sdsforge 0.5.0

CLI and desktop app for generating, converting, validating, and rendering chemical safety data sheets in MHLW/JIS Z 7253 structured formats.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268

/// Source-text evidence extraction and JSON coverage matching.
///
/// Extracts chemical identifiers (CAS, H/P-codes, UN numbers) and signal words
/// from raw SDS text, then measures how well the generated JSON reflects them.
/// Used by `eval-corpus` to compute evidence_coverage without requiring gold labels.

// ---------------------------------------------------------------------------
// Types
// ---------------------------------------------------------------------------

#[derive(Debug, Default, serde::Serialize)]
pub struct Evidence {
    pub cas: Vec<String>,
    pub h_codes: Vec<String>,
    pub p_codes: Vec<String>,
    pub un_numbers: Vec<String>,
    pub signal_words: Vec<String>,
}

#[derive(Debug, Default, serde::Serialize)]
pub struct EvidenceCoverage {
    /// Fraction of source CAS numbers found anywhere in the JSON (0.0–1.0).
    pub cas: f32,
    /// Fraction of source H-codes found in the JSON.
    pub h_codes: f32,
    /// Fraction of source P-codes found in the JSON.
    pub p_codes: f32,
    /// Fraction of source UN numbers found in the JSON.
    pub un_numbers: f32,
}

// ---------------------------------------------------------------------------
// Extraction
// ---------------------------------------------------------------------------

/// Extract chemical identifiers and signal words from raw SDS text.
pub fn extract_evidence(text: &str) -> Evidence {
    Evidence {
        cas:          extract_cas(text),
        h_codes:      extract_h_codes(text),
        p_codes:      extract_p_codes(text),
        un_numbers:   extract_un_numbers(text),
        signal_words: extract_signal_words(text),
    }
}

/// CAS: digits-digits-digit (e.g. "67-56-1", "1336-21-6")
fn extract_cas(text: &str) -> Vec<String> {
    let mut out = Vec::new();
    let bytes = text.as_bytes();
    let len = bytes.len();
    let mut i = 0;
    while i < len {
        // Find a digit
        if !bytes[i].is_ascii_digit() { i += 1; continue; }
        // Scan digits (2–7)
        let a_start = i;
        while i < len && bytes[i].is_ascii_digit() { i += 1; }
        let a_len = i - a_start;
        if a_len < 2 || a_len > 7 { continue; }
        // Expect '-'
        if i >= len || bytes[i] != b'-' { continue; }
        i += 1;
        // Expect 2 digits
        let b_start = i;
        while i < len && bytes[i].is_ascii_digit() { i += 1; }
        if i - b_start != 2 { continue; }
        // Expect '-'
        if i >= len || bytes[i] != b'-' { continue; }
        i += 1;
        // Expect 1 digit
        if i >= len || !bytes[i].is_ascii_digit() { continue; }
        i += 1;
        // Must not be followed by a digit (avoid matching in larger numbers)
        if i < len && bytes[i].is_ascii_digit() { continue; }
        // Must not be preceded by a digit at a_start-1
        if a_start > 0 && bytes[a_start - 1].is_ascii_digit() { continue; }
        let cas = &text[a_start..i];
        if !out.contains(&cas.to_string()) {
            out.push(cas.to_string());
        }
    }
    out
}

/// H-codes: H followed by 3 digits, optionally joined by '+' (e.g. H225, H301+H311)
fn extract_h_codes(text: &str) -> Vec<String> {
    let mut out = Vec::new();
    let chars: Vec<char> = text.chars().collect();
    let len = chars.len();
    let mut i = 0;
    while i < len {
        if chars[i] != 'H' && chars[i] != 'h' { i += 1; continue; }
        // Preceded by word char? Skip (e.g. "REACH" shouldn't match)
        if i > 0 && (chars[i-1].is_alphanumeric() || chars[i-1] == '_') { i += 1; continue; }
        let start = i;
        i += 1;
        // 3 digits
        let d_start = i;
        while i < len && chars[i].is_ascii_digit() { i += 1; }
        if i - d_start != 3 { continue; }
        // Check it's followed by a non-alphanumeric (or end)
        if i < len && chars[i].is_alphanumeric() { continue; }
        let code = chars[start..i].iter().collect::<String>().to_uppercase();
        if !out.contains(&code) {
            out.push(code);
        }
    }
    out
}

/// P-codes: P followed by 3 digits, optionally joined by '+' (e.g. P260, P301+P330)
fn extract_p_codes(text: &str) -> Vec<String> {
    let mut out = Vec::new();
    let chars: Vec<char> = text.chars().collect();
    let len = chars.len();
    let mut i = 0;
    while i < len {
        if chars[i] != 'P' && chars[i] != 'p' { i += 1; continue; }
        if i > 0 && (chars[i-1].is_alphanumeric() || chars[i-1] == '_') { i += 1; continue; }
        let start = i;
        i += 1;
        let d_start = i;
        while i < len && chars[i].is_ascii_digit() { i += 1; }
        if i - d_start != 3 { continue; }
        if i < len && chars[i].is_alphanumeric() { continue; }
        let code = chars[start..i].iter().collect::<String>().to_uppercase();
        if !out.contains(&code) {
            out.push(code);
        }
    }
    out
}

