feat: add Parsed and deinit

src/expr.zig

@@ -0,0 +1,157 @@
+const std = @import("std");
+const rowmod = @import("row.zig");
+const idx = @import("index.zig");
+const querymod = @import("query.zig");
+const dbmod = @import("db.zig");
+
+const Allocator = std.mem.Allocator;
+const Predicate = querymod.Predicate;
+const Condition = querymod.Condition;
+
+/// Errors a caller can act on when a `table select` expression is rejected.
+/// Parsing never panics on malformed input - every failure mode a user can
+/// trigger from the CLI maps to one of these.
+pub const ParseError = error{
+    UnexpectedToken,
+    UnexpectedEnd,
+    InvalidIdentifier,
+    UnknownOperator,
+    InvalidValue,
+} || Allocator.Error;
+
+/// Owns every allocation a parsed Predicate tree depends on (field-name and
+/// text-value copies, and_/or_ slices, not_ boxes) in one arena, so freeing
+/// is a single `deinit()` regardless of how deep the tree is.
+pub const Parsed = struct {
+    arena: std.heap.ArenaAllocator,
+    predicate: Predicate,
+
+    pub fn deinit(self: *Parsed) void {
+        self.arena.deinit();
+    }
+};
+
+const TokenKind = enum { lparen, rparen, word, eof };
+
+const Token = struct {
+    kind: TokenKind,
+    text: []const u8,
+};
+
+const Lexer = struct {
+    src: []const u8,
+    pos: usize = 0,
+
+    fn next(self: *Lexer) Token {
+        while (self.pos < self.src.len and std.ascii.isWhitespace(self.src[self.pos])) : (self.pos += 1) {}
+        if (self.pos >= self.src.len) return .{ .kind = .eof, .text = "" };
+
+        const c = self.src[self.pos];
+        if (c == '(') {
+            self.pos += 1;
+            return .{ .kind = .lparen, .text = "(" };
+        }
+        if (c == ')') {
+            self.pos += 1;
+            return .{ .kind = .rparen, .text = ")" };
+        }
+
+        const start = self.pos;
+        while (self.pos < self.src.len) {
+            const ch = self.src[self.pos];
+            if (std.ascii.isWhitespace(ch) or ch == '(' or ch == ')') break;
+            self.pos += 1;
+        }
+        return .{ .kind = .word, .text = self.src[start..self.pos] };
+    }
+};
+
+fn isKeyword(tok: Token, kw: []const u8) bool {
+    return tok.kind == .word and std.mem.eql(u8, tok.text, kw);
+}
+
+/// Recursive-descent parser for the `table select` predicate language:
+///
+///   expr    := and ("or" and)*
+///   and     := unary ("and" unary)*
+///   unary   := "not" unary | primary
+///   primary := "(" expr ")" | field op value | field ("is_null" | "is_not_null")
+///   op      := eq | ne | lt | lte | gt | gte | like
+///
+/// Precedence (loosest to tightest) is or < and < not, matching every other
+/// C-family boolean expression language, so `a eq 1 or b eq 2 and not c eq 3`
+/// parses as `a eq 1 or (b eq 2 and (not (c eq 3)))`.
+const Parser = struct {
+    lexer: Lexer,
+    allocator: Allocator,
+    cur: Token,
+
+    fn advance(self: *Parser) void {
+        self.cur = self.lexer.next();
+    }
+
+    fn parseOr(self: *Parser) ParseError!Predicate {
+        var list = std.ArrayList(Predicate).init(self.allocator);
+        try list.append(try self.parseAnd());
+        while (isKeyword(self.cur, "or")) {
+            self.advance();
+            try list.append(try self.parseAnd());
+        }
+        if (list.items.len == 1) return list.items[0];
+        return .{ .or_ = try list.toOwnedSlice() };
+    }
+
+    fn parseAnd(self: *Parser) ParseError!Predicate {
+        var list = std.ArrayList(Predicate).init(self.allocator);
+        try list.append(try self.parseUnary());
+        while (isKeyword(self.cur, "and")) {
+            self.advance();
+            try list.append(try self.parseUnary());
+        }
+        if (list.items.len == 1) return list.items[0];
+        return .{ .and_ = try list.toOwnedSlice() };
+    }
+
+    fn parseUnary(self: *Parser) ParseError!Predicate {
+        if (isKeyword(self.cur, "not")) {
+            self.advance();
+            const inner = try self.parseUnary();
+            const boxed = try self.allocator.create(Predicate);
+            boxed.* = inner;
+            return .{ .not_ = boxed };
+        }
+        return self.parsePrimary();
+    }
+
+    fn parsePrimary(self: *Parser) ParseError!Predicate {
+        if (self.cur.kind == .lparen) {
+            self.advance();
+            const inner = try self.parseOr();
+            if (self.cur.kind != .rparen) return error.UnexpectedToken;
+            self.advance();
+            return inner;
+        }
+        return self.parseCondition();
+    }
+
+    fn parseCondition(self: *Parser) ParseError!Predicate {
+        if (self.cur.kind != .word) return if (self.cur.kind == .eof) error.UnexpectedEnd else error.UnexpectedToken;
+        const field = self.cur.text;
+        if (!idx.validIdentifier(field)) return error.InvalidIdentifier;
+        const fieldCopy = try self.allocator.dupe(u8, field);
+        self.advance();
+
+        if (self.cur.kind != .word) return error.UnexpectedEnd;
+        const op = self.cur.text;
+        self.advance();
+
+        if (std.mem.eql(u8, op, "in")) return self.parseIn(fieldCopy);
+        if (std.mem.eql(u8, op, "between")) return self.parseBetween(fieldCopy);
+        if (std.mem.eql(u8, op, "is_null")) return .{ .is_null = fieldCopy };
+        if (std.mem.eql(u8, op, "is_not_null")) return .{ .is_not_null = fieldCopy };
+
+        if (self.cur.kind != .word) return error.UnexpectedEnd;
+        const rawValue = rowmod.Value.parseCli(self.cur.text) catch return error.InvalidValue;
+        const value = try rawValue.dupe(self.allocator);
+        self.advance();
+