rdbms-playground/src/dsl/grammar/data.rs

//! Data command nodes (ADR-0024 §migration Phase D).
//!
//! Five commands at four entry words: `show` (show data /
//! show table), `insert`, `update`, `delete`. The walker route
//! owns these end-to-end.
//!
//! Phase D scope deviation note: ADR-0024's Phase D describes
//! "full schema awareness" via `DynamicSubgrammar
//! (column_value_list)` that unfolds typed slots per column. This
//! milestone lands the data commands at functional parity with
//! the existing chumsky parser — value slots accept any
//! literal regardless of column type, with type validation
//! happening at bind time (matching today's behaviour). The
//! `DynamicSubgrammar` machinery and schema-cache plumbing are
//! deferred to a follow-up refinement; the trie shape is
//! ready to consume them when the schema reference flows
//! through `parse_command`.

use crate::dsl::command::{Command, RowFilter};
use crate::dsl::grammar::{
    CommandNode, HintMode, IdentSource, Node, ValidationError, Word,
    shared::{column_value_list, current_column_value},
};
use crate::dsl::value::Value;
use crate::dsl::walker::outcome::{MatchedItem, MatchedKind, MatchedPath};

// =================================================================
// Building blocks
// =================================================================

const TABLE_NAME_EXISTING: Node = Node::Ident {
    source: IdentSource::Tables,
    role: "table_name",
    validator: None,
    highlight_override: None,
    writes_table: false,
    writes_column: false,
        writes_user_listed_column: false,
};

/// Table-name slot variant that populates
/// `WalkContext::current_table_columns` (ADR-0024 §Phase D).
/// Used by `insert into <T> …` so the inner value list can
/// dispatch typed slots per column.
const TABLE_NAME_INSERT: Node = Node::Ident {
    source: IdentSource::Tables,
    role: "table_name",
    validator: None,
    highlight_override: None,
    writes_table: true,
    writes_column: false,
        writes_user_listed_column: false,
};

// `value_literal` — null / true / false / number / string. The
// chumsky-side equivalent (`value_literal()` in dsl/parser.rs).
const VALUE_LITERAL_CHOICES: &[Node] = &[
    Node::Word(Word::keyword("null")),
    Node::Word(Word::keyword("true")),
    Node::Word(Word::keyword("false")),
    Node::NumberLit { validator: None },
    Node::StringLit,
];
const VALUE_LITERAL_INNER: Node = Node::Choice(VALUE_LITERAL_CHOICES);
/// Value-literal slot with the `ProseOnly` HintMode
/// (ADR-0024 §HintMode-per-node) — the hint resolver surfaces
/// the generic "Type a value: …" prose rather than the
/// misleading `null`/`true`/`false` candidate trio.
const VALUE_LITERAL: Node = Node::Hinted {
    mode: HintMode::ProseOnly("hint.value_literal_slot"),
    inner: &VALUE_LITERAL_INNER,
};

// =================================================================
// show — `show (data|table) <T>`
// =================================================================

const SHOW_DATA_NODES: &[Node] = &[
    Node::Word(Word::keyword("data")),
    TABLE_NAME_EXISTING,
];
const SHOW_DATA: Node = Node::Seq(SHOW_DATA_NODES);

const SHOW_TABLE_NODES: &[Node] = &[
    Node::Word(Word::keyword("table")),
    TABLE_NAME_EXISTING,
];
const SHOW_TABLE: Node = Node::Seq(SHOW_TABLE_NODES);

const SHOW_CHOICES: &[Node] = &[SHOW_DATA, SHOW_TABLE];
const SHOW_SHAPE: Node = Node::Choice(SHOW_CHOICES);

// =================================================================
// insert — `insert into <T> (<a>,<b>,…) values (<v>,<v>,…)`
//        | `insert into <T> values (<v>,…)`
//        | `insert into <T> (<v>,…)`
// =================================================================
//
// Forms A (with column list) and C (bare value list) both start
// with `(`. To avoid the walker's "first commit wins" semantics
// rejecting Form C when the inner content is values rather than
// column names, the inside of the first paren is parsed as a
// repeated `Choice(Ident, ValueLiteral)`. The AST builder then
// disambiguates: if a `values` keyword follows the first paren,
// the inner content was column names; otherwise it was values.

