claude@clouddev1
4ff054ca75
Sub-phase 2b checkpoints 4 and 5 combined — adds the placeholder CTE binding push (§10.3 stage 1) and the projection alias accumulator (§10.4). Node::Ident gains two more flags, mechanically applied to every existing site: - `writes_cte_name: bool` — push a placeholder `CteBinding` (name only, empty columns) onto the top `ScopeFrame`'s `cte_bindings`. Set on `CTE_NAME_IDENT` in sql_select.rs. Fires BEFORE the body's `ScopedSubgrammar` enters (the CTE-def Seq's ident slot precedes the body's `(`), so the body can self-reference the CTE name as a valid table source (WITH RECURSIVE). - `writes_projection_alias: bool` — append the matched name to the top frame's `projection_aliases`. Set on `PROJECTION_BARE_ALIAS_IDENT` so both the AS-form (`a AS alpha`) and bare-form (`a alpha`) paths capture cleanly. The ident is shared by both paths through `PROJECTION_AS_ALIAS` and the lookahead factory, so capturing on the ident itself covers both forms with no duplication. The §10.3 stage-2 harvest (deriving CTE output columns from the body's projection per the six derivation rules in the ADR's table) is structurally deferred — the placeholder's `columns` stays empty until the harvest is wired. This is intentional scope honesty: the placeholder-name presence is sufficient for the schema-existence diagnostic (2d) to recognize CTE names as valid table sources, and the qualified-prefix completion (2e) will populate the columns when the harvest hook is added there. Tests below assert the placeholder-name behavior; the column-derivation tests from plan §2b's exit gate will be satisfied incrementally as later sub-phases need them. Tests (8 new, all green): - Single CTE → one placeholder binding with the matched name. - Multiple CTEs → placeholders in declaration order. - Recursive CTE → name visible inside body (the body's `from r` reference parses; verified by the walk completing). - Projection aliases via AS form → captured into the top frame's `projection_aliases`. - Projection aliases via bare form → captured. - Mixed alias forms → captured in projection order, with unaliased projection items absent from the alias list. - No aliases → empty `projection_aliases`. - CTE body aliases do not leak to outer scope (the body's frame pops on `ScopedSubgrammar` exit, taking its projection_aliases with it). All 1358 previous tests still pass. Test totals: 1366 passing, 0 failed, 1 ignored. Clippy clean. This closes out the scope-accumulator side of sub-phase 2b. The remaining 2b-style work — full CTE column-derivation harvest per §10.3's six rules — folds into 2d (where the arity-check pass needs declared-vs-derived column counts) and 2e (where qualified-prefix completion needs CTE columns).
751 lines
25 KiB
Rust
751 lines
25 KiB
Rust
//! The SQL expression grammar fragment (ADR-0031).
|
|
//!
|
|
//! This is the advanced-mode counterpart of `grammar::expr` (the
|
|
//! DSL `WHERE` grammar, ADR-0026). It fills every expression slot
|
|
//! in advanced-mode SQL — `WHERE`, `HAVING`, `SELECT` projections,
|
|
//! `CHECK`, `DEFAULT` — and is the *superset* of ADR-0026's
|
|
//! grammar: it adds arithmetic, string concatenation, function
|
|
//! calls, and `CASE` on top of the comparison / `LIKE` / `IN` /
|
|
//! `BETWEEN` / `IS NULL` predicate set.
|
|
//!
|
|
//! # One unified ladder
|
|
//!
|
|
//! ADR-0026's DSL grammar stratifies a *boolean* layer above a
|
|
//! *predicate* layer because the DSL forbids a boolean
|
|
//! sub-expression as a comparison operand. Standard SQL draws no
|
|
//! such line — a boolean is a value — so this grammar is a single
|
|
//! precedence ladder, loosest tier first:
|
|
//!
|
|
//! ```text
|
|
//! or_expr := and_expr ( OR and_expr )*
|
|
//! and_expr := not_expr ( AND not_expr )*
|
|
//! not_expr := NOT not_expr | predicate
|
|
//! predicate := additive predicate_tail?
|
|
//! predicate_tail := cmp_op additive
|
|
//! | [ NOT ] LIKE additive
|
|
//! | [ NOT ] BETWEEN additive AND additive
|
|
//! | [ NOT ] IN ( additive ( , additive )* )
|
|
//! | IS [ NOT ] NULL
|
|
//! additive := multiplicative ( ( + | - | || ) multiplicative )*
|
|
//! multiplicative := unary ( ( * | / | % ) unary )*
|
|
//! unary := ( - | + ) unary | primary
|
|
//! primary := literal | ( or_expr ) | case_expr | name_or_call
|
|
//! name_or_call := identifier [ '(' call_args? ')' ]
|
|
//! call_args := '*' | DISTINCT arg_list | arg_list
|
|
//! case_expr := CASE [ or_expr ] ( WHEN or_expr THEN or_expr )+
|
|
//! [ ELSE or_expr ] END
|
|
//! ```
|
|
//!
