@askdb/core 1.0.0-beta.51 → 1.0.0-beta.52

package/dist/sql/readings.js

@@ -0,0 +1,287 @@
+/**
+ * How each SQL check reads a statement, per dialect: the lexer readings it must pass under. This
+ * module is the one place that decides them (#815, ADR 0010's stage (c), "one shared analysis"),
+ * and {@link readSql} lexes each reading once per statement, so a reading changes in one place.
+ * The checks stay heuristics (ADR 0012).
+ *
+ * The sets differ by check. #815 moved them here unchanged; merging them would tighten some checks
+ * and loosen others, so that is its own decision.
+ *
+ * | Check | A built-in engine | Any other id, or no dialect |
+ * |---|---|---|
+ * | read-only ({@link readOnlyReadings}) | the engine's profile | every engine's profile |
+ * | sensitive ({@link sensitiveReadings}) | the engine's (Postgres family: plus CockroachDB's comment end) | the generic profile, plus each engine's that lexes cleanly |
+ * | tenant: statements and calls ({@link tenantStatementReadings}) | the engine's (Postgres family: both comment ends) | every engine's, Postgres and MySQL with both string settings |
+ * | tenant: predicates ({@link tenantCodeReadings}) | the engine's view | the standard, Postgres, MySQL and SQL Server views, the first two with both comment ends; the Postgres and SQL Server views with block comments nested and not |
+ * | placeholders ({@link placeholderReading}) | the engine's | the generic profile |
+ * | names outside printable ASCII, tenant and sensitive ({@link unmatchedNonAsciiName}) | `sqlserver`: the engine's; any other engine: none | as for tenant statements |
+ * | `'…'` read as a name or an alias, tenant and sensitive ({@link stringsSqliteReadsAsNames}) | `sqlite`: its readings; any other engine: none | the generic and standard readings, and SQLite's |
+ * | Postgres-only forms (`postgresReadingProfiles` in `lexer.ts`) | the engine's, when it is on the Postgres lexer | Postgres's, with backslash escapes too when set |
+ *
+ * Every MySQL reading is taken twice by the tenant check, with `/*! … *\/` bodies as code and as
+ * comments, since a server older than the comment's version skips the body. Its Postgres and SQL
+ * Server views, which nest block comments as those engines do, are taken twice too, with the
+ * comments nested and not, so a predicate counts only where the two agree on what is code. A `postgres` dialect
+ * may point at CockroachDB, which is wire-compatible but ends a `--` comment only at `\n`, so the
+ * Postgres family is read with both comment ends where that decides what is code.
+ */
+import { isBuiltInDialectId } from "./dialect-spec.js";
+import { ENGINE_LEXER_PROFILES, GENERIC_LEXER, lexerProfileFor, lexSql } from "./lexer.js";
+/**
+ * Whether `dialect` names a built-in engine family, which the checks read its own way. Any other
+ * id is read every engine's way, and checked against every built-in spec's denylist.
+ */
+export function isKnownEngine(dialect) {
+    return lexerProfileFor(dialect) !== undefined;
+}
+/** A Postgres-family profile, ended at `\n` only the way CockroachDB ends a `--` comment. */
+function asCockroachDb(profile) {
+    return { ...profile, lineCommentEndsAtLfOnly: true };
+}
+/**
+ * The read-only check (`validateSelectSql`): the engine's profile, or every engine's. It refuses
+ * every comment, so where a comment ends doesn't change its verdict.
+ */
+export function readOnlyReadings(dialect) {
+    const profile = lexerProfileFor(dialect);
+    return profile ? [profile] : ENGINE_LEXER_PROFILES;
+}
+/**
+ * The sensitive check. With a built-in engine, its profile (for the Postgres family, the references
+ * CockroachDB's comment end shows are added too). Otherwise the generic profile decides scope issues
+ * (`primary`), and the references found under each engine's profile that lexes cleanly are added to
+ * it (`union`), so a column one engine's quoting rules hide is still reported.
+ */
+export function sensitiveReadings(dialect) {
+    const profile = dialect ? lexerProfileFor(dialect) : undefined;
+    if (!profile)
+        return { primary: GENERIC_LEXER, union: ENGINE_LEXER_PROFILES };
+    return { primary: profile, union: profile.id === "postgres" ? [asCockroachDb(profile)] : [] };
+}
+/**
+ * Every way a server may lex `sql` into statements, for the tenant check's statement and call
+ * checks: the dialect's profile when it names a built-in engine, otherwise every engine's, and for
+ * Postgres and MySQL both string settings (a backslash ends a literal under one and escapes the
+ * quote under the other). A MySQL profile is read with its `/*!NNNNN … *\/` bodies as code and as
+ * comments, since a server older than NNNNN skips them, and a Postgres-family one with both
+ * comment ends.
+ */
+export function tenantStatementReadings(dialect) {
+    const known = dialect?.id !== undefined ? lexerProfileFor(dialect) : undefined;
+    const profiles = known
+        ? known.id === "postgres" ? [known, asCockroachDb(known)] : [known]
+        : ENGINE_LEXER_PROFILES.flatMap((p) => p.id === "postgres" || p.id === "mysql" ? [p, { ...p, backslashEscapes: !p.backslashEscapes }] : [p]);
+    return bothExecutableCommentReadings(profiles);
+}
+/** Printable ASCII, the only characters a name can hold for the checks to match it by spelling everywhere. */
+const PRINTABLE_ASCII = /^[\x20-\x7e]*$/;
+/**
+ * The first name in `read`, quoted or not, that holds a character outside printable ASCII and isn't
+ * one of `known` exactly, as written; undefined when there is none. SQL Server resolves a name
+ * through the database collation, which ignores some characters (U+200D, U+FEFF, any emoji) and
+ * reads others as ASCII letters (fullwidth letters, small capitals), so such a name can reach a
+ * table or column that the tenant and sensitive checks, which match names by spelling, never see.
+ * Read under `sqlserver`, and under any id that is no built-in engine or no dialect, each reading
+ * from {@link tenantStatementReadings}; the other built-in engines compare such a name as written.
