Skip to main content

document_svg/document/
tbx.rs

1//! Bounded TermBase eXchange (TBX/MARTIF) previews.
2//!
3//! Terminology entries often contain proprietary definitions and references.
4//! This adapter validates the TBX core structure and renders element counts and
5//! language metadata while keeping term text, identifiers and links inert.
6
7use std::collections::BTreeSet;
8use std::path::Path;
9
10use crate::convert::{ConvertOptions, PageConsumer, read_limited_file};
11use crate::document::html::{HtmlBlock, render_blocks_to_pages};
12use crate::error::{Error, Result};
13use crate::geospatial::xml_tree::{XmlElement, XmlLimits, parse_xml_tree};
14use crate::table::{TableAlign, TableData};
15
16const MAX_TBX_BYTES: u64 = 128 * 1024 * 1024;
17const MAX_TBX_EVENTS: usize = 1_000_000;
18const MAX_TBX_NODES: usize = 500_000;
19const MAX_TBX_DEPTH: usize = 128;
20const MAX_TBX_TEXT_BYTES: usize = 32 * 1024 * 1024;
21const MAX_TBX_ROWS: usize = 200_000;
22const MAX_TBX_LANGUAGES: usize = 10_000;
23const MAX_TBX_DISPLAY_BYTES: usize = 512;
24
25#[derive(Default)]
26struct Summary {
27    dialect: String,
28    entries: usize,
29    terms: usize,
30    language_sets: usize,
31    term_items: usize,
32    descriptions: usize,
33    administrative: usize,
34    notes: usize,
35    transactions: usize,
36    references: usize,
37    languages: BTreeSet<String>,
38    rows: Vec<Vec<String>>,
39}
40
41struct TbxPageSink<'a> {
42    inner: &'a mut dyn PageConsumer,
43    warnings: &'a [String],
44}
45
46impl PageConsumer for TbxPageSink<'_> {
47    fn consume(&mut self, mut page: crate::ir::Page) -> Result<()> {
48        page.source_format = "tbx".into();
49        if page.title.is_empty() {
50            page.title = "TBX terminology base".into();
51        }
52        page.description =
53            "TBX terminology structure is rendered as bounded inert metadata; term text and external resources are not exposed".into();
54        for warning in self.warnings {
55            page.warn(warning.clone());
56        }
57        self.inner.consume(page)
58    }
59}
60
61pub(crate) fn looks_like_prefix(prefix: &[u8]) -> bool {
62    crate::geospatial::xml_tree::looks_like_root(prefix, b"tbx", None)
63        || crate::geospatial::xml_tree::looks_like_root(prefix, b"martif", None)
64}
65
66pub(crate) fn convert(
67    path: &Path,
68    options: &ConvertOptions,
69    sink: &mut dyn PageConsumer,
70) -> Result<Vec<String>> {
71    let bytes = read_limited_file(
72        path,
73        options.max_input_bytes.min(MAX_TBX_BYTES),
74        "TBX input",
75    )?;
76    let root = parse_xml_tree(
77        &bytes,
78        &XmlLimits {
79            max_events: options.max_xml_events.min(MAX_TBX_EVENTS),
80            max_nodes: MAX_TBX_NODES,
81            max_depth: MAX_TBX_DEPTH,
82            max_text_bytes: MAX_TBX_TEXT_BYTES,
83        },
84        "TBX",
85    )?;
86    if !root.name.eq_ignore_ascii_case("martif") && !root.name.eq_ignore_ascii_case("tbx") {
87        return Err(Error::InvalidInput(
88            "TBX root must be <tbx> or legacy <martif>".into(),
89        ));
90    }
91    let dialect = attr_local(&root, "type")
92        .or_else(|| attr_local(&root, "style"))
93        .unwrap_or(root.name.as_str());
94    let mut summary = Summary {
95        dialect: truncate(dialect),
96        entries: count_named(&root, "termEntry"),
97        terms: count_named(&root, "term"),
98        language_sets: count_named(&root, "langSet"),
99        term_items: count_named(&root, "tig") + count_named(&root, "termSec"),
100        descriptions: count_named(&root, "descrip") + count_named(&root, "definition"),
101        administrative: count_named(&root, "admin"),
102        notes: count_named(&root, "note"),
103        transactions: count_named(&root, "transac") + count_named(&root, "transacGrp"),
104        references: count_named(&root, "ref") + count_named(&root, "xref"),
105        ..Summary::default()
106    };
107    collect_languages(&root, &mut summary.languages)?;
108    push_row(
109        &mut summary.rows,
110        "Document",
111        &root.name,
112        &format!("dialect={}", summary.dialect),
113    )?;
114    push_row(
115        &mut summary.rows,
116        "Terminology",
117        &summary.entries.to_string(),
118        &format!(
119            "terms={} languageSets={}",