|
|
//! Stratification removes left recursion (every recursion is
|
|
//! guarded by a token) and encodes precedence in the layering.
|
|
//! Each tier transition, and each genuine recursion, goes through
|
|
//! [`Node::Subgrammar`] — a `Seq` / `Choice` embeds its children
|
|
//! by value and cannot close a cycle. The walker counts active
|
|
//! `Subgrammar` frames and refuses past `MAX_SUBGRAMMAR_DEPTH`
|
|
//! (ADR-0026 §1) — reused here unchanged.
|
|
//!
|
|
//! # No AST (ADR-0031 §2)
|
|
//!
|
|
//! Unlike `grammar::expr`, this fragment carries **no AST
|
|
//! builder**. Its consumers run or store SQL as text (ADR-0030
|
|
//! §4/§6), so there is no `Expr` to build. The fragment's output
|
|
//! is the three other walker products — accept/reject, the flat
|
|
//! `MatchedPath` of terminals (which drives highlighting,
|
|
//! completion, the expected-set, and hints), and, via those
|
|
//! terminals' `span`s, the source range a consumer slices when it
|
|
//! needs the expression as text. The grammar tier owns
|
|
//! validation, highlight, completion, and the no-left-recursion
|
|
//! guarantee; it simply has no tree to hand back.
|
|
|
|
use crate::dsl::grammar::{IdentSource, Node, Word, sql_select};
|
|
|
|
// =================================================================
|
|
// Shared leaf nodes
|
|
// =================================================================
|
|
|
|
/// `,` — the separator inside `IN ( … )` and a function-call
|
|
/// argument list.
|
|
static COMMA: Node = Node::Punct(',');
|
|
|
|
/// A column reference inside an expression. `IdentSource::Columns`
|
|
/// drives Tab completion against the statement's table(s) from the
|
|
/// same `SchemaCache` the DSL uses (ADR-0031 §5). The same slot
|
|
/// also begins a `name_or_call` function call — the grammar does
|
|
/// not decide which (ADR-0031 §1, §2), so the slot optimises for
|
|
/// the common case (a column) and a function name simply is not a
|
|
/// completion candidate. `const` so it can be embedded by value.
|
|
const EXPR_IDENT: Node = Node::Ident {
|
|
source: IdentSource::Columns,
|
|
role: "sql_expr_ident",
|
|
validator: None,
|
|
highlight_override: None,
|
|
writes_table: false,
|
|
writes_column: false,
|
|
writes_user_listed_column: false,
|
|
writes_table_alias: false,
|
|
writes_cte_name: false,
|
|
writes_projection_alias: false,
|
|
};
|
|
|
|
// =================================================================
|
|
// or_expr := and_expr ( OR and_expr )* — the fragment entry point
|
|
// =================================================================
|
|
|
|
static OR_TAIL_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("or")),
|
|
Node::Subgrammar(&AND_EXPR),
|
|
];
|
|
static OR_TAIL: Node = Node::Seq(OR_TAIL_NODES);
|
|
static SQL_OR_EXPR_NODES: &[Node] = &[
|
|
Node::Subgrammar(&AND_EXPR),
|
|
Node::Repeated {
|
|
inner: &OR_TAIL,
|
|
separator: None,
|
|
min: 0,
|
|
},
|
|
];
|
|
/// `or_expr` — the loosest tier and the grammar's entry point.
|
|
pub static SQL_OR_EXPR: Node = Node::Seq(SQL_OR_EXPR_NODES);
|
|
|
|
/// The SQL expression fragment, ready to drop into a SQL command
|
|
/// `Seq` as one node — the advanced-mode counterpart of
|
|
/// [`crate::dsl::grammar::expr::EXPRESSION`]. Walking it consumes a
|
|
/// whole `or_expr`.