+ */
+export function unmatchedNonAsciiName(read, dialect, known) {
+    if (dialect !== undefined && isBuiltInDialectId(dialect.id) && dialect.id !== "sqlserver")
+        return undefined;
+    for (const profile of tenantStatementReadings(dialect)) {
+        for (const token of read.tokens(profile)) {
+            const name = token.kind === "word" ? token.text : token.kind === "quoted_identifier" ? token.value : undefined;
+            if (name !== undefined && !PRINTABLE_ASCII.test(name) && !known.has(name))
+                return token.text;
+        }
+    }
+    return undefined;
+}
+/** Whether `profile` may be SQLite's reading: SQLite's own, or the generic and standard ones a dialect that may be SQLite gets. */
+export function mayBeSqlite(profile) {
+    return profile.id === "sqlite" || profile.id === "generic";
+}
+/** Words that start a clause other than `FROM`, so a `,` after one is in no `FROM` list. */
+const NOT_A_FROM_LIST = new Set(["select", "where", "group", "having", "order", "limit", "offset", "window", "values", "returning", "union", "intersect", "except", "set", "with"]);
+/**
+ * The indices in `tokens` of the `'…'` strings that SQLite reads as names. SQLite takes a
+ * single-quoted literal where only a name may stand as that name, so `FROM 'orders'` reads the
+ * table `orders`. A string counts right after `FROM`, `JOIN` or `IN` (`IN 'orders'` reads a table),
+ * after a `,` in a `FROM` list, or beside a `.` (`main.'orders'`, `'main'.orders`). With
+ * `withAliases`, it also counts right after a table or CTE reference, or after `AS` following one,
+ * where it is that read's alias (`FROM orders 'o'`). The tenant and sensitive checks read such a
+ * string as a quoted name under a reading that {@link mayBeSqlite may be SQLite's}; only the tenant
+ * check reads aliases, since the sensitive check already fails closed on a qualifier it can't bind.
+ */
+export function stringsSqliteReadsAsNames(tokens, withAliases) {
+    const code = [...tokens.keys()].filter((i) => tokens[i].kind !== "comment");
+    const at = (j) => (j >= 0 && j < code.length ? tokens[code[j]] : undefined);
+    const is = (t, kind, text) => t?.kind === kind && t.lower === text;
+    const inFromList = (comma) => {
+        let depth = 0;
+        for (let j = comma - 1; j >= 0; j--) {
+            const t = at(j);
+            if (is(t, "punct", ")"))
+                depth++;
+            else if (is(t, "punct", "(")) {
+                if (depth === 0)
+                    return false;
+                depth--;
+            }
+            else if (depth === 0 && is(t, "punct", ";"))
+                return false;
+            else if (depth === 0 && t.kind === "word") {
+                if (t.lower === "from")
+                    return true;
+                if (NOT_A_FROM_LIST.has(t.lower))
+                    return false;
+            }
+        }
+        return false;
+    };
+    const names = new Set();
+    /** Positions in `code` of the names that stand in a table's place: `FROM orders`, `JOIN main.orders`. */
+    const sources = new Set();
+    for (let j = 0; j < code.length; j++) {
+        const t = at(j);
+        const prev = at(j - 1);
+        const plainString = t.kind === "string" && t.text.startsWith("'");
+        const startsSource = is(prev, "word", "from") || is(prev, "word", "join") ||
+            (is(prev, "punct", ",") && inFromList(j - 1)) || (is(prev, "punct", ".") && sources.has(j - 2));
+        if (startsSource && (t.kind === "word" || t.kind === "quoted_identifier" || plainString))
+            sources.add(j);
+        if (!plainString)
+            continue;
+        const alias = withAliases && (sources.has(j - 1) || (is(prev, "word", "as") && sources.has(j - 2)));
+        const named = alias || is(prev, "punct", ".") || is(at(j + 1), "punct", ".") ||
+            is(prev, "word", "from") || is(prev, "word", "join") || is(prev, "word", "in") ||
+            (is(prev, "punct", ",") && inFromList(j - 1));
+        if (named)
+            names.add(code[j]);
+    }
+    return names;
+}
+/** A view that nests block comments, also read with them ending at the first `*\/`. */
+function bothBlockCommentReadings(profiles) {
+    return profiles.flatMap((p) => (p.nestedBlockComments ? [p, { ...p, nestedBlockComments: false }] : [p]));
+}
+/** A profile that reads `/*!NNNNN … *\/` bodies as code, also read with them as comments. */
+function bothExecutableCommentReadings(profiles) {
+    return profiles.flatMap((p) => (p.executableComments ? [p, { ...p, executableComments: false }] : [p]));
+}
+/**
+ * The tenant check's standard view: how its own scanner split code from strings and comments, kept
+ * as a lexer profile (#815). `'…'` strings with `''` escapes; `"…"`, `` `…` `` and `[…]` quoted
+ * names under every engine, where the first `]` ends a bracketed name; `--` and `/* *\/` comments,
+ * never nested, where a line comment ends only at `\n`.
+ */
+const STANDARD_VIEW = {
+    id: "generic",
+    backslashEscapes: false,
+    doubleQuote: "identifier",
+    backtickIdentifiers: true,
+    bracketIdentifiers: "simple",
+    dollarQuotes: false,
+    escapeStrings: false,
+    hashComments: false,
+    dashCommentNeedsSpace: false,
+    nestedBlockComments: false,
+    executableComments: false,
+    atVariables: false,
+    hashIdentifiers: false,
+    tsqlLiterals: false,
+    lineCommentEndsAtLfOnly: true,
+    strictStringPrefixes: true,
+};
+/**
+ * The standard view, with Postgres `E'…'` escape strings, `$tag$…$tag$` strings, and block comments
+ * nested the way Postgres and CockroachDB nest them.
+ */
+const POSTGRES_VIEW = {
+    ...STANDARD_VIEW,
+    id: "postgres",
+    escapeStrings: true,
+    dollarQuotes: true,
+    nestedBlockComments: true,
+};
+/**
+ * The standard view, read the MySQL way: `"…"` is a string, backslash escapes, `#` comments, `--`
+ * only before whitespace, and `/*! … *\/` bodies as code.
+ */
+const MYSQL_VIEW = {
+    ...STANDARD_VIEW,
+    id: "mysql",
+    doubleQuote: "string",
+    backslashEscapes: true,
+    hashComments: true,
+    dashCommentNeedsSpace: true,
+    executableComments: true,
+};
+/** A view whose line comments end at `\r` too, as on Postgres and SQL Server. */
+const crEnds = (view) => ({ ...view, lineCommentEndsAtLfOnly: false });
+/**
+ * The standard view, read the SQL Server way (#623): a line comment ends at `\r` too, a name or
+ * number ends where SQL Server ends it (`tsqlLiterals`: after `0x` or `1E`, and at whitespace such
+ * as U+00A0, U+200B and U+0085), so a keyword written right after one is a word of its own, `]]`
+ * inside a bracketed name is an escaped `]`, and block comments nest.
