rdbms-playground/tests/it/seed.rs

//! Tier-3 integration tests for the `seed` command (ADR-0048, the
//! Phase-1 walking skeleton). Covers the parse path (grammar → AST),
//! the worker round-trip (rows generated + persisted to CSV),
//! reproducibility via a fixed `--seed`, and the single `history.log`
//! line for the whole command (ADR-0048 D15 / U3).

use rdbms_playground::db::Database;
use rdbms_playground::dsl::{ColumnSpec, Command, ReferentialAction, Type, parse_command};
use rdbms_playground::persistence::Persistence;
use rdbms_playground::project;

fn rt() -> tokio::runtime::Runtime {
    tokio::runtime::Builder::new_current_thread()
        .enable_all()
        .build()
        .expect("tokio rt")
}

fn open_project_db() -> (project::Project, Database, tempfile::TempDir) {
    let dir = tempfile::tempdir().expect("create tempdir");
    let project =
        project::open_or_create(None, Some(dir.path())).expect("open or create project");
    let persistence = Persistence::new(project.path().to_path_buf());
    let db = Database::open_with_persistence(project.db_path(), persistence)
        .expect("open db with persistence");
    (project, db, dir)
}

fn read_csv(project: &project::Project, table: &str) -> Option<String> {
    std::fs::read_to_string(project.path().join("data").join(format!("{table}.csv"))).ok()
}

/// `People(id serial pk, name text, email text)` — `id` is autogen
/// (excluded from generation, so no PK collisions), `name`/`email`
/// are generated.
fn create_people(db: &Database, rt: &tokio::runtime::Runtime) {
    rt.block_on(db.create_table(
        "People".to_string(),
        vec![
            ColumnSpec::new("id", Type::Serial),
            ColumnSpec::new("name", Type::Text),
            ColumnSpec::new("email", Type::Text),
        ],
        vec!["id".to_string()],
        None,
    ))
    .expect("create People");
}

/// Data rows in a CSV = non-empty lines minus the header.
fn data_row_count(csv: &str) -> usize {
    csv.lines()
        .filter(|l| !l.trim().is_empty())
        .count()
        .saturating_sub(1)
}

#[test]
fn seed_parses_with_and_without_count() {
    match parse_command("seed People 5").expect("`seed People 5` parses") {
        Command::Seed {
            table,
            count,
            rng_seed,
        } => {
            assert_eq!(table, "People");
            assert_eq!(count, Some(5));
            assert_eq!(rng_seed, None);
        }
        other => panic!("expected Command::Seed, got {other:?}"),
    }
    match parse_command("seed People").expect("`seed People` parses") {
        Command::Seed { table, count, .. } => {
            assert_eq!(table, "People");
            assert_eq!(count, None, "omitted count is None (executor defaults to 20)");
        }
        other => panic!("expected Command::Seed, got {other:?}"),
    }
}

#[test]
fn seed_populates_a_table_and_persists_rows() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    create_people(&db, &rt);

    let result = rt
        .block_on(db.seed("People".into(), Some(7), Some(42), Some("seed People 7".into())))
        .expect("seed succeeds");
    assert_eq!(result.produced, 7);

    let csv = read_csv(&project, "People").expect("People CSV exists after seed");
    assert_eq!(
        data_row_count(&csv),
        7,
        "CSV should hold 7 generated rows:\n{csv}"
    );
    // The generated `email` column produces address-shaped values.
    assert!(csv.contains('@'), "seeded emails should appear in the CSV:\n{csv}");
}

#[test]
fn seed_count_defaults_to_twenty() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    create_people(&db, &rt);

    let result = rt
        .block_on(db.seed("People".into(), None, Some(1), Some("seed People".into())))
        .expect("seed succeeds");
    assert_eq!(result.produced, 20, "omitted count defaults to 20");
    let csv = read_csv(&project, "People").expect("People CSV exists");
    assert_eq!(data_row_count(&csv), 20);
}

#[test]
fn seed_is_reproducible_with_a_fixed_seed() {
    let (p1, db1, _d1) = open_project_db();
    let (p2, db2, _d2) = open_project_db();
    let rt = rt();
    create_people(&db1, &rt);
    create_people(&db2, &rt);

    rt.block_on(db1.seed("People".into(), Some(4), Some(123), Some("seed People 4".into())))
        .expect("seed run 1");
    rt.block_on(db2.seed("People".into(), Some(4), Some(123), Some("seed People 4".into())))
        .expect("seed run 2");

    let csv1 = read_csv(&p1, "People").expect("csv 1");
    let csv2 = read_csv(&p2, "People").expect("csv 2");
    assert_eq!(csv1, csv2, "the same --seed must reproduce identical data");
}