|
|
pub static SQL_EXPRESSION: Node = Node::Subgrammar(&SQL_OR_EXPR);
|
|
|
|
// =================================================================
|
|
// and_expr := not_expr ( AND not_expr )*
|
|
// =================================================================
|
|
|
|
static AND_TAIL_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("and")),
|
|
Node::Subgrammar(&NOT_EXPR),
|
|
];
|
|
static AND_TAIL: Node = Node::Seq(AND_TAIL_NODES);
|
|
static AND_EXPR_NODES: &[Node] = &[
|
|
Node::Subgrammar(&NOT_EXPR),
|
|
Node::Repeated {
|
|
inner: &AND_TAIL,
|
|
separator: None,
|
|
min: 0,
|
|
},
|
|
];
|
|
static AND_EXPR: Node = Node::Seq(AND_EXPR_NODES);
|
|
|
|
// =================================================================
|
|
// not_expr := NOT not_expr | predicate
|
|
// =================================================================
|
|
|
|
static NOT_FORM_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("not")),
|
|
Node::Subgrammar(&NOT_EXPR),
|
|
];
|
|
static NOT_EXPR_CHOICES: &[Node] = &[
|
|
Node::Seq(NOT_FORM_NODES),
|
|
Node::Subgrammar(&PREDICATE),
|
|
];
|
|
static NOT_EXPR: Node = Node::Choice(NOT_EXPR_CHOICES);
|
|
|
|
// =================================================================
|
|
// predicate := additive predicate_tail?
|
|
// =================================================================
|
|
//
|
|
// `predicate_tail` is optional: a bare `additive` with no
|
|
// comparison operator is itself a valid expression — that is what
|
|
// a `SELECT` projection item (`a + b`) or a literal `WHERE 1`
|
|
// needs. ADR-0026's DSL grammar made the tail mandatory because it
|
|
// forbade a bare column as a boolean; SQL does not.
|
|
|
|
static PREDICATE_NODES: &[Node] = &[
|
|
Node::Subgrammar(&ADDITIVE),
|
|
Node::Optional(&PREDICATE_TAIL),
|
|
];
|
|
static PREDICATE: Node = Node::Seq(PREDICATE_NODES);
|
|
|
|
// ---- cmp_op := <= | <> | >= | != | < | > | = --------------------
|
|
//
|
|
// Two-character operators come before their one-character
|
|
// prefixes: `walk_literal` matches `<` against the `<` of `<=`,
|
|
// so `<=` / `<>` must be tried first (ADR-0026's note, inherited).
|
|
|
|
static CMP_OP_CHOICES: &[Node] = &[
|
|
Node::Literal("<="),
|
|
Node::Literal("<>"),
|
|
Node::Literal(">="),
|
|
Node::Literal("!="),
|
|
Node::Literal("<"),
|
|
Node::Literal(">"),
|
|
Node::Literal("="),
|
|
];
|
|
|
|
// ---- predicate_tail branches ------------------------------------
|
|
|
|
/// `cmp_op additive`.
|
|
static COMPARE_FORM_NODES: &[Node] = &[
|
|
Node::Choice(CMP_OP_CHOICES),
|
|
Node::Subgrammar(&ADDITIVE),
|
|
];
|
|
|
|
/// `IS [NOT] NULL`.
|
|
static IS_NULL_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("is")),
|
|
Node::Optional(&Node::Word(Word::keyword("not"))),
|
|
Node::Word(Word::keyword("null")),
|
|
];
|
|
|
|
/// `LIKE additive`.
|
|
static LIKE_FORM_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("like")),
|
|
Node::Subgrammar(&ADDITIVE),
|
|
];
|
|
|
|
/// `BETWEEN additive AND additive`. The inner `and` is consumed
|
|
/// here, so a stray `and` at the `and_expr` tier is always a
|
|
/// connective.
|
|
static BETWEEN_FORM_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("between")),
|
|
Node::Subgrammar(&ADDITIVE),
|
|
Node::Word(Word::keyword("and")),
|
|
Node::Subgrammar(&ADDITIVE),
|
|
];
|
|
|
|
/// `IN ( additive [, additive]* | compound_select )` —
|
|
/// ADR-0032 §6.2. The `IN (` prefix is factored; after the
|
|
/// opening paren a `Choice` dispatches between the
|
|
/// compound-select subquery and the comma-separated value
|
|
/// list. The first inside token disambiguates the same way the
|
|
/// scalar-subquery `primary` does. The subquery recurses
|
|
/// through `ScopedSubgrammar`.
|
|
static IN_ITEM: Node = Node::Subgrammar(&ADDITIVE);
|
|
static IN_INSIDE_CHOICES: &[Node] = &[
|
|
Node::ScopedSubgrammar(&sql_select::SQL_SELECT_COMPOUND),
|
|
Node::Repeated {
|
|
inner: &IN_ITEM,
|
|
separator: Some(&COMMA),
|
|
min: 1,
|
|
},
|
|
];
|
|
static IN_FORM_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("in")),
|
|
Node::Punct('('),
|
|
Node::Choice(IN_INSIDE_CHOICES),
|
|
Node::Punct(')'),
|
|
];
|
|
|
|
/// The negatable predicate bodies — each starts with a distinct
|
|
/// keyword, so this `Choice` discriminates cleanly.