+ */
+const SQLSERVER_VIEW = {
+    ...crEnds(STANDARD_VIEW),
+    id: "sqlserver",
+    tsqlLiterals: true,
+    bracketIdentifiers: "doubled",
+    nestedBlockComments: true,
+};
+/**
+ * The tenant check's predicate readings. A built-in engine gets its own view: the MySQL one twice,
+ * with executable comments as code and as comments; the Postgres one with both comment ends; the
+ * SQL Server one. The Postgres and SQL Server views nest block comments, and each is read again with
+ * them not nested. Backslash escapes follow the spec as the lexer reads it: an unset value on MySQL
+ * or MariaDB is the server default, on. Any other id, or none, gets the standard and Postgres views
+ * with both comment ends, the MySQL view and the SQL Server view: a table counts as referenced when
+ * any of them sees it, a predicate only when every one does.
+ */
+export function tenantCodeReadings(dialect) {
+    const views = (profiles) => bothExecutableCommentReadings(bothBlockCommentReadings(profiles)).map((profile) => ({ profile, pipesAreOr: profile.id === "mysql" }));
+    if (dialect === undefined || !isBuiltInDialectId(dialect.id)) {
+        return views([STANDARD_VIEW, crEnds(STANDARD_VIEW), POSTGRES_VIEW, crEnds(POSTGRES_VIEW), MYSQL_VIEW, SQLSERVER_VIEW]);
+    }
+    const mysqlFamily = dialect.id === "mysql" || dialect.id === "mariadb";
+    const backslashEscapes = mysqlFamily ? dialect.backslashEscapes !== false : dialect.backslashEscapes === true;
+    const engine = mysqlFamily
+        ? [MYSQL_VIEW]
+        : dialect.id === "postgres" || dialect.id === "cockroachdb"
+            ? [POSTGRES_VIEW, crEnds(POSTGRES_VIEW)]
+            : dialect.id === "sqlserver"
+                ? [SQLSERVER_VIEW]
+                : [STANDARD_VIEW];
+    return views(engine.map((view) => ({ ...view, backslashEscapes })));
+}
+/** Placeholder scans (`tokenizeSqlSpans`): the engine's profile, or the generic one. */
+export function placeholderReading(dialect) {
+    return (dialect && lexerProfileFor(dialect)) ?? GENERIC_LEXER;
+}
+/** Read `sql`: {@link SqlRead.tokens} lexes it under a profile the first time it is asked for one. */
+export function readSql(sql) {
+    const lexed = new Map();
+    return {
+        sql,
+        tokens(profile) {
+            const key = profileKey(profile);
+            let tokens = lexed.get(key);
+            if (tokens === undefined) {
+                tokens = lexSql(sql, profile);
+                lexed.set(key, tokens);
+            }
+            return tokens;
+        },
+    };
+}
+/** Two profiles with the same rules lex the same tokens. */
+function profileKey(p) {
+    return [
+        p.id, p.backslashEscapes, p.doubleQuote, p.backtickIdentifiers, p.bracketIdentifiers, p.dollarQuotes,
+        p.escapeStrings, p.hashComments, p.dashCommentNeedsSpace, p.nestedBlockComments, p.executableComments,
+        p.atVariables, p.hashIdentifiers, p.tsqlLiterals, p.lineCommentEndsAtLfOnly === true, p.strictStringPrefixes === true,
+    ].join("|");
+}
+//# sourceMappingURL=readings.js.map

package/dist/sql/readings.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"readings.js","sourceRoot":"","sources":["../../src/sql/readings.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,OAAO,EAAE,kBAAkB,EAAoB,MAAM,mBAAmB,CAAC;AACzE,OAAO,EAAE,qBAAqB,EAAE,aAAa,EAAE,eAAe,EAAE,MAAM,EAAuC,MAAM,YAAY,CAAC;AAKhI;;;GAGG;AACH,MAAM,UAAU,aAAa,CAAC,OAAuB;IACnD,OAAO,eAAe,CAAC,OAAO,CAAC,KAAK,SAAS,CAAC;AAChD,CAAC;AAED,6FAA6F;AAC7F,SAAS,aAAa,CAAC,OAAwB;IAC7C,OAAO,EAAE,GAAG,OAAO,EAAE,uBAAuB,EAAE,IAAI,EAAE,CAAC;AACvD,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,gBAAgB,CAAC,OAAuB;IACtD,MAAM,OAAO,GAAG,eAAe,CAAC,OAAO,CAAC,CAAC;IACzC,OAAO,OAAO,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,qBAAqB,CAAC;AACrD,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,iBAAiB,CAAC,OAAmC;IAInE,MAAM,OAAO,GAAG,OAAO,CAAC,CAAC,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/D,IAAI,CAAC,OAAO;QAAE,OAAO,EAAE,OAAO,EAAE,aAAa,EAAE,KAAK,EAAE,qBAAqB,EAAE,CAAC;IAC9E,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,OAAO,CAAC,EAAE,KAAK,UAAU,CAAC,CAAC,CAAC,CAAC,aAAa,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC;AAChG,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,uBAAuB,CAAC,OAAmC;IACzE,MAAM,KAAK,GAAG,OAAO,EAAE,EAAE,KAAK,SAAS,CAAC,CAAC,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/E,MAAM,QAAQ,GAAG,KAAK;QACpB,CAAC,CAAC,KAAK,CAAC,EAAE,KAAK,UAAU,CAAC,CAAC,