
237            .metadata
238            .warnings
239            .iter()
240            .any(|warning| warning.contains("Missing word/document.xml")));
241    }
242
243    #[test]
244    fn build_doc_model_parses_empty_document() {
245        let mut parts = HashMap::new();
246        parts.insert(
247            "word/document.xml".to_string(),
248            OoxmlPart {
249                name: "word/document.xml".to_string(),
250                content: "<w:document><w:body/></w:document>".to_string(),
251            },
252        );
253        let pkg = OoxmlPackage {
254            parts,
255            binary_assets: HashMap::new(),
256            warnings: Vec::new(),
257        };
258        let model = build_doc_model(&pkg);
259        assert_eq!(model.nodes.len(), 1);
260        assert!(model.metadata.sections.is_some());
261    }
262
263    #[test]
264    fn build_doc_model_preserves_adapter_warnings() {
265        let pkg = OoxmlPackage {
266            parts: HashMap::new(),
267            binary_assets: HashMap::new(),
268            warnings: vec!["Legacy DOC adapter warning".to_string()],
269        };
270        let model = build_doc_model(&pkg);
271        assert!(model
272            .metadata
273            .warnings
274            .iter()
275            .any(|warning| warning == "Legacy DOC adapter warning"));
276    }
277}
278
279#[cfg(test)]
280mod package_preparation_tests {
281    #[test]
282    fn parallel_build_preserves_identity_and_normalized_source_count() {
283        let mut parts = std::collections::HashMap::new();
284        for index in 0..6 {
285            let name = format!("part-{index}.xml");
286            parts.insert(name.clone(), crate::package::OoxmlPart { name, content: format!("<w:root xmlns:w=\"http://schemas.openxmlformats.org/wordprocessingml/2006/main\"><w:t>{}</w:t></w:root>", "text &amp; detail ".repeat(1000)) });
287        }
288        let package = crate::package::OoxmlPackage {
289            parts,
290            binary_assets: std::collections::HashMap::from([("asset".into(), vec![5; 32 * 1024])]),
291            warnings: vec!["prior warning".into()],
292        };
293        let expected_identity = crate::structured::package_identity(&package);
294        let (expected, _) = crate::parse::namespaces::normalized_package(&package);
295        let mut serial = super::build_doc_model_contents(&package);
296        crate::structured::assign_source_ids(&mut serial, &expected_identity);
297        let actual = super::build_doc_model(&package);
298        assert_eq!(actual.metadata.document_id.as_deref(), Some(expected_identity.as_str()));
299        assert_eq!(actual.metadata.source_parts, expected.parts.len() as i64);
300        assert_eq!(serde_json::to_string(&actual).unwrap(), serde_json::to_string(&serial).unwrap());
301    }
302}
303
304#[cfg(all(test, not(target_arch = "wasm32")))]
305mod identity_overlap_tests {
306    use super::*;
307    use std::collections::HashMap;
308    use crate::package::OoxmlPart;
309
310    #[test]
311    fn overlapped_identity_matches_serial_construction() {
312        for asset_size in [0, 65536, 262145] {
313            let document = r#"<w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:body><w:p><w:r><w:t>content</w:t></w:r></w:p></w:body></w:document>"#;
314            let part = OoxmlPart { name: "word/document.xml".into(), content: document.into() };
315            let mut package = OoxmlPackage { parts: HashMap::from([(part.name.clone(), part)]), binary_assets: HashMap::new(), warnings: Vec::new() };
316            package.binary_assets.insert("word/media/payload.bin".into(), (0..asset_size).map(|n| n as u8).collect());
317            let mut serial = build_doc_model_contents(&package);
318            crate::structured::assign_source_ids(&mut serial, &crate::structured::package_identity(&package));
319            let parallel = build_doc_model(&package);
320            assert_eq!(serde_json::to_string(&serial).unwrap(), serde_json::to_string(&parallel).unwrap());
321            package.binary_assets.get_mut("word/media/payload.bin").unwrap().push(7);
322            let changed = build_doc_model(&package);
323            assert_ne!(parallel.metadata.document_id, changed.metadata.document_id);
324        }
325    }
326}