|
|
static NEGATABLE_CHOICES: &[Node] = &[
|
|
Node::Seq(LIKE_FORM_NODES),
|
|
Node::Seq(BETWEEN_FORM_NODES),
|
|
Node::Seq(IN_FORM_NODES),
|
|
];
|
|
|
|
/// `NOT (LIKE … | BETWEEN … | IN …)` — the infix `NOT` is factored
|
|
/// in front of the negatable choice (ADR-0026's factoring).
|
|
static NOT_NEGATABLE_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("not")),
|
|
Node::Choice(NEGATABLE_CHOICES),
|
|
];
|
|
|
|
/// `predicate_tail`. Branch discrimination relies on each branch's
|
|
/// *first* child reporting a clean `NoMatch` on a non-match — no
|
|
/// branch starts with an `Optional` (ADR-0026's hazard). The infix
|
|
/// `NOT` is its own explicit `NOT negatable` branch.
|
|
static PREDICATE_TAIL_CHOICES: &[Node] = &[
|
|
Node::Seq(COMPARE_FORM_NODES),
|
|
Node::Seq(IS_NULL_NODES),
|
|
Node::Seq(NOT_NEGATABLE_NODES),
|
|
Node::Seq(LIKE_FORM_NODES),
|
|
Node::Seq(BETWEEN_FORM_NODES),
|
|
Node::Seq(IN_FORM_NODES),
|
|
];
|
|
static PREDICATE_TAIL: Node = Node::Choice(PREDICATE_TAIL_CHOICES);
|
|
|
|
// =================================================================
|
|
// additive := multiplicative ( ( + | - | || ) multiplicative )*
|
|
// =================================================================
|
|
|
|
static ADD_OP_CHOICES: &[Node] = &[
|
|
Node::Punct('+'),
|
|
Node::Punct('-'),
|
|
Node::Literal("||"),
|
|
];
|
|
static ADD_TAIL_NODES: &[Node] = &[
|
|
Node::Choice(ADD_OP_CHOICES),
|
|
Node::Subgrammar(&MULTIPLICATIVE),
|
|
];
|
|
static ADD_TAIL: Node = Node::Seq(ADD_TAIL_NODES);
|
|
static ADDITIVE_NODES: &[Node] = &[
|
|
Node::Subgrammar(&MULTIPLICATIVE),
|
|
Node::Repeated {
|
|
inner: &ADD_TAIL,
|
|
separator: None,
|
|
min: 0,
|
|
},
|
|
];
|
|
static ADDITIVE: Node = Node::Seq(ADDITIVE_NODES);
|
|
|
|
// =================================================================
|
|
// multiplicative := unary ( ( * | / | % ) unary )*
|
|
// =================================================================
|
|
|
|
static MUL_OP_CHOICES: &[Node] = &[
|
|
Node::Punct('*'),
|
|
Node::Punct('/'),
|
|
Node::Punct('%'),
|
|
];
|
|
static MUL_TAIL_NODES: &[Node] = &[
|
|
Node::Choice(MUL_OP_CHOICES),
|
|
Node::Subgrammar(&UNARY),
|
|
];
|
|
static MUL_TAIL: Node = Node::Seq(MUL_TAIL_NODES);
|
|
static MULTIPLICATIVE_NODES: &[Node] = &[
|
|
Node::Subgrammar(&UNARY),
|
|
Node::Repeated {
|
|
inner: &MUL_TAIL,
|
|
separator: None,
|
|
min: 0,
|
|
},
|
|
];
|
|
static MULTIPLICATIVE: Node = Node::Seq(MULTIPLICATIVE_NODES);
|
|
|
|
// =================================================================
|
|
// unary := ( - | + ) unary | primary
|
|
// =================================================================
|
|
|
|
static SIGN_CHOICES: &[Node] = &[Node::Punct('-'), Node::Punct('+')];
|
|
static UNARY_SIGN_NODES: &[Node] = &[
|
|
Node::Choice(SIGN_CHOICES),
|
|
Node::Subgrammar(&UNARY),
|
|
];
|
|
static UNARY_CHOICES: &[Node] = &[
|
|
Node::Seq(UNARY_SIGN_NODES),
|
|
Node::Subgrammar(&PRIMARY),
|
|
];
|
|
static UNARY: Node = Node::Choice(UNARY_CHOICES);
|
|
|
|
// =================================================================
|
|
// primary := literal | ( or_expr ) | case_expr | name_or_call
|
|
// =================================================================
|
|
|
|
/// `( or_expr )` and the scalar subquery `( compound_select )`
|
|
/// share a leading `(`. Per ADR-0032 §6.1, the `(` is matched
|
|
/// once and the inside is a `Choice` between
|
|
/// `compound_select` (the scalar subquery) and `or_expr` (the
|
|
/// parenthesised expression). The first inside token
|
|
/// disambiguates: `SELECT` or `WITH` → subquery; anything else →
|
|
/// expression. Subquery recursion goes through
|
|
/// `ScopedSubgrammar` to push a new lexical scope (§10.2).
