Skip to main content

document_svg/document/
tmx.rs

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