const INSERT_PAREN_ITEM_CHOICES: &[Node] = &[
    // VALUE_LITERAL first so that `true`/`false`/`null` match
    // their Word branch rather than the broader Ident{Columns}
    // catch-all (consume_ident doesn't filter against the
    // keyword set; without this ordering, `(true)` would lex
    // as a column-name list).
    VALUE_LITERAL,
    Node::Ident {
        source: IdentSource::Columns,
        role: "insert_first_item",
        validator: None,
        highlight_override: None,
        writes_table: false,
        writes_column: false,
        // Form A signal: when the user lists explicit columns
        // in `insert into <T> (col1, col2, …)`, the walker
        // appends each matched name to
        // `WalkContext::user_listed_columns`. The inner
        // `values (…)` slot list then mirrors that user
        // selection instead of the auto-filtered default
        // (ADR-0024 §Phase D §column_value_list).
        writes_user_listed_column: true,
    },
];
const INSERT_PAREN_ITEM: Node = Node::Choice(INSERT_PAREN_ITEM_CHOICES);
const INSERT_PAREN_LIST: Node = Node::Repeated {
    inner: &INSERT_PAREN_ITEM,
    separator: Some(&Node::Punct(',')),
    min: 1,
};

/// Schema-aware value list: when the walker has a populated
/// `current_table_columns`, unfolds to a `Seq` of typed slots
/// per column (`int_slot`, `text_slot`, …). When schemaless,
/// falls back to the pre-Phase-D `Repeated(VALUE_LITERAL, ',', 1)`
/// shape (ADR-0024 §Phase D §column_value_list).
const INSERT_VALUES_LIST: Node = Node::DynamicSubgrammar(column_value_list);

const INSERT_OPTIONAL_VALUES_NODES: &[Node] = &[
    Node::Word(Word::keyword("values")),
    Node::Punct('('),
    INSERT_VALUES_LIST,
    Node::Punct(')'),
];
const INSERT_OPTIONAL_VALUES: Node = Node::Optional(&Node::Seq(INSERT_OPTIONAL_VALUES_NODES));

const INSERT_PAREN_FIRST_NODES: &[Node] = &[
    Node::Punct('('),
    INSERT_PAREN_LIST,
    Node::Punct(')'),
    INSERT_OPTIONAL_VALUES,
];
const INSERT_PAREN_FIRST: Node = Node::Seq(INSERT_PAREN_FIRST_NODES);

const INSERT_VALUES_KEYWORD_FIRST_NODES: &[Node] = &[
    Node::Word(Word::keyword("values")),
    Node::Punct('('),
    INSERT_VALUES_LIST,
    Node::Punct(')'),
];
const INSERT_VALUES_KEYWORD_FIRST: Node = Node::Seq(INSERT_VALUES_KEYWORD_FIRST_NODES);

const INSERT_AFTER_TABLE_CHOICES: &[Node] =
    &[INSERT_VALUES_KEYWORD_FIRST, INSERT_PAREN_FIRST];
const INSERT_AFTER_TABLE: Node = Node::Choice(INSERT_AFTER_TABLE_CHOICES);

const INSERT_NODES: &[Node] = &[
    Node::Word(Word::keyword("into")),
    TABLE_NAME_INSERT,
    INSERT_AFTER_TABLE,
];
const INSERT_SHAPE: Node = Node::Seq(INSERT_NODES);

// =================================================================
// update — `update <T> set <col>=<v>[, <col>=<v>] (where … | --all-rows)`
// =================================================================

/// Table-name slot that populates `current_table_columns` so
/// the inner `set <col>=<value>` / `where <col>=<value>` slots
/// can resolve column types (Phase D).
const TABLE_NAME_WRITES: Node = Node::Ident {
    source: IdentSource::Tables,
    role: "table_name",
    validator: None,
    highlight_override: None,
    writes_table: true,
    writes_column: false,
        writes_user_listed_column: false,
};