CAAC,CAAC,KAAK,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;QACnE,CAAC,CAAC,qBAAqB,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,CAClC,CAAC,CAAC,EAAE,KAAK,UAAU,IAAI,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,CAAC,EAAE,gBAAgB,EAAE,CAAC,CAAC,CAAC,gBAAgB,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CACrG,CAAC;IACN,OAAO,6BAA6B,CAAC,QAAQ,CAAC,CAAC;AACjD,CAAC;AAED,8GAA8G;AAC9G,MAAM,eAAe,GAAG,gBAAgB,CAAC;AAEzC;;;;;;;;GAQG;AACH,MAAM,UAAU,qBAAqB,CACnC,IAAa,EACb,OAAmC,EACnC,KAA0B;IAE1B,IAAI,OAAO,KAAK,SAAS,IAAI,kBAAkB,CAAC,OAAO,CAAC,EAAE,CAAC,IAAI,OAAO,CAAC,EAAE,KAAK,WAAW;QAAE,OAAO,SAAS,CAAC;IAC5G,KAAK,MAAM,OAAO,IAAI,uBAAuB,CAAC,OAAO,CAAC,EAAE,CAAC;QACvD,KAAK,MAAM,KAAK,IAAI,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE,CAAC;YACzC,MAAM,IAAI,GAAG,KAAK,CAAC,IAAI,KAAK,MAAM,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,KAAK,mBAAmB,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS,CAAC;YAC/G,IAAI,IAAI,KAAK,SAAS,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC;gBAAE,OAAO,KAAK,CAAC,IAAI,CAAC;QAC/F,CAAC;IACH,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED,mIAAmI;AACnI,MAAM,UAAU,WAAW,CAAC,OAAwB;IAClD,OAAO,OAAO,CAAC,EAAE,KAAK,QAAQ,IAAI,OAAO,CAAC,EAAE,KAAK,SAAS,CAAC;AAC7D,CAAC;AAED,4FAA4F;AAC5F,MAAM,eAAe,GAAG,IAAI,GAAG,CAAC,CAAC,QAAQ,EAAE,OAAO,EAAE,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,OAAO,EAAE,QAAQ,EAAE,QAAQ,EAAE,QAAQ,EAAE,WAAW,EAAE,OAAO,EAAE,WAAW,EAAE,QAAQ,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC,CAAC;AAEpL;;;;;;;;;GASG;AACH,MAAM,UAAU,yBAAyB,CAAC,MAA2B,EAAE,WAAoB;IACzF,MAAM,IAAI,GAAG,CAAC,GAAG,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,CAAC,CAAE,CAAC,IAAI,KAAK,SAAS,CAAC,CAAC;IAC7E,MAAM,EAAE,GAAG,CAAC,CAAS,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC;IACrF,MAAM,EAAE,GAAG,CAAC,CAAuB,EAAE,IAAsB,EAAE,IAAY,EAAE,EAAE,CAAC,CAAC,EAAE,IAAI,KAAK,IAAI,IAAI,CAAC,CAAC,KAAK,KAAK,IAAI,CAAC;IACnH,MAAM,UAAU,GAAG,CAAC,KAAa,EAAE,EAAE;QACnC,IAAI,KAAK,GAAG,CAAC,CAAC;QACd,KAAK,IAAI,CAAC,GAAG,KAAK,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;YACpC,MAAM,CAAC,GAAG,EAAE,CAAC,CAAC,CAAE,CAAC;YACjB,IAAI,EAAE,CAAC,CAAC,EAAE,OAAO,EAAE,GAAG,CAAC;gBAAE,KAAK,EAAE,CAAC;iBAC5B,IAAI,EAAE,CAAC,CAAC,EAAE,OAAO,EAAE,GAAG,CAAC,EAAE,CAAC;gBAC7B,IAAI,KAAK,KAAK,CAAC;oBAAE,OAAO,KAAK,CAAC;gBAC9B,KAAK,EAAE,CAAC;YACV,CAAC;iBAAM,IAAI,KAAK,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,EAAE,OAAO,EAAE,GAAG,CAAC;gBAAE,OAAO,KAAK,CAAC;iBACvD,IAAI,KAAK,KAAK,CAAC,IAAI,CAAC,CAAC,IAAI,KAAK,MAAM,EAAE,CAAC;gBAC1C,IAAI,CAAC,CAAC,KAAK,KAAK,MAAM;oBAAE,OAAO,IAAI,CAAC;gBACpC,IAAI,eAAe,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC;oBAAE,OAAO,KAAK,CAAC;YACjD,CAAC;QACH,CAAC;QACD,OAAO,KAAK,CAAC;IACf,CAAC,CAAC;IACF,MAAM,KAAK,GAAG,IAAI,GAAG,EAAU,CAAC;IAChC,yGAAyG;IACzG,MAAM,OAAO,GAAG,IAAI,GAAG,EAAU,CAAC;IAClC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACrC,MAAM,CAAC,GAAG,EAAE,CAAC,CAAC,CAAE,CAAC;QACjB,MAAM,IAAI,GAAG,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;QACvB,MAAM,WAAW,GAAG,CAAC,CAAC,IAAI,KAAK,QAAQ,IAAI,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC;QAClE,MAAM,YAAY,GAChB,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC;YACpD,CAAC,EAAE,CAAC,IAAI,EAAE,OAAO,EAAE,GAAG,CAAC,IAAI,UAAU,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,OAAO,EAAE,GAAG,CAAC,IAAI,OAAO,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAClG,IAAI,YAAY,IAAI,CAAC,CAAC,CAAC,IAAI,KAAK,MAAM,IAAI,CAAC,CAAC,IAAI,KAAK,mBAAmB,IAAI,WAAW,CAAC;YAAE,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QACzG,IAAI,CAAC,WAAW;YAAE,SAAS;QAC3B,MAAM,KAAK,GAAG,WAAW,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,CAAC,IAAI,OAAO,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC;QACpG,MAAM,KAAK,GACT,KAAK,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,EAAE,GAAG,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,EAAE,GAAG,CAAC;YAC9D,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,CAAC;YAC9E,CAAC,EAAE,CAAC,IAAI,EAAE,OAAO,EAAE,GAAG,CAAC,IAAI,UAAU,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAChD,IAAI,KAAK;YAAE,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,CAAE,CAAC,CAAC;IACjC,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,uFAAuF;AACvF,SAAS,wBAAwB,CAAC,QAAoC;IACpE,OAAO,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,mBAAmB,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,CAAC,EAAE,mBAAmB,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAC5G,CAAC;AAED,6FAA6F;AAC7F,SAAS,6BAA6