#[test]
fn seed_writes_exactly_one_history_line() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    create_people(&db, &rt);

    rt.block_on(db.seed("People".into(), Some(5), Some(1), Some("seed People 5".into())))
        .expect("seed succeeds");

    let history = std::fs::read_to_string(project.path().join("history.log"))
        .expect("history.log exists");
    let seed_lines = history.lines().filter(|l| l.contains("seed People 5")).count();
    assert_eq!(
        seed_lines, 1,
        "a seed of 5 rows must write exactly one history line:\n{history}"
    );
}

// — FK sampling, empty-parent error, block guard (ADR-0048 D14 / D1) —

/// `Users(id serial pk, name text)` + `Orders(id serial pk, user_id
/// int, total decimal)` with `Orders.user_id -> Users.id`.
fn create_users_and_orders(db: &Database, rt: &tokio::runtime::Runtime, add_fk: bool) {
    rt.block_on(async {
        db.create_table(
            "Users".to_string(),
            vec![
                ColumnSpec::new("id", Type::Serial),
                ColumnSpec::new("name", Type::Text),
            ],
            vec!["id".to_string()],
            None,
        )
        .await
        .expect("create Users");
        db.create_table(
            "Orders".to_string(),
            vec![
                ColumnSpec::new("id", Type::Serial),
                ColumnSpec::new("user_id", Type::Int),
                ColumnSpec::new("total", Type::Decimal),
            ],
            vec!["id".to_string()],
            None,
        )
        .await
        .expect("create Orders");
        if add_fk {
            db.add_relationship(
                None,
                "Users".to_string(),
                vec!["id".to_string()],
                "Orders".to_string(),
                vec!["user_id".to_string()],
                ReferentialAction::NoAction,
                ReferentialAction::NoAction,
                false,
                None,
            )
            .await
            .expect("add Orders->Users FK");
        }
    });
}

/// `user_id` is column index 1 of `Orders(id, user_id, total)`.
fn order_user_ids(csv: &str) -> Vec<String> {
    let mut lines = csv.lines().filter(|l| !l.trim().is_empty());
    lines.next(); // header
    lines
        .map(|l| l.split(',').nth(1).unwrap_or_default().to_string())
        .collect()
}

#[test]
fn seed_fills_foreign_keys_from_existing_parents() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    create_users_and_orders(&db, &rt, true);

    // 5 parents → serial ids 1..=5.
    rt.block_on(db.seed("Users".into(), Some(5), Some(1), Some("seed Users 5".into())))
        .expect("seed Users");
    let res = rt
        .block_on(db.seed("Orders".into(), Some(10), Some(2), Some("seed Orders 10".into())))
        .expect("seed Orders");
    assert_eq!(res.produced, 10, "every child row must insert (valid FK)");

    let csv = read_csv(&project, "Orders").expect("Orders CSV");
    let valid: std::collections::HashSet<String> = (1..=5).map(|i| i.to_string()).collect();
    let user_ids = order_user_ids(&csv);
    assert_eq!(user_ids.len(), 10);
    for uid in &user_ids {
        assert!(
            valid.contains(uid),
            "user_id `{uid}` does not reference an existing parent:\n{csv}"
        );
    }
}

#[test]
fn seed_refuses_when_a_parent_table_is_empty() {
    let (_project, db, _dir) = open_project_db();
    let rt = rt();
    create_users_and_orders(&db, &rt, true);

    // Users is empty — no valid FK can be fabricated.
    let err = rt
        .block_on(db.seed("Orders".into(), Some(3), Some(1), Some("seed Orders 3".into())))
        .expect_err("seed must refuse an empty parent");
    let msg = err.to_string();
    assert!(msg.contains("Users"), "error should name the empty parent: {msg}");
    let lower = msg.to_lowercase();
    assert!(
        lower.contains("no rows") || lower.contains("first"),
        "error should explain how to fix it: {msg}"
    );
}

