data-beans 0.12.0

use super::*;

fn names(v: &[&str]) -> Vec<Box<str>> {
    v.iter().map(|s| Box::from(*s)).collect()
}

#[test]
fn gene_index_resolves_hgnc_renames_both_ways() {
    // A matrix on the CURRENT HGNC release; a marker panel on an older one. Without the
    // alias tier every one of these silently drops out of the panel.
    let dict = names(&[
        "H4C3", "H2BC21", "H1-0", "MARCHF2", "SEPTIN7", "MTARC1", "CD8A",
    ]);
    let idx = GeneIndex::build(&dict);
    assert_eq!(idx.match_gene("HIST1H4C"), Some(0));
    assert_eq!(idx.match_gene("HIST2H2BE"), Some(1));
    assert_eq!(idx.match_gene("H1F0"), Some(2));
    assert_eq!(idx.match_gene("MARCH2"), Some(3), "rule-based MARCH family");
    assert_eq!(idx.match_gene("SEPT7"), Some(4), "rule-based SEPT family");
    assert_eq!(
        idx.match_gene("MARC1"),
        Some(5),
        "MARC1 is MTARC1, not a MARCH-family member"
    );

    // ...and the reverse: an old-release matrix, a current panel.
    let old = names(&["HIST1H4C", "MARCH2", "SEPT7", "CD8A"]);
    let idx_old = GeneIndex::build(&old);
    assert_eq!(idx_old.match_gene("H4C3"), Some(0));
    assert_eq!(idx_old.match_gene("MARCHF2"), Some(1));
    assert_eq!(idx_old.match_gene("SEPTIN7"), Some(2));

    // The alias tier resolves through the faba `{gene}/count/{track}` row keys and the
    // `ENSG…_SYMBOL` form too, since it runs on the extracted symbol.
    let faba = names(&["ENSG00000197061_H4C3/count/spliced", "CD8A/count/spliced"]);
    let idx_faba = GeneIndex::build(&faba);
    assert_eq!(idx_faba.match_gene("HIST1H4C"), Some(0));

    // A gene with no alias must not be coerced into one.
    assert_eq!(idx.match_gene("ZZZ9"), None);
}

#[test]
fn gene_index_tiers_and_idf() {
    let dict = names(&["ENSG00000153563_CD8A", "MS4A1", "A_FOO", "FOO"]);
    let idx = GeneIndex::build(&dict);

    // exact (case-insensitive)
    assert_eq!(idx.match_gene("ms4a1"), Some(1));
    // symbol = last `_`-segment
    assert_eq!(idx.match_gene("CD8A"), Some(0));
    // exact preferred over an earlier-indexed suffix match (the refinement)
    assert_eq!(idx.match_gene("FOO"), Some(3));
    // unmatched
    assert_eq!(idx.match_gene("ZZZ9"), None);

    // bare ENSG query resolves to the `ENSG…_SYMBOL` row (and the combined
    // form resolves too) — HGNC / ENSG / ENSG_HGNC reconcile both ways.
    assert_eq!(idx.match_gene("ENSG00000153563"), Some(0));
    assert_eq!(idx.match_gene("ENSG00000153563_CD8A"), Some(0));

    // dict keyed by bare ENSG: a combined query still resolves via the
    // leading-id tier.
    let dict2 = names(&["ENSG00000153563", "MS4A1"]);
    let idx2 = GeneIndex::build(&dict2);
    assert_eq!(idx2.match_gene("ENSG00000153563_CD8A"), Some(0));
    assert_eq!(idx2.match_gene("ensg00000153563"), Some(0));

    // IDF: ubiquitous (df == C) → 0; exclusive → ln(C)
    assert_eq!(idf_weight(4, 4), 0.0);
    assert!(idf_weight(4, 1) > 1.38 && idf_weight(4, 1) < 1.39);
}

#[test]
fn a_row_is_named_by_id_alone_when_its_name_adds_nothing() {
    assert_eq!(id_name("FID1", "GENE1"), Box::from("FID1_GENE1"));
    assert_eq!(id_name("FID2", "FID2"), Box::from("FID2"));
    assert_eq!(id_name("FID3", ""), Box::from("FID3"));
}