|
|
static PAREN_INSIDE_CHOICES: &[Node] = &[
|
|
Node::ScopedSubgrammar(&sql_select::SQL_SELECT_COMPOUND),
|
|
Node::Subgrammar(&SQL_OR_EXPR),
|
|
];
|
|
static PAREN_GROUP_NODES: &[Node] = &[
|
|
Node::Punct('('),
|
|
Node::Choice(PAREN_INSIDE_CHOICES),
|
|
Node::Punct(')'),
|
|
];
|
|
|
|
// ADR-0032 §6.3 — `EXISTS ( compound_select )` as a primary.
|
|
// `[ NOT ] EXISTS` falls out via the existing `not_expr := NOT
|
|
// not_expr` tier above `primary`, so only the bare `EXISTS`
|
|
// form lives here.
|
|
static EXISTS_PRIMARY_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("exists")),
|
|
Node::Punct('('),
|
|
Node::ScopedSubgrammar(&sql_select::SQL_SELECT_COMPOUND),
|
|
Node::Punct(')'),
|
|
];
|
|
|
|
// ---- case_expr --------------------------------------------------
|
|
//
|
|
// `CASE [operand] (WHEN cond THEN result)+ [ELSE result] END`.
|
|
// Searched-CASE (no operand) and simple-CASE (with operand) are two
|
|
// `Choice` branches *after* the shared `CASE` keyword is factored
|
|
// out: the searched branch opens with `WHEN`, the simple branch
|
|
// with an operand expression. Branch 1's leading `Repeated(min 1)`
|
|
// reports a clean `NoMatch` when the next token is not `WHEN`, so
|
|
// the `Choice` falls through to the simple branch cleanly.
|
|
|
|
static WHEN_CLAUSE_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("when")),
|
|
Node::Subgrammar(&SQL_OR_EXPR),
|
|
Node::Word(Word::keyword("then")),
|
|
Node::Subgrammar(&SQL_OR_EXPR),
|
|
];
|
|
static WHEN_CLAUSE: Node = Node::Seq(WHEN_CLAUSE_NODES);
|
|
|
|
static ELSE_CLAUSE_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("else")),
|
|
Node::Subgrammar(&SQL_OR_EXPR),
|
|
];
|
|
static ELSE_CLAUSE: Node = Node::Seq(ELSE_CLAUSE_NODES);
|
|
|
|
/// Searched-CASE body: `(WHEN … THEN …)+ [ELSE …] END`.
|
|
static SEARCHED_CASE_NODES: &[Node] = &[
|
|
Node::Repeated {
|
|
inner: &WHEN_CLAUSE,
|
|
separator: None,
|
|
min: 1,
|
|
},
|
|
Node::Optional(&ELSE_CLAUSE),
|
|
Node::Word(Word::keyword("end")),
|
|
];
|
|
/// Simple-CASE body: `operand (WHEN … THEN …)+ [ELSE …] END`.
|
|
static SIMPLE_CASE_NODES: &[Node] = &[
|
|
Node::Subgrammar(&SQL_OR_EXPR),
|
|
Node::Repeated {
|
|
inner: &WHEN_CLAUSE,
|
|
separator: None,
|
|
min: 1,
|
|
},
|
|
Node::Optional(&ELSE_CLAUSE),
|
|
Node::Word(Word::keyword("end")),
|
|
];
|
|
static CASE_BODY_CHOICES: &[Node] = &[
|
|
Node::Seq(SEARCHED_CASE_NODES),
|
|
Node::Seq(SIMPLE_CASE_NODES),
|
|
];
|
|
static CASE_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("case")),
|
|
Node::Choice(CASE_BODY_CHOICES),
|
|
];
|
|
|
|
// ---- name_or_call -----------------------------------------------
|
|
//
|
|
// `identifier [ '(' call_args? ')' ]`. The identifier is matched
|
|
// once; the `( call_args )` group is an `Optional` tail — present
|
|
// is a function call, absent is a column reference. Factoring this
|
|
// (rather than two `Choice` branches sharing the identifier
|
|
// prefix) avoids the function-call branch committing on the
|
|
// identifier and discarding the column-ref branch.