B,CAAC,QAAoC;IACzE,OAAO,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,kBAAkB,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,CAAC,EAAE,kBAAkB,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAC1G,CAAC;AASD;;;;;GAKG;AACH,MAAM,aAAa,GAAoB;IACrC,EAAE,EAAE,SAAS;IACb,gBAAgB,EAAE,KAAK;IACvB,WAAW,EAAE,YAAY;IACzB,mBAAmB,EAAE,IAAI;IACzB,kBAAkB,EAAE,QAAQ;IAC5B,YAAY,EAAE,KAAK;IACnB,aAAa,EAAE,KAAK;IACpB,YAAY,EAAE,KAAK;IACnB,qBAAqB,EAAE,KAAK;IAC5B,mBAAmB,EAAE,KAAK;IAC1B,kBAAkB,EAAE,KAAK;IACzB,WAAW,EAAE,KAAK;IAClB,eAAe,EAAE,KAAK;IACtB,YAAY,EAAE,KAAK;IACnB,uBAAuB,EAAE,IAAI;IAC7B,oBAAoB,EAAE,IAAI;CAC3B,CAAC;AACF;;;GAGG;AACH,MAAM,aAAa,GAAoB;IACrC,GAAG,aAAa;IAChB,EAAE,EAAE,UAAU;IACd,aAAa,EAAE,IAAI;IACnB,YAAY,EAAE,IAAI;IAClB,mBAAmB,EAAE,IAAI;CAC1B,CAAC;AACF;;;GAGG;AACH,MAAM,UAAU,GAAoB;IAClC,GAAG,aAAa;IAChB,EAAE,EAAE,OAAO;IACX,WAAW,EAAE,QAAQ;IACrB,gBAAgB,EAAE,IAAI;IACtB,YAAY,EAAE,IAAI;IAClB,qBAAqB,EAAE,IAAI;IAC3B,kBAAkB,EAAE,IAAI;CACzB,CAAC;AACF,iFAAiF;AACjF,MAAM,MAAM,GAAG,CAAC,IAAqB,EAAmB,EAAE,CAAC,CAAC,EAAE,GAAG,IAAI,EAAE,uBAAuB,EAAE,KAAK,EAAE,CAAC,CAAC;AACzG;;;;;GAKG;AACH,MAAM,cAAc,GAAoB;IACtC,GAAG,MAAM,CAAC,aAAa,CAAC;IACxB,EAAE,EAAE,WAAW;IACf,YAAY,EAAE,IAAI;IAClB,kBAAkB,EAAE,SAAS;IAC7B,mBAAmB,EAAE,IAAI;CAC1B,CAAC;AAEF;;;;;;;;GAQG;AACH,MAAM,UAAU,kBAAkB,CAAC,OAAmC;IACpE,MAAM,KAAK,GAAG,CAAC,QAAoC,EAAE,EAAE,CACrD,6BAA6B,CAAC,wBAAwB,CAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC,EAAE,KAAK,OAAO,EAAE,CAAC,CAAC,CAAC;IACxI,IAAI,OAAO,KAAK,SAAS,IAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,CAAC;QAC7D,OAAO,KAAK,CAAC,CAAC,aAAa,EAAE,MAAM,CAAC,aAAa,CAAC,EAAE,aAAa,EAAE,MAAM,CAAC,aAAa,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,CAAC,CAAC;IACzH,CAAC;IACD,MAAM,WAAW,GAAG,OAAO,CAAC,EAAE,KAAK,OAAO,IAAI,OAAO,CAAC,EAAE,KAAK,SAAS,CAAC;IACvE,MAAM,gBAAgB,GAAG,WAAW,CAAC,CAAC,CAAC,OAAO,CAAC,gBAAgB,KAAK,KAAK,CAAC,CAAC,CAAC,OAAO,CAAC,gBAAgB,KAAK,IAAI,CAAC;IAC9G,MAAM,MAAM,GACV,WAAW;QACT,CAAC,CAAC,CAAC,UAAU,CAAC;QACd,CAAC,CAAC,OAAO,CAAC,EAAE,KAAK,UAAU,IAAI,OAAO,CAAC,EAAE,KAAK,aAAa;YACzD,CAAC,CAAC,CAAC,aAAa,EAAE,MAAM,CAAC,aAAa,CAAC,CAAC;YACxC,CAAC,CAAC,OAAO,CAAC,EAAE,KAAK,WAAW;gBAC1B,CAAC,CAAC,CAAC,cAAc,CAAC;gBAClB,CAAC,CAAC,CAAC,aAAa,CAAC,CAAC;IAC1B,OAAO,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,EAAE,GAAG,IAAI,EAAE,gBAAgB,EAAE,CAAC,CAAC,CAAC,CAAC;AACtE,CAAC;AAED,wFAAwF;AACxF,MAAM,UAAU,kBAAkB,CAAC,OAAmC;IACpE,OAAO,CAAC,OAAO,IAAI,eAAe,CAAC,OAAO,CAAC,CAAC,IAAI,aAAa,CAAC;AAChE,CAAC;AAQD,sGAAsG;AACtG,MAAM,UAAU,OAAO,CAAC,GAAW;IACjC,MAAM,KAAK,GAAG,IAAI,GAAG,EAA+B,CAAC;IACrD,OAAO;QACL,GAAG;QACH,MAAM,CAAC,OAAO;YACZ,MAAM,GAAG,GAAG,UAAU,CAAC,OAAO,CAAC,CAAC;YAChC,IAAI,MAAM,GAAG,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;YAC5B,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;gBACzB,MAAM,GAAG,MAAM,CAAC,GAAG,EAAE,OAAO,CAAC,CAAC;gBAC9B,KAAK,CAAC,GAAG,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;YACzB,CAAC;YACD,OAAO,MAAM,CAAC;QAChB,CAAC;KACF,CAAC;AACJ,CAAC;AAED,4DAA4D;AAC5D,SAAS,UAAU,CAAC,CAAkB;IACpC,OAAO;QACL,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,gBAAgB,EAAE,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC,mBAAmB,EAAE,CAAC,CAAC,kBAAkB,EAAE,CAAC,CAAC,YAAY;QACpG,CAAC,CAAC,aAAa,EAAE,CAAC,CAAC,YAAY,EAAE,CAAC,CAAC,qBAAqB,EAAE,CAAC,CAAC,mBAAmB,EAAE,CAAC,CAAC,kBAAkB;QACrG,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC,eAAe,EAAE,CAAC,CAAC,YAAY,EAAE,CAAC,CAAC,uBAAuB,KAAK,IAAI,EAAE,CAAC,CAAC,oBAAoB,KAAK,IAAI;KACtH,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;AACd,CAAC"}

package/dist/sql/schema-question-precheck.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"schema-question-precheck.d.ts","sourceRoot":"","sources":["../../src/sql/schema-question-precheck.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAS9D,wBAAgB,mBAAmB,CAAC,MAAM,EAAE,mBAAmB,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI,CASvF;AAED;;;;;;;GAOG;AACH,wBAAgB,qBAAqB,CAAC,QAAQ,EAAE,MAAM,EAAE,MAAM,EAAE,mBAAmB,GAAG,MAAM,EAAE,CAwB7F"}
+{"version":3,"file":"schema-question-precheck.d.ts","sourceRoot":"","sources":["../../src/sql/schema-question-precheck.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAS9D,wBAAgB,mBAAmB,CAAC,MAAM,EAAE,mBAAmB,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI,CAOvF;AAED;;;;;;;GAOG;AACH,wBAAgB,qBAAqB,CAAC,QAAQ,EAAE,MAAM,EAAE,MAAM,EAAE,mBAAmB,GAAG,MAAM,EAAE,CAwB7F"}

package/dist/sql/schema-question-precheck.js

@@ -5,10 +5,10 @@ const JOIN_RELATION_HINT = /\b(join|joined|joins|inner\s+join|left\s+join|right\
