open-standards/wai/wai-rs/tests/audio_mark.rs

//! The audio scheme of the second marking layer, `wai-mark:band64` (extensions/synthesis.md
//! §7.10), and its corpus `live-conformance/audio-mark/`: every generated file regenerates byte
//! for byte and no other file but the measured records is there; every embedding, detection,
//! case and record gets what it states; §7.10.1's definition, §7.10.3's shuffle and §7.10.6's
//! bound are the reference's; and §7.10.7's tables are the measurements' and its claims attacks
//! every measurement shows survived.

use serde_json::Value;
use std::collections::BTreeMap;
use std::path::PathBuf;
use wai::audio_mark::corpus;
use wai::synthesis::mark::audio as band;

fn wai_dir() -> PathBuf {
    PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("..")
}

fn tree(root: &PathBuf) -> BTreeMap<String, Vec<u8>> {
    let mut out = BTreeMap::new();
    let mut stack = vec![root.clone()];
    while let Some(d) = stack.pop() {
        for e in std::fs::read_dir(&d).expect("the corpus directory") {
            let p = e.expect("an entry").path();
            if p.is_dir() {
                stack.push(p);
                continue;
            }
            let rel = p.strip_prefix(root).expect("under the root").to_string_lossy().replace('\\', "/");
            if rel != "README.md" {
                out.insert(rel, std::fs::read(&p).expect("a file"));
            }
        }
    }
    out
}

fn committed() -> BTreeMap<String, Vec<u8>> {
    tree(&wai_dir().join("live-conformance/audio-mark"))
}

fn section(text: &str, head: &str) -> String {
    let at = text.find(&format!("\n{head}")).unwrap_or_else(|| panic!("a section {head}"));
    let level = head.chars().take_while(|c| *c == '#').count();
    let end = text[at + 1..]
        .match_indices("\n#")
        .find(|(i, _)| text[at + 1 + i + 1..].chars().take_while(|c| *c == '#').count() <= level)
        .map_or(text.len(), |(i, _)| at + 1 + i);
    text[at..end].to_string()
}

fn fold(s: &str) -> String {
    s.split_whitespace().collect::<Vec<_>>().join(" ")
}

fn synthesis() -> String {
    std::fs::read_to_string(wai_dir().join("extensions/synthesis.md")).expect("synthesis.md")
}

#[test]
fn the_committed_corpus_regenerates_from_its_seeds() {
    let generated: BTreeMap<String, Vec<u8>> = corpus::generate().into_iter().collect();
    let files = committed();
    for k in generated.keys() {
        assert!(files.get(k) == generated.get(k), "{k} differs from its regeneration, or is missing");
    }
    for k in files.keys() {
        assert!(generated.contains_key(k) || k.starts_with("robustness/measured/"), "{k} is not generated");
    }
}

#[test]
fn every_vector_case_and_record_gets_what_it_states() {
    let fails = corpus::verify(&committed());
    assert!(fails.is_empty(), "{fails:#?}");
}

#[test]
fn the_extension_states_the_references_definition_shuffle_and_bound() {
    let s = synthesis();
    let def = section(&s, "#### 7.10.1 ");
    let block = def.split("```").nth(1).expect("the definition's block");
    let joined: String = block.lines().map(str::trim).collect();
    assert_eq!(joined, band::definition_text(), "§7.10.1's definition is the reference's");
    let digest: String = band::digest().iter().map(|b| format!("{b:02x}")).collect();
    assert!(fold(&def).contains(&format!("its digest is `{digest}`")), "§7.10.1 states the digest");
    let lay = band::layout();
    let head = format!("π begins {}, {}, {}, {}", lay.perm[0], lay.perm[1], lay.perm[2], lay.perm[3]);
    assert!(fold(&section(&s, "#### 7.10.3 ")).contains(&head), "§7.10.3 states the shuffle's first entries");
    let bound = corpus::false_positive_log2(corpus::ATTEMPTS);
    assert!(fold(&section(&s, "#### 7.10.6 ")).contains(&format!("2^−{}", bound.trim_start_matches('-'))), "§7.10.6 states {bound}");
}

/// §7.10.7's tables, by their first data column's header: attack → column → cell.
fn tables(s: &str) -> BTreeMap<String, BTreeMap<String, BTreeMap<String, String>>> {
    let mut out: BTreeMap<String, BTreeMap<String, BTreeMap<String, String>>> = BTreeMap::new();
    let mut head: Option<Vec<String>> = None;
    for line in s.lines() {
        let cells: Vec<String> = line.trim().trim_matches('|').split('|').map(|c| c.trim().trim_matches('`').to_string()).collect();
        if line.starts_with("| attack") {
            out.entry(cells[1].clone()).or_default();
            head = Some(cells);
        } else if let (Some(h), true) = (&head, line.starts_with("| `")) {
            let row: BTreeMap<String, String> = h[1..].iter().cloned().zip(cells[1..].iter().cloned()).collect();
            out.get_mut(&h[1]).expect("a table").insert(cells[0].clone(), row);
        } else if !line.starts_with('|') {
            head = None;
        }
    }
    out
}

#[test]
fn the_tables_are_the_measurements_and_the_claims_were_survived() {
    let files = committed();
    let mut rows: BTreeMap<(String, String), Value> = BTreeMap::new();
    let mut add = |b: &[u8]| {
        let v: Value = serde_json::from_slice(b).expect("a measurement");
        for r in v["table"].as_array().expect("a table") {
            rows.insert((r["attack"].as_str().unwrap_or("").into(), r["set"].as_str().unwrap_or("").into()), r.clone());
        }
    };
    add(&files["battery-band64.json"]);
    for (k, b) in &files {
        if k.starts_with("robustness/measured/measurement-") {
            add(b);
        }
    }
    let text = section(&synthesis(), "#### 7.10.7 ");
    let t = tables(&text);
    assert!(t.contains_key("voice"), "§7.10.7 has the pinned attacks' table");
    for table in t.values() {
        for (attack, cells) in table {
            for (col, cell) in cells {
                let r = rows.get(&(attack.clone(), col.clone())).unwrap_or_else(|| panic!("no measurement of {attack} on {col}"));
                let mut want = format!("{} / {}", r["survived"], r["items"]);
                if r["survived"].as_u64() != Some(0) {
                    want += &format!(" ({})", r["errors_max"]);
                }
                assert_eq!(cell, &want, "§7.10.7 {attack}, {col}");
            }
        }
    }
    let f = fold(&text);
    let claims = f.split("**claims to survive** ").nth(1).and_then(|x| x.split('.').next()).expect("§7.10.7's claims");
    for a in claims.split('`').skip(1).step_by(2) {
        let sets: Vec<&Value> = rows.iter().filter(|((at, _), _)| at == a).map(|(_, r)| r).collect();
        assert!(!sets.is_empty(), "{a} is measured");
        for r in sets {
            assert!(r["survived"] == r["items"] && r["false_positives"] == 0, "§7.10.7 claims {a}, which {} did not survive", r["set"]);
        }
    }
}