|
|
|
|
/// One function-call argument — a whole expression.
|
|
static CALL_ARG: Node = Node::Subgrammar(&SQL_OR_EXPR);
|
|
|
|
/// `call_args := '*' | DISTINCT arg_list | arg_list`. `count(*)`
|
|
/// is the one place `*` is an argument; `count(distinct col)` the
|
|
/// one place `DISTINCT` leads the list. The grammar admits the
|
|
/// call shape structurally — it does not know which names are
|
|
/// aggregates (ADR-0031 §1).
|
|
static DISTINCT_ARGS_NODES: &[Node] = &[
|
|
Node::Word(Word::keyword("distinct")),
|
|
Node::Repeated {
|
|
inner: &CALL_ARG,
|
|
separator: Some(&COMMA),
|
|
min: 1,
|
|
},
|
|
];
|
|
static CALL_ARGS_CHOICES: &[Node] = &[
|
|
Node::Punct('*'),
|
|
Node::Seq(DISTINCT_ARGS_NODES),
|
|
Node::Repeated {
|
|
inner: &CALL_ARG,
|
|
separator: Some(&COMMA),
|
|
min: 1,
|
|
},
|
|
];
|
|
static CALL_ARGS: Node = Node::Choice(CALL_ARGS_CHOICES);
|
|
|
|
static CALL_TAIL_NODES: &[Node] = &[
|
|
Node::Punct('('),
|
|
Node::Optional(&CALL_ARGS),
|
|
Node::Punct(')'),
|
|
];
|
|
|
|
// ADR-0032 §5 — qualified column reference. `name_or_call` gains
|
|
// a `.identifier` suffix as a Choice sibling of the function-call
|
|
// `( args )` tail. The leading identifier is matched once (no
|
|
// Choice branch begins with an identifier per ADR-0031 §1's
|
|
// factoring); the optional tail dispatches between the two
|
|
// suffixes by their first character (`.` vs `(`).
|
|
const QUALIFIED_REF_IDENT: Node = Node::Ident {
|
|
source: IdentSource::Columns,
|
|
role: "sql_expr_qualified_ref",
|
|
validator: None,
|
|
highlight_override: None,
|
|
writes_table: false,
|
|
writes_column: false,
|
|
writes_user_listed_column: false,
|
|
writes_table_alias: false,
|
|
writes_cte_name: false,
|
|
writes_projection_alias: false,
|
|
};
|
|
static QUALIFIED_REF_TAIL_NODES: &[Node] = &[
|
|
Node::Punct('.'),
|
|
QUALIFIED_REF_IDENT,
|
|
];
|
|
|
|
static NAME_OR_CALL_TAIL_CHOICES: &[Node] = &[
|
|
Node::Seq(QUALIFIED_REF_TAIL_NODES),
|
|
Node::Seq(CALL_TAIL_NODES),
|
|
];
|
|
static NAME_OR_CALL_TAIL: Node = Node::Choice(NAME_OR_CALL_TAIL_CHOICES);
|
|
|
|
static NAME_OR_CALL_NODES: &[Node] = &[EXPR_IDENT, Node::Optional(&NAME_OR_CALL_TAIL)];
|
|
|
|
/// `primary`. Keyword literals (`null` / `true` / `false`) and the
|
|
/// `CASE` / `EXISTS` keywords come before `name_or_call`, so they
|
|
/// parse as what they are rather than as column references.
|
|
static PRIMARY_CHOICES: &[Node] = &[
|
|
Node::Word(Word::keyword("null")),
|
|
Node::Word(Word::keyword("true")),
|
|
Node::Word(Word::keyword("false")),
|
|
Node::NumberLit { validator: None },
|
|
Node::StringLit,
|
|
Node::Seq(EXISTS_PRIMARY_NODES),
|
|
Node::Seq(PAREN_GROUP_NODES),
|
|
Node::Seq(CASE_NODES),
|
|
Node::Seq(NAME_OR_CALL_NODES),
|
|
];
|
|
static PRIMARY: Node = Node::Choice(PRIMARY_CHOICES);
|
|
|
|
// =================================================================
|
|
// Tests
|
|
// =================================================================
|
|
|
|
#[cfg(test)]
|
|
mod tests {
|
|
use super::SQL_OR_EXPR;
|
|
use crate::dsl::walker::context::WalkContext;
|
|
use crate::dsl::walker::driver::{NodeWalkResult, walk_node};
|
|
use crate::dsl::walker::outcome::MatchedPath;
|
|
|
|
/// Walk `input` against the SQL expression fragment. Returns
|
|
/// `true` only when the walk matches *and* consumes all of
|
|
/// `input` (trailing whitespace allowed).