/// Column-name slot in `set col = …` — resolves the column's
/// type into `current_column` so the value slot dispatches per
/// column type (Phase D).
const SET_COLUMN: Node = Node::Ident {
    source: IdentSource::Columns,
    role: "update_set_column",
    validator: None,
    highlight_override: None,
    writes_table: false,
    writes_column: true,
    writes_user_listed_column: false,
};

/// Column-name slot in `where col = …` — same writes-column
/// semantics as SET_COLUMN, distinct role for the AST builder.
const FILTER_COLUMN: Node = Node::Ident {
    source: IdentSource::Columns,
    role: "filter_column",
    validator: None,
    highlight_override: None,
    writes_table: false,
    writes_column: true,
    writes_user_listed_column: false,
};

/// Value slot resolved at walk time from
/// `WalkContext::current_column`. Falls back to the schemaless
/// value-literal choice when no current_column is bound.
const PER_COLUMN_VALUE: Node = Node::DynamicSubgrammar(current_column_value);

const UPDATE_ASSIGNMENT_NODES: &[Node] = &[
    SET_COLUMN,
    Node::Punct('='),
    PER_COLUMN_VALUE,
];
const UPDATE_ASSIGNMENT: Node = Node::Seq(UPDATE_ASSIGNMENT_NODES);
const UPDATE_ASSIGNMENTS: Node = Node::Repeated {
    inner: &UPDATE_ASSIGNMENT,
    separator: Some(&Node::Punct(',')),
    min: 1,
};

const WHERE_CLAUSE_NODES: &[Node] = &[
    Node::Word(Word::keyword("where")),
    FILTER_COLUMN,
    Node::Punct('='),
    PER_COLUMN_VALUE,
];
const WHERE_CLAUSE: Node = Node::Seq(WHERE_CLAUSE_NODES);

const FILTER_CHOICES: &[Node] = &[WHERE_CLAUSE, Node::Flag("all-rows")];
const FILTER_CLAUSE: Node = Node::Choice(FILTER_CHOICES);

const UPDATE_NODES: &[Node] = &[
    TABLE_NAME_WRITES,
    Node::Word(Word::keyword("set")),
    UPDATE_ASSIGNMENTS,
    FILTER_CLAUSE,
];
const UPDATE_SHAPE: Node = Node::Seq(UPDATE_NODES);

// =================================================================
// delete — `delete from <T> (where … | --all-rows)`
// =================================================================

const DELETE_NODES: &[Node] = &[
    Node::Word(Word::keyword("from")),
    TABLE_NAME_WRITES,
    FILTER_CLAUSE,
];
const DELETE_SHAPE: Node = Node::Seq(DELETE_NODES);

// =================================================================
// AST builders
// =================================================================

fn ident_text<'a>(path: &'a MatchedPath, role: &str) -> Option<&'a str> {
    path.items.iter().find_map(|i| match &i.kind {
        MatchedKind::Ident { role: r } if *r == role => Some(i.text.as_str()),
        _ => None,
    })
}

fn require_ident(path: &MatchedPath, role: &'static str) -> Result<String, ValidationError> {
    ident_text(path, role)
        .map(str::to_string)
        .ok_or_else(|| ValidationError {
            message_key: "parse.error_wrapper",
            args: vec![("detail", format!("missing {role}"))],
        })
}