/// UN numbers: UN followed by optional whitespace and 4 digits (e.g. "UN 1230", "UN1090")
fn extract_un_numbers(text: &str) -> Vec<String> {
    let mut out = Vec::new();
    let bytes = text.as_bytes();
    let len = bytes.len();
    let mut i = 0;
    while i < len {
        // Match "UN" (case-insensitive)
        if i + 1 < len
            && (bytes[i] == b'U' || bytes[i] == b'u')
            && (bytes[i+1] == b'N' || bytes[i+1] == b'n')
        {
            // Not preceded by alphanumeric
            if i > 0 && bytes[i-1].is_ascii_alphanumeric() { i += 1; continue; }
            let mut j = i + 2;
            // Optional whitespace / separator
            while j < len && (bytes[j] == b' ' || bytes[j] == b'\t' || bytes[j] == b':' || bytes[j] == b'.') {
                j += 1;
            }
            // Expect exactly 4 digits
            let d_start = j;
            while j < len && bytes[j].is_ascii_digit() { j += 1; }
            if j - d_start == 4 {
                // Not followed by digit
                if j < len && bytes[j].is_ascii_digit() { i += 1; continue; }
                let num = std::str::from_utf8(&bytes[d_start..j]).unwrap_or("").to_string();
                if !out.contains(&num) {
                    out.push(num);
                }
                i = j;
                continue;
            }
        }
        i += 1;
    }
    out
}

const SIGNAL_PATTERNS: &[&str] = &[
    "危険", "警告", "Danger", "danger", "DANGER",
    "Warning", "warning", "WARNING",
    "危险", "警告", // zh-cn/zh-tw (same kanji)
    "Not classified", "not classified",
    "Not applicable",
];

fn extract_signal_words(text: &str) -> Vec<String> {
    let mut out = Vec::new();
    for &pat in SIGNAL_PATTERNS {
        if text.contains(pat) && !out.iter().any(|s: &String| s.eq_ignore_ascii_case(pat)) {
            out.push(pat.to_string());
        }
    }
    out
}

// ---------------------------------------------------------------------------
// Coverage matching
// ---------------------------------------------------------------------------

/// Compute what fraction of each evidence category appears somewhere in the JSON.
pub fn match_evidence(ev: &Evidence, json: &serde_json::Value) -> EvidenceCoverage {
    let json_str = json.to_string().to_uppercase();
    EvidenceCoverage {
        cas:       coverage_ratio(&ev.cas,        |s| json_str.contains(&s.to_uppercase())),
        h_codes:   coverage_ratio(&ev.h_codes,    |s| json_str.contains(s)),
        p_codes:   coverage_ratio(&ev.p_codes,    |s| json_str.contains(s)),
        un_numbers:coverage_ratio(&ev.un_numbers, |s| json_str.contains(s)),
    }
}

fn coverage_ratio<F: Fn(&str) -> bool>(items: &[String], found: F) -> f32 {
    if items.is_empty() { return 1.0; } // nothing to check = perfect
    let matched = items.iter().filter(|s| found(s)).count();
    matched as f32 / items.len() as f32
}

// ---------------------------------------------------------------------------
// Tests
// ---------------------------------------------------------------------------

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn cas_basic() {
        let ev = extract_evidence("成分: Methanol CAS 67-56-1, Ethanol 64-17-5.");
        assert!(ev.cas.contains(&"67-56-1".to_string()), "got: {:?}", ev.cas);
        assert!(ev.cas.contains(&"64-17-5".to_string()), "got: {:?}", ev.cas);
    }

    #[test]
    fn h_p_codes() {
        let ev = extract_evidence("H225, H301+H311\nP260, P301+P330+P331");
        assert!(ev.h_codes.contains(&"H225".to_string()));
        assert!(ev.h_codes.contains(&"H301".to_string()));
        assert!(ev.p_codes.contains(&"P260".to_string()));
        assert!(ev.p_codes.contains(&"P301".to_string()));
    }

    #[test]
    fn un_number() {
        let ev = extract_evidence("UN 1230  UN1090 UN: 3077");
        assert!(ev.un_numbers.contains(&"1230".to_string()), "{:?}", ev.un_numbers);
        assert!(ev.un_numbers.contains(&"1090".to_string()), "{:?}", ev.un_numbers);
        assert!(ev.un_numbers.contains(&"3077".to_string()), "{:?}", ev.un_numbers);
    }

    #[test]
    fn signal_words() {
        let ev = extract_evidence("Signal word: Danger.\nWARNING for storage.");
        assert!(ev.signal_words.iter().any(|s| s.to_uppercase() == "DANGER"));
    }

    #[test]
    fn coverage_empty_source_is_perfect() {
        let ev = Evidence::default();
        let cov = match_evidence(&ev, &serde_json::Value::Null);
        assert_eq!(cov.cas, 1.0);
        assert_eq!(cov.h_codes, 1.0);
    }

    #[test]
    fn coverage_partial() {
        let ev = Evidence {
            cas: vec!["67-56-1".into(), "99-99-9".into()],
            ..Default::default()
        };
        let json = serde_json::json!({"CASno": "67-56-1"});
        let cov = match_evidence(&ev, &json);
        assert!((cov.cas - 0.5).abs() < 0.01, "expected 0.5, got {}", cov.cas);
    }
}