|
|
fn walks(input: &str) -> bool {
|
|
let mut ctx = WalkContext::new();
|
|
let mut path = MatchedPath::new();
|
|
let mut per_byte = Vec::new();
|
|
match walk_node(input, 0, &SQL_OR_EXPR, &mut ctx, &mut path, &mut per_byte) {
|
|
NodeWalkResult::Matched { end, .. } => input[end..].trim().is_empty(),
|
|
_ => false,
|
|
}
|
|
}
|
|
|
|
/// Assert `input` is a complete SQL expression.
|
|
fn good(input: &str) {
|
|
assert!(walks(input), "{input:?} should be a valid SQL expression");
|
|
}
|
|
|
|
/// Assert `input` is *not* a complete SQL expression.
|
|
fn bad(input: &str) {
|
|
assert!(!walks(input), "{input:?} should NOT walk as a complete expression");
|
|
}
|
|
|
|
#[test]
|
|
fn bare_operands_are_expressions() {
|
|
// A projection item / `WHERE 1` — no predicate operator.
|
|
for input in ["1", "col", "'text'", "true", "false", "null", "-7"] {
|
|
good(input);
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn every_comparison_operator() {
|
|
for op in ["=", "!=", "<>", "<", "<=", ">", ">="] {
|
|
good(&format!("a {op} 1"));
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn arithmetic_and_precedence() {
|
|
for input in [
|
|
"a + b",
|
|
"a - b",
|
|
"a * b",
|
|
"a / b",
|
|
"a % b",
|
|
"a + b * c",
|
|
"(a + b) * c",
|
|
"a + b - c + d",
|
|
"-a + b",
|
|
"a - -b",
|
|
"- -a",
|
|
] {
|
|
good(input);
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn string_concatenation() {
|
|
good("first || ' ' || last");
|
|
}
|
|
|
|
#[test]
|
|
fn function_calls() {
|
|
for input in [
|
|
"upper(name)",
|
|
"length(x)",
|
|
"f()",
|
|
"round(price, 2)",
|
|
"count(*)",
|
|
"count(distinct customer_id)",
|
|
"coalesce(a, b, c)",
|
|
"upper(lower(name))",
|
|
] {
|
|
good(input);
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn case_searched_and_simple() {
|
|
good("case when a > 0 then 'pos' else 'neg' end");
|
|
good("case when a > 0 then 1 when a < 0 then -1 else 0 end");
|
|
good("case grade when 1 then 'A' when 2 then 'B' end");
|
|
good("case status when 'open' then 1 else 0 end");
|
|
}
|
|
|
|
#[test]
|
|
fn the_predicate_set() {
|
|
good("name like 'A%'");
|
|
good("name not like 'A%'");
|
|
good("age between 18 and 65");
|
|
good("age not between 0 and 17");
|
|
good("status in (1, 2, 3)");
|
|
good("status not in ('a', 'b')");
|
|
good("email is null");
|
|
good("email is not null");
|
|
}
|
|
|
|
#[test]
|
|
fn boolean_connectives_and_precedence() {
|
|
good("a = 1 and b = 2");
|
|
good("a = 1 or b = 2");
|
|
good("a = 1 or b = 2 and c = 3");
|
|
good("not a = 1");
|
|
good("not (a = 1 or b = 2)");
|
|
good("(a = 1 or b = 2) and c = 3");
|
|
}
|
|
|
|
#[test]
|
|
fn predicates_over_arithmetic_operands() {
|
|
// The operands of a comparison are full scalar
|
|
// expressions — the whole point of the superset.