 const FROM_JOIN_TABLE = /\b(?:from|join)\s+([a-zA-Z_][\w]*)\b/gi;
 export function assertNlToSqlInputs(schema, question) {
     if (schema.tables.length === 0) {
-        throw new AskDbError("Schema has no tables. Add at least one table to the AskDB schema JSON (see fixtures/schemas/ in this repo).");
+        throw new AskDbError("Schema has no tables. Pass a `schema` with at least one table.");
     }
     if (!question.trim()) {
-        throw new AskDbError("Question is empty. Pass a non-blank natural-language question with -q / --question.");
+        throw new AskDbError("Question is empty. Pass a non-blank natural-language `question`.");
     }
 }
 /**

package/dist/sql/schema-question-precheck.js.map

@@ -1 +1 @@
-{"version":3,"file":"schema-question-precheck.js","sourceRoot":"","sources":["../../src/sql/schema-question-precheck.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,cAAc,CAAC;AAE1C,OAAO,EAAE,cAAc,EAAE,MAAM,yBAAyB,CAAC;AAEzD,MAAM,kBAAkB,GACtB,oMAAoM,CAAC;AAEvM,oGAAoG;AACpG,MAAM,eAAe,GAAG,wCAAwC,CAAC;AAEjE,MAAM,UAAU,mBAAmB,CAAC,MAA2B,EAAE,QAAgB;IAC/E,IAAI,MAAM,CAAC,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC/B,MAAM,IAAI,UAAU,CAClB,6GAA6G,CAC9G,CAAC;IACJ,CAAC;IACD,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,EAAE,CAAC;QACrB,MAAM,IAAI,UAAU,CAAC,qFAAqF,CAAC,CAAC;IAC9G,CAAC;AACH,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,qBAAqB,CAAC,QAAgB,EAAE,MAA2B;IACjF,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,wEAAwE;IACxE,MAAM,OAAO,GAAG,UAAU,IAAI,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC;IAC5F,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC;IAErE,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC,IAAI,kBAAkB,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC9D,KAAK,CAAC,IAAI,CACR,kIAAkI,CACnI,CAAC;IACJ,CAAC;IAED,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAC;IAC/B,KAAK,MAAM,CAAC,IAAI,QAAQ,CAAC,QAAQ,CAAC,eAAe,CAAC,EAAE,CAAC;QACnD,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,CAAE,CAAC;QAClB,MAAM,GAAG,GAAG,GAAG,CAAC,WAAW,EAAE,CAAC;QAC9B,IAAI,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC;YAAE,SAAS;QAC5B,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACd,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;YACzB,KAAK,CAAC,IAAI,CAAC,gCAAgC,GAAG,+BAA+B,CAAC,CAAC;QACjF,CAAC;IACH,CAAC;IAED,OAAO,KAAK,CAAC;AACf,CAAC"}
+{"version":3,"file":"schema-question-precheck.js","sourceRoot":"","sources":["../../src/sql/schema-question-precheck.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,cAAc,CAAC;AAE1C,OAAO,EAAE,cAAc,EAAE,MAAM,yBAAyB,CAAC;AAEzD,MAAM,kBAAkB,GACtB,oMAAoM,CAAC;AAEvM,oGAAoG;AACpG,MAAM,eAAe,GAAG,wCAAwC,CAAC;AAEjE,MAAM,UAAU,mBAAmB,CAAC,MAA2B,EAAE,QAAgB;IAC/E,IAAI,MAAM,CAAC,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC/B,MAAM,IAAI,UAAU,CAAC,gEAAgE,CAAC,CAAC;IACzF,CAAC;IACD,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,EAAE,CAAC;QACrB,MAAM,IAAI,UAAU,CAAC,kEAAkE,CAAC,CAAC;IAC3F,CAAC;AACH,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,qBAAqB,CAAC,QAAgB,EAAE,MAA2B;IACjF,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,wEAAwE;IACxE,MAAM,OAAO,GAAG,UAAU,IAAI,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC;IAC5F,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC;IAErE,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC,IAAI,kBAAkB,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC9D,KAAK,CAAC,IAAI,CACR,kIAAkI,CACnI,CAAC;IACJ,CAAC;IAED,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAC;IAC/B,KAAK,MAAM,CAAC,IAAI,QAAQ,CAAC,QAAQ,CAAC,eAAe,CAAC,EAAE,CAAC;QACnD,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,CAAE,CAAC;QAClB,MAAM,GAAG,GAAG,GAAG,CAAC,WAAW,EAAE,CAAC;QAC9B,IAAI,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC;YAAE,SAAS;QAC5B,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACd,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;YACzB,KAAK,CAAC,IAAI,CAAC,gCAAgC,GAAG,+BAA+B,CAAC,CAAC;QACjF,CAAC;IACH,CAAC;IAED,OAAO,KAAK,CAAC;AACf,CAAC"}

package/dist/sql/sensitive-guardrail.d.ts

@@ -64,4 +64,9 @@ export declare function scanSensitiveReferences(sql: string, schema: AnyNormaliz
  */
 export declare function schemaHasSensitiveIdentifiers(schema: AnyNormalizedSchema): boolean;
 export { formatSensitiveReference };
+/**
+ * Every table, namespace and column name in `schema`, exactly as it spells them. Internal to
+ * `@askdb/core`: the tenant check counts a name outside printable ASCII only when it is one of them.