/// Convert a `MatchedItem` whose kind is one of the `value_literal`
/// variants (Word("null"|"true"|"false"), NumberLit, StringLit) to
/// a `Value`. Returns None for non-value items.
fn item_to_value(item: &MatchedItem) -> Option<Value> {
    match &item.kind {
        MatchedKind::Word("null") => Some(Value::Null),
        MatchedKind::Word("true") => Some(Value::Bool(true)),
        MatchedKind::Word("false") => Some(Value::Bool(false)),
        MatchedKind::NumberLit => Some(Value::Number(item.text.clone())),
        MatchedKind::StringLit => Some(Value::Text(item.text.clone())),
        _ => None,
    }
}

fn build_show(path: &MatchedPath) -> Result<Command, ValidationError> {
    let sub = path
        .items
        .iter()
        .filter_map(|i| match &i.kind {
            MatchedKind::Word(w) => Some(*w),
            _ => None,
        })
        .nth(1);
    let name = require_ident(path, "table_name")?;
    match sub {
        Some("data") => Ok(Command::ShowData { name }),
        Some("table") => Ok(Command::ShowTable { name }),
        _ => Err(ValidationError {
            message_key: "parse.error_wrapper",
            args: vec![("detail", "unknown show subcommand".to_string())],
        }),
    }
}

fn build_insert(path: &MatchedPath) -> Result<Command, ValidationError> {
    let table = require_ident(path, "table_name")?;

    // Locate the second `values` keyword (the first is the
    // command word `insert`'s sibling — but `insert` isn't a
    // matched Word here since it's the entry word and the
    // entry-word push uses the literal "insert"; only later
    // `values` matches as Word("values")).
    //
    // Strategy: walk the path. After the table name:
    //   - If we see Word("values") next (Form B), the next
    //     parenthesized values are the value list.
    //   - If we see Punct('('), the first paren's content was
    //     either column names (Form A) or values (Form C).
    //     If a Word("values") follows the closing paren, it's
    //     Form A.
    //
    // Easier discriminator: collect all matched keyword words;
    // count occurrences of "values".
    let saw_values = path
        .items
        .iter()
        .any(|i| matches!(i.kind, MatchedKind::Word("values")));

    // Find the index of the table_name match — the first paren
    // afterwards starts the parsed list.
    let table_idx = path
        .items
        .iter()
        .position(|i| matches!(&i.kind, MatchedKind::Ident { role: "table_name" }))
        .ok_or_else(|| ValidationError {
            message_key: "parse.error_wrapper",
            args: vec![("detail", "missing table".to_string())],
        })?;

    // Form B (values keyword right after table): no column list,
    // values come from the single paren-bounded list.
    let first_token_after_table = path.items.get(table_idx + 1);
    let form_b = matches!(
        first_token_after_table.map(|i| &i.kind),
        Some(MatchedKind::Word("values"))
    );

    if form_b {
        // Form B: the only value run is between the only `(` … `)`.
        let values = collect_values_in_parens(path, table_idx + 1)?;
        return Ok(Command::Insert {
            table,
            columns: None,
            values,
        });
    }

    // Form A or C: the first paren after the table is a Choice
    // of either column-idents or value-literals.
    let first_paren_idx = path
        .items
        .iter()
        .enumerate()
        .skip(table_idx + 1)
        .find(|(_, i)| matches!(i.kind, MatchedKind::Punct('(')))
        .map(|(idx, _)| idx)
        .ok_or_else(|| ValidationError {
            message_key: "parse.error_wrapper",
            args: vec![("detail", "missing `(`".to_string())],
        })?;