#[test]
fn seed_refuses_a_not_null_blob_column() {
    let (_project, db, _dir) = open_project_db();
    let rt = rt();
    let mut payload = ColumnSpec::new("payload", Type::Blob);
    payload.not_null = true;
    rt.block_on(db.create_table(
        "Files".to_string(),
        vec![ColumnSpec::new("id", Type::Serial), payload],
        vec!["id".to_string()],
        None,
    ))
    .expect("create Files");

    let err = rt
        .block_on(db.seed("Files".into(), Some(2), Some(1), Some("seed Files 2".into())))
        .expect_err("seed must refuse a NOT NULL blob");
    let msg = err.to_string();
    assert!(
        msg.contains("payload") && msg.to_lowercase().contains("blob"),
        "error should name the un-generatable blob column: {msg}"
    );
}

#[test]
fn seed_omits_a_nullable_blob_column() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    rt.block_on(db.create_table(
        "Files".to_string(),
        vec![
            ColumnSpec::new("id", Type::Serial),
            ColumnSpec::new("name", Type::Text),
            // nullable blob → omitted (→ NULL), seed still succeeds.
            ColumnSpec::new("payload", Type::Blob),
        ],
        vec!["id".to_string()],
        None,
    ))
    .expect("create Files");

    let res = rt
        .block_on(db.seed("Files".into(), Some(3), Some(1), Some("seed Files 3".into())))
        .expect("seed succeeds despite the nullable blob");
    assert_eq!(res.produced, 3);
    let csv = read_csv(&project, "Files").expect("Files CSV");
    assert_eq!(data_row_count(&csv), 3);
}

// — uniqueness, junction distinct-combos, IN-CHECK (D10 / D14 / D17) —

/// The `n`th comma-separated field of each data row (the generated
/// values here never contain commas).
fn nth_column_values(csv: &str, n: usize) -> Vec<String> {
    csv.lines()
        .filter(|l| !l.trim().is_empty())
        .skip(1)
        .map(|l| l.split(',').nth(n).unwrap_or_default().trim().to_string())
        .collect()
}

#[test]
fn seed_keeps_unique_columns_distinct() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    let mut label = ColumnSpec::new("label", Type::Text);
    label.unique = true;
    rt.block_on(db.create_table(
        "Tags".to_string(),
        vec![ColumnSpec::new("id", Type::Serial), label],
        vec!["id".to_string()],
        None,
    ))
    .expect("create Tags");

    let res = rt
        .block_on(db.seed("Tags".into(), Some(8), Some(3), Some("seed Tags 8".into())))
        .expect("seed");
    assert_eq!(res.produced, 8);

    let csv = read_csv(&project, "Tags").expect("Tags CSV");
    let labels = nth_column_values(&csv, 1);
    let distinct: std::collections::HashSet<&String> = labels.iter().collect();
    assert_eq!(distinct.len(), labels.len(), "UNIQUE column has duplicates:\n{csv}");
}

#[test]
fn seed_sequences_identifier_int_columns() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    // `code` is an identifier-named int (D10) but not a constraint —
    // uniqueness comes from the identifier rule.
    rt.block_on(db.create_table(
        "Items".to_string(),
        vec![
            ColumnSpec::new("id", Type::Serial),
            ColumnSpec::new("code", Type::Int),
            ColumnSpec::new("name", Type::Text),
        ],
        vec!["id".to_string()],
        None,
    ))
    .expect("create Items");

    let res = rt
        .block_on(db.seed("Items".into(), Some(5), Some(1), Some("seed Items 5".into())))
        .expect("seed");
    assert_eq!(res.produced, 5);

    let csv = read_csv(&project, "Items").expect("Items CSV");
    let codes: Vec<i64> = nth_column_values(&csv, 1)
        .iter()
        .map(|s| s.parse().expect("code is an int"))
        .collect();
    let distinct: std::collections::HashSet<i64> = codes.iter().copied().collect();
    assert_eq!(distinct.len(), 5, "identifier ints must be unique: {codes:?}");
}

#[test]
fn seed_junction_produces_distinct_combinations_and_caps() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    rt.block_on(async {
        // Two parents, 2 rows each → 2x2 = 4 possible (a, b) pairs.
        for t in ["P1", "P2"] {
            db.create_table(
                t.to_string(),
                vec![
                    ColumnSpec::new("id", Type::Serial),
                    ColumnSpec::new("name", Type::Text),
                ],
                vec!["id".to_string()],
                None,
            )
            .await
            .expect("create parent");
            db.seed(t.into(), Some(2), Some(1), Some(format!("seed {t} 2")))
                .await
                .expect("seed parent");
        }
        // Junction with a compound PK over its two FK columns.
        db.create_table(
            "J".to_string(),
            vec![ColumnSpec::new("a", Type::Int), ColumnSpec::new("b", Type::Int)],
            vec!["a".to_string(), "b".to_string()],
            None,
        )
        .await
        .expect("create J");
        db.add_relationship(
            None,
            "P1".into(),
            vec!["id".into()],
            "J".into(),
            vec!["a".into()],
            ReferentialAction::NoAction,
            ReferentialAction::NoAction,
            false,
            None,
        )
        .await
        .expect("fk a");
        db.add_relationship(
            None,
            "P2".into(),
            vec!["id".into()],
            "J".into(),
            vec!["b".into()],
            ReferentialAction::NoAction,
            ReferentialAction::NoAction,
            false,
            None,
        )
        .await
        .expect("fk b");