#[test]
fn loci_match_loci_strictly_by_key() {
    let dict = names(&["X:0-100", "chr1:0-100", "GENE1", "ENSG000_GENE2"]);
    let idx = GeneIndex::build(&dict);
    assert_eq!(idx.match_gene("chrX:0-100"), Some(0));
    assert_eq!(idx.match_gene("CHR1:0-100"), Some(1));
    assert_eq!(idx.match_gene("x:0-100"), None, "chromosome case is kept");
    assert_eq!(idx.match_gene("X:0-101"), None);
    assert_eq!(idx.match_gene("GENE2"), Some(3));
    // A gene query never lands on a locus row through the fallback tiers.
    assert_eq!(idx.match_gene("X"), None);
    assert_eq!(idx.match_gene("chr1"), None);
    // A locus query still reaches a row whose `/`-core is that locus.
    let idx = GeneIndex::build(&names(&["GENE1", "chr2:5-9/count/spliced"]));
    assert_eq!(idx.match_gene("chr2:5-9"), Some(1));
    assert_eq!(idx.match_gene("chr2:5-9/count/spliced"), Some(1));
}

#[test]
fn peak_rows_import_in_colon_form_and_gene_rows_stay() {
    let mut ids = names(&["chr1-100-200", "GENE_1_2", "chrUn_CTG1v1_5_10", "chr2:1-9"]);
    let mut nm = names(&["chr1-100-200", "GENE_1_2", "", "chr2:1-9"]);
    let types = names(&["Peaks", "Gene Expression", "Peaks", "Peaks"]);
    colon_peak_names(&mut ids, &mut nm, Some(&types));
    assert_eq!(
        ids,
        names(&["chr1:100-200", "GENE_1_2", "chrUn_CTG1v1:5-10", "chr2:1-9"])
    );
    assert_eq!(nm, names(&["chr1:100-200", "GENE_1_2", "", "chr2:1-9"]));
}

#[test]
fn an_untyped_list_converts_only_when_every_row_is_an_interval() {
    let mut ids = names(&["chr1-100-200", "chr2_5_9"]);
    let mut nm = names(&["", ""]);
    assert_eq!(colon_peak_names(&mut ids, &mut nm, None), 2);
    assert_eq!(ids, names(&["chr1:100-200", "chr2:5-9"]));
    let mut ids = names(&["chr1-100-200", "GENE1"]);
    let mut nm = names(&["", ""]);
    assert_eq!(colon_peak_names(&mut ids, &mut nm, None), 0);
    assert_eq!(ids, names(&["chr1-100-200", "GENE1"]));
}

#[test]
fn locus_queries_prefer_the_exact_row_and_reach_composite_rows() {
    let idx = GeneIndex::build(&names(&["chr2:5-9/count/spliced", "chr2:5-9"]));
    assert_eq!(idx.match_gene("chr2:5-9"), Some(1), "exact name first");
    assert_eq!(
        idx.match_gene("2:5-9"),
        Some(1),
        "a whole-locus row wins its key"
    );
    assert_eq!(idx.match_gene("chr2:5-9/count/spliced"), Some(0));
    assert_eq!(
        idx.match_gene("2:5-9/count/spliced"),
        Some(0),
        "chr prefix free"
    );
    let idx = GeneIndex::build(&names(&["chrX:0-100", "X:0-100"]));
    assert_eq!(idx.match_gene("X:0-100"), Some(1));
    assert_eq!(idx.match_gene("chrX:0-100"), Some(0));
    // An id_name composite peak row is reached by its locus.
    let idx = GeneIndex::build(&names(&["GENE9", "chr1:100-200_GENE1"]));
    assert_eq!(idx.match_gene("chr1:100-200"), Some(1));
    assert_eq!(idx.match_gene("x:0-100/count/spliced"), None);
}

#[test]
fn a_suffixed_name_never_collides_with_a_name_in_the_list() {
    let mut v = names(&["ABC", "ABC-1", "ABC", "ABC", "X", "ABC-1"]);
    assert_eq!(make_names_unique(&mut v), 2);
    assert_eq!(
        v,
        names(&["ABC", "ABC-1", "ABC-2", "ABC-3", "X", "ABC-1-1"])
    );
    let distinct: HashSet<&str> = v.iter().map(|n| n.as_ref()).collect();
    assert_eq!(distinct.len(), v.len());

    // Plain duplicates number as before.
    let mut v = names(&["A", "B", "A", "A", "B"]);
    assert_eq!(make_names_unique(&mut v), 2);
    assert_eq!(v, names(&["A", "B", "A-1", "A-2", "B-1"]));

    let mut v = names(&["A", "B"]);
    assert_eq!(make_names_unique(&mut v), 0);
    assert_eq!(v, names(&["A", "B"]));
}