120            summary.terms, summary.language_sets
121        ),
122    )?;
123    push_row(
124        &mut summary.rows,
125        "Languages",
126        &summary.languages.len().to_string(),
127        &display_languages(&summary.languages),
128    )?;
129    push_row(
130        &mut summary.rows,
131        "Metadata",
132        &format!(
133            "termItems={} descriptions={}",
134            summary.term_items, summary.descriptions
135        ),
136        &format!(
137            "admin={} notes={} transactions={} references={}",
138            summary.administrative, summary.notes, summary.transactions, summary.references
139        ),
140    )?;
141    let blocks = vec![
142        HtmlBlock::Heading {
143            level: 1,
144            text: "TBX terminology base".into(),
145        },
146        HtmlBlock::Paragraph {
147            text: "TermBase eXchange entries and language structure are summarized without displaying proprietary terminology text.".into(),
148        },
149        HtmlBlock::Table(TableData {
150            headers: vec!["Kind".into(), "Value".into(), "Detail".into()],
151            rows: summary.rows,
152            alignments: vec![TableAlign::Left; 3],
153            raw_source: String::new(),
154        }),
155    ];
156    let warnings = vec![
157        "TBX terms, definitions, IDs, notes, references, URLs and proprietary data categories are omitted or redacted; only bounded counts and language tags are shown".into(),
158        "TBX DTDs, XCS/Schematron resources, external entities, references and terminology services are never loaded or executed".into(),
159    ];
160    let mut page_sink = TbxPageSink {
161        inner: sink,
162        warnings: &warnings,
163    };
164    render_blocks_to_pages(&blocks, &mut page_sink, options)?;
165    Ok(warnings)
166}
167
168fn collect_languages(element: &XmlElement, languages: &mut BTreeSet<String>) -> Result<()> {
169    if (element.name.eq_ignore_ascii_case("langSet") || element.name.eq_ignore_ascii_case("lang"))
170        && let Some(language) =
171            attr_local(element, "lang").or_else(|| attr_local(element, "xml:lang"))
172    {
173        let language = truncate(language);
174        if !language.is_empty() {
175            if languages.len() >= MAX_TBX_LANGUAGES && !languages.contains(&language) {
176                return Err(Error::LimitExceeded(format!(
177                    "TBX language tags exceed {MAX_TBX_LANGUAGES}"
178                )));
179            }
180            languages.insert(language);
181        }
182    }
183    for child in &element.children {
184        collect_languages(child, languages)?;
185    }
186    Ok(())
187}
188
189fn count_named(element: &XmlElement, name: &str) -> usize {
190    element
191        .children
192        .iter()
193        .map(|child| usize::from(child.name.eq_ignore_ascii_case(name)) + count_named(child, name))
194        .sum()
195}
196
197fn attr_local<'a>(element: &'a XmlElement, name: &str) -> Option<&'a str> {
198    element.attributes.iter().find_map(|(key, value)| {
199        key.rsplit(':')
200            .next()
201            .filter(|local| local.eq_ignore_ascii_case(name))
202            .map(|_| value.as_str())
203    })
204}
205
206fn display_languages(languages: &BTreeSet<String>) -> String {
207    if languages.is_empty() {
208        return "none".into();
209    }
210    let mut values = languages.iter().take(16).cloned().collect::<Vec<_>>();
211    if languages.len() > values.len() {
212        values.push(format!("+{} more", languages.len() - values.len()));
213    }
214    values.join(", ")
215}
216
217fn push_row(rows: &mut Vec<Vec<String>>, kind: &str, value: &str, detail: &str) -> Result<()> {
218    if rows.len() >= MAX_TBX_ROWS {
219        return Err(Error::LimitExceeded(format!(
220            "TBX rows exceed {MAX_TBX_ROWS}"
221        )));
222    }
223    rows.push(vec![truncate(kind), truncate(value), truncate(detail)]);
224    Ok(())
225}
226
227fn truncate(value: &str) -> String {
228    if value.len() <= MAX_TBX_DISPLAY_BYTES {
229        value.to_owned()
230    } else {
231        let mut end = MAX_TBX_DISPLAY_BYTES;
232        while end > 0 && !value.is_char_boundary(end) {
233            end -= 1;
234        }
235        format!("{}…", &value[..end])
236    }
237}