    if saw_values {
        // Form A: first paren = column names; second paren = values.
        // The Repeated inside the first paren tagged matched idents
        // with role "insert_first_item".
        let columns: Vec<String> = path
            .items
            .iter()
            .filter_map(|i| match &i.kind {
                MatchedKind::Ident {
                    role: "insert_first_item",
                } => Some(i.text.clone()),
                _ => None,
            })
            .collect();
        if columns.is_empty() {
            return Err(ValidationError {
                message_key: "parse.error_wrapper",
                args: vec![("detail", "expected column names in `insert into T (…)`".to_string())],
            });
        }
        // Find the `values` keyword and the next `(` — the values
        // run starts after that `(`.
        let values_idx = path
            .items
            .iter()
            .enumerate()
            .skip(first_paren_idx)
            .find(|(_, i)| matches!(i.kind, MatchedKind::Word("values")))
            .map(|(i, _)| i)
            .ok_or_else(|| ValidationError {
                message_key: "parse.error_wrapper",
                args: vec![("detail", "missing `values` keyword".to_string())],
            })?;
        let values = collect_values_in_parens(path, values_idx + 1)?;
        Ok(Command::Insert {
            table,
            columns: Some(columns),
            values,
        })
    } else {
        // Form C: the first paren contained the value list. The
        // Repeated tagged the matched values via their natural
        // MatchedKind (Word/NumberLit/StringLit); collect them.
        //
        // Form-A-without-`values` recovery: the shared
        // INSERT_PAREN_ITEM choice accepts both VALUE_LITERAL
        // and Ident{Columns} so that Form A can resolve
        // column-name items inside its `( cols )` list. When the
        // user types `insert into T (col)` (column-shaped item,
        // no `values` keyword), the grammar walks to a complete
        // match but the user almost certainly meant Form A and
        // forgot the `values (...)` suffix. Reject here with a
        // ValidationError — the walker classifies validation
        // errors as `at_eof: true`, so the input renderer
        // surfaces this as IncompleteAtEof (mid-typing) rather
        // than dispatching a logically-broken Form C insert with
        // an empty value list.
        let user_listed_columns: Vec<String> = path
            .items
            .iter()
            .filter_map(|i| match &i.kind {
                MatchedKind::Ident {
                    role: "insert_first_item",
                } => Some(i.text.clone()),
                _ => None,
            })
            .collect();
        if !user_listed_columns.is_empty() {
            return Err(ValidationError {
                message_key: "parse.custom.insert_form_a_missing_values",
                args: vec![("columns", user_listed_columns.join(", "))],
            });
        }
        let values = collect_values_in_parens(path, first_paren_idx)?;
        Ok(Command::Insert {
            table,
            columns: None,
            values,
        })
    }
}

/// Collect Value items inside the next `(…)` block at or after
/// `start_idx`. Stops at the matching `)`.
fn collect_values_in_parens(
    path: &MatchedPath,
    start_idx: usize,
) -> Result<Vec<Value>, ValidationError> {
    let mut out = Vec::new();
    let mut inside = false;
    for item in path.items.iter().skip(start_idx) {
        match &item.kind {
            MatchedKind::Punct('(') => inside = true,
            MatchedKind::Punct(')') if inside => return Ok(out),
            _ if inside => {
                if let Some(v) = item_to_value(item) {
                    out.push(v);
                }
            }
            _ => {}
        }
    }
    if out.is_empty() && !inside {
        return Err(ValidationError {
            message_key: "parse.error_wrapper",
            args: vec![("detail", "missing `(`".to_string())],
        });
    }
    Ok(out)
}

fn build_update(path: &MatchedPath) -> Result<Command, ValidationError> {
    let table = require_ident(path, "table_name")?;
    let assignments = collect_assignments(path)?;
    let filter = collect_filter(path)?;
    Ok(Command::Update {
        table,
        assignments,
        filter,
    })
}

fn collect_assignments(
    path: &MatchedPath,
) -> Result<Vec<(String, Value)>, ValidationError> {
    let mut out = Vec::new();
    let mut iter = path.items.iter();
    while let Some(item) = iter.next() {
        if matches!(
            item.kind,
            MatchedKind::Ident {
                role: "update_set_column"
            }
        ) {
            let column = item.text.clone();
            // Skip the `=` punct.
            for next in iter.by_ref() {
                if matches!(next.kind, MatchedKind::Punct('=')) {
                    break;
                }
            }
            // Next item is the value.
            let value_item = iter.next().ok_or_else(|| ValidationError {
                message_key: "parse.error_wrapper",
                args: vec![("detail", "missing assignment value".to_string())],
            })?;
            let value = item_to_value(value_item).ok_or_else(|| ValidationError {
                message_key: "parse.error_wrapper",
                args: vec![("detail", "expected value literal".to_string())],
            })?;
            out.push((column, value));
        }
    }
    Ok(out)
}