#[test]
fn a_locus_id_is_not_split_inside_its_contig_name() {
    for id in [
        "chrUn_KI270742v1:100-200",
        "chr1_KI270706v1_random:5-9",
        "chr1:100-200",
        "chrUn_KI270742v1:100-200/count/spliced",
    ] {
        assert_eq!(split_id_name(id), (id, id), "{id}");
        assert_eq!(split_id_name(&id_name(id, id)), (id, id), "{id}");
    }
    assert_eq!(
        split_id_name(&id_name("chrUn_KI270742v1:100-200", "GENE1")),
        ("chrUn_KI270742v1:100-200", "GENE1")
    );
    // A composite that is one locus as a whole stays whole: a genome-tagged
    // contig of a multi-genome reference, and equally an id joined to a
    // locus name, which reads the same.
    for id in [
        "GRCh38_chr1:100-200",
        "mm10_chr1:100-200",
        "peak1_chr1:100-200",
    ] {
        assert_eq!(split_id_name(id), (id, id), "{id}");
    }
    // Gene rows split on the first separator, as composed.
    assert_eq!(
        split_id_name("ENSG00000153563_CD8A"),
        ("ENSG00000153563", "CD8A")
    );
    assert_eq!(split_id_name("CD8A"), ("CD8A", "CD8A"));
}

#[test]
fn substring_matches_come_back_once_and_blank_queries_match_nothing() {
    let all = names(&["CD3E", "CD3D", "CD8A", "MS4A1"]);

    // A blank line of a name file is not a query for every name.
    let (idx, got) = match_by_substring(&all, &names(&["CD8A", ""]), "row").unwrap();
    assert_eq!(idx, vec![2]);
    assert_eq!(got, names(&["CD8A"]));
    assert!(match_by_substring(&all, &names(&[""]), "row").is_err());

    // Overlapping queries: each name once, in first-matched order.
    let (idx, got) = match_by_substring(&all, &names(&["CD3E", "CD3", "MS4"]), "row").unwrap();
    assert_eq!(idx, vec![0, 1, 3]);
    assert_eq!(got, names(&["CD3E", "CD3D", "MS4A1"]));
}

/// The fallback tier finds the row a scan of the whole vocabulary finds.
#[test]
fn the_fallback_tier_agrees_with_a_full_scan() {
    let parts = ["a", "b", "ab", "c", "", "ensg0001", "x1", "gene"];
    // Names of one to four parts, thinned so that many queries have no row,
    // then a few with an aux suffix or a locus.
    let mut dict: Vec<String> = Vec::new();
    for n in 1..=4usize {
        for code in 0..parts.len().pow(n as u32) {
            let mut c = code;
            let segs: Vec<&str> = (0..n)
                .map(|_| {
                    let p = parts[c % parts.len()];
                    c /= parts.len();
                    p
                })
                .collect();
            if code % 7 < 2 {
                dict.push(segs.join("_"));
            }
        }
    }
    dict.push("A_B/count/spliced".into());
    dict.push("chr1:1-2".into());
    dict.push("chr1:1-2_A".into());
    let dict: Vec<Box<str>> = dict.into_iter().map(String::into_boxed_str).collect();
    let idx = GeneIndex::build(&dict);

    let scan = |gl: &str| {
        let suffix = format!("_{gl}");
        let prefix = format!("{gl}_");
        let middle = format!("_{gl}_");
        idx.lowered
            .iter()
            .position(|t| t.ends_with(&suffix) || t.starts_with(&prefix) || t.contains(&middle))
    };
    let mut queries: Vec<String> = vec![
        String::new(),
        "_".into(),
        "__".into(),
        "zzz".into(),
        "a_".into(),
        "_a".into(),
        "b/count/spliced".into(),
        "chr1:1-2".into(),
        "spliced".into(),
    ];
    for a in parts {
        queries.push(a.to_string());
        for b in parts {
            queries.push(format!("{a}_{b}"));
            for c in parts {
                queries.push(format!("{a}_{b}_{c}"));
            }
        }
    }
    let mut n_hits = 0;
    for q in &queries {
        let want = scan(q);
        assert_eq!(idx.match_segments(q), want, "query `{q}`");
        n_hits += usize::from(want.is_some());
    }
    assert!(n_hits > 50 && n_hits < queries.len(), "{n_hits} hits");
}