        // Requesting 10 caps at the 4 available distinct combinations.
        let res = db
            .seed("J".into(), Some(10), Some(7), Some("seed J 10".into()))
            .await
            .expect("seed J");
        assert_eq!(res.produced, 4, "junction caps at available combos");
        assert_eq!(res.requested, 10, "the requested count is reported for the cap note");
    });

    let csv = read_csv(&project, "J").expect("J CSV");
    let pairs: Vec<String> = csv
        .lines()
        .filter(|l| !l.trim().is_empty())
        .skip(1)
        .map(str::to_string)
        .collect();
    let distinct: std::collections::HashSet<&String> = pairs.iter().collect();
    assert_eq!(distinct.len(), pairs.len(), "junction rows must be distinct:\n{csv}");
}

#[test]
fn seed_draws_enum_values_from_an_in_check() {
    let (project, db, _dir) = open_project_db();
    let rt = rt();
    let mut status = ColumnSpec::new("status", Type::Text);
    status.check_sql = Some("status IN ('active', 'closed')".to_string());
    rt.block_on(db.create_table(
        "Tickets".to_string(),
        vec![ColumnSpec::new("id", Type::Serial), status],
        vec!["id".to_string()],
        None,
    ))
    .expect("create Tickets");

    // Every generated status must satisfy the CHECK, so all rows insert.
    let res = rt
        .block_on(db.seed("Tickets".into(), Some(12), Some(2), Some("seed Tickets 12".into())))
        .expect("seed");
    assert_eq!(res.produced, 12, "all rows insert — values satisfy the CHECK");

    let csv = read_csv(&project, "Tickets").expect("Tickets CSV");
    for v in nth_column_values(&csv, 1) {
        assert!(
            matches!(v.as_str(), "active" | "closed"),
            "status `{v}` was not drawn from the IN check:\n{csv}"
        );
    }
    // The IN-check column is derived, not generic, so it is NOT flagged.
    assert!(
        res.advisory_columns.is_empty(),
        "an IN-check column should not be flagged: {:?}",
        res.advisory_columns
    );
}

#[test]
fn seed_advises_on_enum_ish_columns() {
    let (_project, db, _dir) = open_project_db();
    let rt = rt();
    // `status` has no CHECK and no name heuristic → generic text, so it
    // is flagged for the advisory (D12/D13).
    rt.block_on(db.create_table(
        "Tasks".to_string(),
        vec![
            ColumnSpec::new("id", Type::Serial),
            ColumnSpec::new("status", Type::Text),
        ],
        vec!["id".to_string()],
        None,
    ))
    .expect("create Tasks");

    let res = rt
        .block_on(db.seed("Tasks".into(), Some(3), Some(1), Some("seed Tasks 3".into())))
        .expect("seed");
    assert!(
        res.advisory_columns.contains(&"status".to_string()),
        "enum-ish `status` should be flagged: {:?}",
        res.advisory_columns
    );
}

#[test]
fn seed_refuses_an_excessive_count() {
    let (_project, db, _dir) = open_project_db();
    let rt = rt();
    create_people(&db, &rt);
    let err = rt
        .block_on(db.seed("People".into(), Some(1_000_000), Some(1), Some("seed People 1000000".into())))
        .expect_err("an excessive count must be refused");
    assert!(
        err.to_string().to_lowercase().contains("maximum"),
        "error should mention the maximum: {err}"
    );
}

#[test]
fn seed_preview_is_capped_but_count_is_full() {
    let (_project, db, _dir) = open_project_db();
    let rt = rt();
    create_people(&db, &rt);

    let res = rt
        .block_on(db.seed("People".into(), Some(25), Some(1), Some("seed People 25".into())))
        .expect("seed");
    assert_eq!(res.produced, 25, "the full count is produced");
    assert_eq!(res.data.rows.len(), 20, "the preview is capped at 20 rows");
}