fn collect_filter(path: &MatchedPath) -> Result<RowFilter, ValidationError> {
    if path
        .items
        .iter()
        .any(|i| matches!(i.kind, MatchedKind::Flag("all-rows")))
    {
        return Ok(RowFilter::AllRows);
    }
    // Walk for filter_column ident, then `=`, then value.
    let mut iter = path.items.iter();
    while let Some(item) = iter.next() {
        if matches!(
            item.kind,
            MatchedKind::Ident {
                role: "filter_column"
            }
        ) {
            let column = item.text.clone();
            // Skip until `=`.
            for next in iter.by_ref() {
                if matches!(next.kind, MatchedKind::Punct('=')) {
                    break;
                }
            }
            let value_item = iter.next().ok_or_else(|| ValidationError {
                message_key: "parse.error_wrapper",
                args: vec![("detail", "missing where value".to_string())],
            })?;
            let value = item_to_value(value_item).ok_or_else(|| ValidationError {
                message_key: "parse.error_wrapper",
                args: vec![("detail", "expected value literal".to_string())],
            })?;
            return Ok(RowFilter::Where { column, value });
        }
    }
    Err(ValidationError {
        message_key: "parse.error_wrapper",
        args: vec![("detail", "missing where or --all-rows".to_string())],
    })
}

fn build_delete(path: &MatchedPath) -> Result<Command, ValidationError> {
    let table = require_ident(path, "table_name")?;
    let filter = collect_filter(path)?;
    Ok(Command::Delete { table, filter })
}

// =================================================================
// replay — `replay <bare-path>` | `replay '<path>'`
// =================================================================
//
// Phase E (ADR-0024 §migration). The chumsky-side
// `try_parse_replay_with_bare_path` source-slice helper is
// retired here: walker BarePath consumes the unquoted form
// (terminating at whitespace per the path-bearing UX change),
// and StringLit consumes the quoted form. Paths with spaces
// must use the quoted form — same UX that `import` / `export`
// adopted in Phase A.

const REPLAY_PATH_CHOICES: &[Node] = &[Node::StringLit, Node::BarePath];
const REPLAY_PATH: Node = Node::Choice(REPLAY_PATH_CHOICES);

fn build_replay(path: &MatchedPath) -> Result<Command, ValidationError> {
    let payload = path
        .items
        .iter()
        .find_map(|i| match &i.kind {
            MatchedKind::StringLit | MatchedKind::BarePath => Some(i.text.clone()),
            _ => None,
        })
        .ok_or_else(|| ValidationError {
            message_key: "parse.error_wrapper",
            args: vec![("detail", "missing path".to_string())],
        })?;
    Ok(Command::Replay { path: payload })
}

// =================================================================
// CommandNodes
// =================================================================

pub static SHOW: CommandNode = CommandNode {
    entry: Word::keyword("show"),
    shape: SHOW_SHAPE,
    ast_builder: build_show,
    help_id: Some("data.show"),
    usage_ids: &["parse.usage.show_data", "parse.usage.show_table"],
    hint_mode: None,
};

pub static INSERT: CommandNode = CommandNode {
    entry: Word::keyword("insert"),
    shape: INSERT_SHAPE,
    ast_builder: build_insert,
    help_id: Some("data.insert"),
    usage_ids: &["parse.usage.insert"],
    hint_mode: None,
};

pub static UPDATE: CommandNode = CommandNode {
    entry: Word::keyword("update"),
    shape: UPDATE_SHAPE,
    ast_builder: build_update,
    help_id: Some("data.update"),
    usage_ids: &["parse.usage.update"],
    hint_mode: None,
};

pub static DELETE: CommandNode = CommandNode {
    entry: Word::keyword("delete"),
    shape: DELETE_SHAPE,
    ast_builder: build_delete,
    help_id: Some("data.delete"),
    usage_ids: &["parse.usage.delete"],
    hint_mode: None,
};

pub static REPLAY: CommandNode = CommandNode {
    entry: Word::keyword("replay"),
    shape: REPLAY_PATH,
    ast_builder: build_replay,
    help_id: Some("data.replay"),
    usage_ids: &["parse.usage.replay"],
    hint_mode: None,
};