+ */
+export declare function schemaNames(schema: AnyNormalizedSchema): ReadonlySet<string>;
 //# sourceMappingURL=sensitive-guardrail.d.ts.map

package/dist/sql/sensitive-guardrail.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"sensitive-guardrail.d.ts","sourceRoot":"","sources":["../../src/sql/sensitive-guardrail.ts"],"names":[],"mappings":"AAAA,OAAO,EAEL,KAAK,kBAAkB,EAEvB,KAAK,oBAAoB,EAC1B,MAAM,cAAc,CAAC;AACtB,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAC9D,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,EAAU,wBAAwB,EAAoC,MAAM,uBAAuB,CAAC;AAW3G,MAAM,MAAM,sBAAsB,GAAG,MAAM,GAAG,QAAQ,CAAC;AAEvD,MAAM,MAAM,kCAAkC,GAAG;IAC/C;;;;OAIG;IACH,IAAI,CAAC,EAAE,sBAAsB,CAAC;IAC9B;;;;;OAKG;IACH,OAAO,CAAC,EAAE,IAAI,CAAC,WAAW,EAAE,IAAI,GAAG,kBAAkB,CAAC,CAAC;CACxD,CAAC;AAEF,MAAM,MAAM,wBAAwB,GAAG;IACrC,8FAA8F;IAC9F,MAAM,EAAE,OAAO,CAAC;IAChB,UAAU,EAAE,kBAAkB,EAAE,CAAC;IACjC,8FAA8F;IAC9F,eAAe,CAAC,EAAE,oBAAoB,CAAC;CACxC,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,wBAAgB,2BAA2B,CACzC,GAAG,EAAE,MAAM,EACX,MAAM,EAAE,mBAAmB,EAC3B,OAAO,CAAC,EAAE,kCAAkC,GAC3C,wBAAwB,CAM1B;AAED;;;GAGG;AACH,wBAAgB,uBAAuB,CACrC,GAAG,EAAE,MAAM,EACX,MAAM,EAAE,mBAAmB,EAC3B,OAAO,EAAE,IAAI,CAAC,WAAW,EAAE,IAAI,GAAG,kBAAkB,CAAC,GAAG,SAAS,GAChE,wBAAwB,CAG1B;AAED;;;GAGG;AACH,wBAAgB,6BAA6B,CAAC,MAAM,EAAE,mBAAmB,GAAG,OAAO,CAElF;AAED,OAAO,EAAE,wBAAwB,EAAE,CAAC"}
+{"version":3,"file":"sensitive-guardrail.d.ts","sourceRoot":"","sources":["../../src/sql/sensitive-guardrail.ts"],"names":[],"mappings":"AAAA,OAAO,EAEL,KAAK,kBAAkB,EAEvB,KAAK,oBAAoB,EAC1B,MAAM,cAAc,CAAC;AACtB,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAC9D,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,EAAU,wBAAwB,EAAoC,MAAM,uBAAuB,CAAC;AAI3G,MAAM,MAAM,sBAAsB,GAAG,MAAM,GAAG,QAAQ,CAAC;AAEvD,MAAM,MAAM,kCAAkC,GAAG;IAC/C;;;;OAIG;IACH,IAAI,CAAC,EAAE,sBAAsB,CAAC;IAC9B;;;;;OAKG;IACH,OAAO,CAAC,EAAE,IAAI,CAAC,WAAW,EAAE,IAAI,GAAG,kBAAkB,CAAC,CAAC;CACxD,CAAC;AAEF,MAAM,MAAM,wBAAwB,GAAG;IACrC,8FAA8F;IAC9F,MAAM,EAAE,OAAO,CAAC;IAChB,UAAU,EAAE,kBAAkB,EAAE,CAAC;IACjC,8FAA8F;IAC9F,eAAe,CAAC,EAAE,oBAAoB,CAAC;CACxC,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,wBAAgB,2BAA2B,CACzC,GAAG,EAAE,MAAM,EACX,MAAM,EAAE,mBAAmB,EAC3B,OAAO,CAAC,EAAE,kCAAkC,GAC3C,wBAAwB,CAM1B;AAED;;;GAGG;AACH,wBAAgB,uBAAuB,CACrC,GAAG,EAAE,MAAM,EACX,MAAM,EAAE,mBAAmB,EAC3B,OAAO,EAAE,IAAI,CAAC,WAAW,EAAE,IAAI,GAAG,kBAAkB,CAAC,GAAG,SAAS,GAChE,wBAAwB,CAG1B;AAED;;;GAGG;AACH,wBAAgB,6BAA6B,CAAC,MAAM,EAAE,mBAAmB,GAAG,OAAO,CAElF;AAED,OAAO,EAAE,wBAAwB,EAAE,CAAC;AAoFpC;;;GAGG;AACH,wBAAgB,WAAW,CAAC,MAAM,EAAE,mBAAmB,GAAG,WAAW,CAAC,MAAM,CAAC,CAE5E"}

package/dist/sql/sensitive-guardrail.js

@@ -1,5 +1,6 @@
 import { decide, formatSensitiveReference, sensitiveFindings, throwIfDenied } from "./guardrail-decide.js";
-import { ENGINE_LEXER_PROFILES, GENERIC_LEXER, findPostgresOnlyForms, hasUnterminatedToken, lexSql, lexerProfileFor, } from "./lexer.js";
+import { findPostgresOnlyForms, hasUnterminatedToken } from "./lexer.js";
+import { isKnownEngine, mayBeSqlite, readSql, sensitiveReadings, stringsSqliteReadsAsNames, unmatchedNonAsciiName } from "./readings.js";
 /**
  * Check whether a SQL statement references tables/columns marked `sensitive`
  * in the schema artifact.
@@ -88,17 +89,38 @@ function indexSchema(schema) {
         else
             byName.set(entry.lower, [entry]);
     });
-    const index = { tables, byName, hasSensitive };
+    const names = new Set();
+    for (const t of tables) {
+        names.add(t.name);
+        if (t.schema !== undefined)
+            names.add(t.schema);
+        for (const c of t.columns)
+            names.add(c.name);
+    }
+    const index = { tables, byName, hasSensitive, names };
     indexCache.set(schema, index);
     return index;
 }
+/**
+ * Every table, namespace and column name in `schema`, exactly as it spells them. Internal to
+ * `@askdb/core`: the tenant check counts a name outside printable ASCII only when it is one of them.