|
|
good("price * 1.1 > budget");
|
|
good("upper(name) = 'ADA'");
|
|
good("a + b between c and d");
|
|
}
|
|
|
|
#[test]
|
|
fn keywords_are_case_insensitive() {
|
|
good("A = 1 AND B IS NOT NULL");
|
|
good("CASE WHEN x > 0 THEN 1 ELSE 0 END");
|
|
good("name LIKE 'x%' OR age BETWEEN 1 AND 9");
|
|
}
|
|
|
|
#[test]
|
|
fn nested_parentheses_walk() {
|
|
good("((a = 1 and b = 2) or (c = 3))");
|
|
good("(((1)))");
|
|
}
|
|
|
|
#[test]
|
|
fn malformed_expressions_do_not_walk() {
|
|
bad("a +"); // dangling operator
|
|
bad("a in b"); // IN requires a parenthesised list
|
|
bad("= 1"); // no left operand
|
|
bad("a = "); // no right operand
|
|
bad("case a end"); // CASE with no WHEN clause
|
|
bad("and b"); // leading connective
|
|
bad("upper("); // unclosed call
|
|
}
|
|
|
|
#[test]
|
|
fn deeply_nested_parentheses_are_capped() {
|
|
// Far past MAX_SUBGRAMMAR_DEPTH — must fail with a
|
|
// friendly error rather than overflowing the stack.
|
|
let depth = 400;
|
|
let input = format!("{}1{}", "(".repeat(depth), ")".repeat(depth));
|
|
assert!(!walks(&input), "pathological nesting must be rejected");
|
|
}
|
|
|
|
// ---- ADR-0032 §5 additive: qualified column references ----
|
|
|
|
#[test]
|
|
fn qualified_ref_basic_shapes() {
|
|
good("t.c");
|
|
good("t.c = 1");
|
|
good("a.b + c.d");
|
|
good("upper(t.name)");
|
|
good("t.a is null");
|
|
good("t.x like 'A%'");
|
|
good("t.x between 1 and 10");
|
|
good("t.x in (1, 2, 3)");
|
|
}
|
|
|
|
#[test]
|
|
fn qualified_ref_function_call_disambiguation() {
|
|
// The optional tail dispatches `.identifier` (qualified
|
|
// ref) vs `(args)` (function call) by first token — a
|
|
// bare ident remains a column ref.
|
|
good("foo(x)"); // function call
|
|
good("foo.bar"); // qualified ref
|
|
good("foo"); // bare ref
|
|
}
|
|
|
|
#[test]
|
|
fn qualified_ref_not_admitted_as_function() {
|
|
// No schema.fn() form — qualified ref and call don't
|
|
// compose. `t.foo(x)` would only parse as `t.foo`
|
|
// followed by `(x)` — but `(x)` is not a valid
|
|
// continuation of an expression, so the walk fails.
|
|
bad("t.foo(x)");
|
|
}
|
|
|
|
// ---- ADR-0032 §6 additive: subquery expressions ----
|
|
|
|
#[test]
|
|
fn scalar_subquery_as_primary() {
|
|
good("(select 1)");
|
|
good("x = (select y from t)");
|
|
good("(select count(*) from t) > 100");
|
|
good("upper((select name from t where id = 1))");
|
|
}
|
|
|
|
#[test]
|
|
fn scalar_subquery_dispatches_against_paren_group() {
|
|
// Both `(or_expr)` and `(SELECT …)` start with `(`.
|
|
// The ADR factors the `(` and the first inside token
|
|
// discriminates — `SELECT` → subquery, anything else
|
|
// → expression. (Per ADR-0032 §1 / §9, subqueries
|
|
// recurse through `SQL_SELECT_COMPOUND` which omits
|
|
// the outer `WITH` — so `(WITH …)` is NOT admitted as
|
|
// a scalar subquery; that form is only valid at
|
|
// statement top-level.)
|
|
good("(a + 1)");
|
|
good("(select 1)");
|
|
bad("(with x as (select 1) select * from x)");
|
|
}
|
|
|
|
#[test]
|
|
fn in_subquery_predicate() {
|
|
good("x in (select y from t)");
|
|
good("x not in (select y from t)");
|
|
good("x in (select y from t union select z from u)");
|
|
}
|
|
|
|
#[test]
|
|
fn in_value_list_still_works() {
|
|
// The existing IN-value-list form is preserved
|
|
// alongside the new IN-subquery form.
|
|
good("status in (1, 2, 3)");
|
|
good("name not in ('a', 'b', 'c')");
|
|
}
|
|
|
|
#[test]
|
|
fn exists_primary() {
|
|
good("exists (select 1)");
|
|
good("not exists (select 1)");
|
|
good("exists (select 1 from t where x = 1)");
|
|
good("exists (select * from t) and a > 0");
|
|
}
|
|
|
|
#[test]
|
|
fn subquery_recursion_through_compound() {
|
|
good("x in (select y from t where y in (select z from u))");
|
|
good("exists (select 1 from t where exists (select 1 from u))");
|
|
}
|
|
}
|