+ */
+export function schemaNames(schema) {
+    return indexSchema(schema).names;
+}
 /**
  * Map lexer tokens onto the scanner's simpler vocabulary. Comments are dropped (MySQL
- * `/*! … *\/` bodies are already lexed as code).
+ * `/*! … *\/` bodies are already lexed as code). A `'…'` at an index in `names` is a quoted name,
+ * as SQLite reads `FROM 'users'`.
  */
-function toToks(tokens) {
+function toToks(tokens, names) {
     const out = [];
-    for (const t of tokens) {
+    for (const [index, t] of tokens.entries()) {
+        if (names.has(index)) {
+            const inner = t.text.slice(1, t.unterminated ? undefined : -1).replaceAll("''", "'");
+            out.push({ kind: "word", value: inner, lower: inner.toLowerCase(), quoted: true });
+            continue;
+        }
         switch (t.kind) {
             case "comment":
                 break;
@@ -219,28 +241,25 @@ function selectModifiersFor(dialect) {
     }
 }
 /**
- * Lex and scan. With a known dialect the statement is lexed the way that engine reads it
- * (a Postgres-family one also the way CockroachDB ends a line comment, with the references
- * found there added). Without one (or for an unrecognized id) the generic reading decides scope
- * issues, and references are unioned across every engine's reading that lexes cleanly —
+ * Lex and scan, under the readings `sensitiveReadings` gives. With a known dialect the statement is
+ * lexed the way that engine reads it (a Postgres-family one also the way CockroachDB ends a line
+ * comment, with the references found there added). Without one (or for an unrecognized id) the
+ * generic reading decides scope issues, and references are unioned across every engine's reading
+ * that lexes cleanly —
  * so a column hidden from one reading by an engine-specific quote or comment rule (for
  * example MySQL `'\''` or Postgres `E'\''`) is still reported. Readings with an
  * unterminated token are skipped: the engine would reject that SQL outright.
  */
 function scanSql(sql, index, dialect) {
-    const profile = dialect ? lexerProfileFor(dialect) : undefined;
-    // A `postgres` dialect may point at CockroachDB, where only `\n` ends a line comment: the
-    // references that reading finds are added too.
-    const modifiers = profile ? selectModifiersFor(dialect) : SELECT_MODIFIERS.any;
-    const primary = scanTokens(lexSql(sql, profile ?? GENERIC_LEXER), index, modifiers);
-    const others = !profile
-        ? ENGINE_LEXER_PROFILES
-        : profile.id === "postgres" ? [{ ...profile, lineCommentEndsAtLfOnly: true }] : [];
-    for (const p of others) {
-        const tokens = lexSql(sql, p);
+    const read = readSql(sql);
+    const readings = sensitiveReadings(dialect);
+    const modifiers = dialect && isKnownEngine(dialect) ? selectModifiersFor(dialect) : SELECT_MODIFIERS.any;
+    const primary = scanTokens(read.tokens(readings.primary), index, modifiers, mayBeSqlite(readings.primary));
+    for (const p of readings.union) {
+        const tokens = read.tokens(p);
         if (hasUnterminatedToken(tokens))
             continue;
-        for (const [key, ref] of scanTokens(tokens, index, modifiers).found) {
+        for (const [key, ref] of scanTokens(tokens, index, modifiers, mayBeSqlite(p)).found) {
             const existing = primary.found.get(key);
             if (!existing)
                 primary.found.set(key, ref);
@@ -248,7 +267,12 @@ function scanSql(sql, index, dialect) {
                 existing.matchKind = "qualified";
         }
     }
-    return finalize(withPostgresOnlyIssues(primary, sql, dialect));
+    const outcome = withPostgresOnlyIssues(primary, sql, dialect);
+    // SQL Server's collation may read a name outside printable ASCII as a sensitive column's, which
+    // the scan above matches by spelling: unless the schema spells it exactly, fail closed.
+    if (unmatchedNonAsciiName(read, dialect, index.names) !== undefined)
+        outcome.issues.add("UNICODE_ESCAPED_IDENTIFIER");
+    return finalize(outcome);
 }
 /**
  * `outcome` with a scope issue for each form only the Postgres reading can hold: a `U&"…"` name,
@@ -278,8 +302,9 @@ function finalize(outcome) {
         ...(unresolvedScope ? { unresolvedScope } : {}),
     };
 }
-function scanTokens(lexed, index, modifiers) {
-    const tokens = toToks(lexed);
+function scanTokens(lexed, index, modifiers, sqlite) {
+    // A reading that may be SQLite's reads a '…' in a table's place as the table's name.
+    const tokens = toToks(lexed, sqlite ? stringsSqliteReadsAsNames(lexed, false) : new Set());
     const isWord = (t) => t?.kind === "word";
     const isPunct = (t, v) => t?.kind === "punct" && t.value === v;
     const isKeywordTok = (t, v) => t?.kind === "word" && !t.quoted && t.lower === v;
@@ -864,7 +889,7 @@ const SCOPE_ISSUE_TEXT = {
     UNKNOWN_QUALIFIER: "a qualified reference used a qualifier that is not a known table, alias, or CTE",
     OPAQUE_TABLE_SOURCE: "a table source is not a relation name (table function or table-valued expression)",
     UNTERMINATED_TOKEN: "a string, quoted identifier, or comment never closes, so the rest of the statement could not be scanned",
-    UNICODE_ESCAPED_IDENTIFIER: "a Unicode-escaped name (U&\"…\") can't be matched against sensitive columns, since Postgres decodes it",
+    UNICODE_ESCAPED_IDENTIFIER: "a name can't be matched against sensitive columns as written: a Unicode-escaped name (U&\"…\"), which Postgres decodes, or a name outside printable ASCII that the schema doesn't spell, which SQL Server's collation may read as another",
     STRING_CONTINUATION: "an E'…' string continues on a later line, which Postgres reads with backslash escapes, so where the string ends is unclear",
     ESCAPED_BYTE_STRING: "a b'…' byte string holds a backslash or continues on a later line, which CockroachDB reads with backslash escapes, so where the string ends is unclear",
 };