feat(grid): paste SQL INSERT statements into blank new rows

This commit is contained in:
yangzidemuou authored and GitHub committed 2026-09-30 15:28:07 +08:00
1 parent b32ef45da1
commit dc02e3bccf
18 files changed
+530 -9

No files matched your search

@@ -0,0 +1,99 @@
import { describe, expect, it } from "vitest";
import { parseInsertStatementPaste } from "@/lib/dataGrid/dataGridInsertPaste";
describe("parseInsertStatementPaste", () => {
it("returns null for non-INSERT text", () => {
expect(parseInsertStatementPaste("name\tage\nalice\t30")).toBeNull();
expect(parseInsertStatementPaste("SELECT * FROM users")).toBeNull();
expect(parseInsertStatementPaste("")).toBeNull();
});
it("parses a single statement with an explicit column list", () => {
const parsed = parseInsertStatementPaste("INSERT INTO users (name, age, bio) VALUES ('Alice', 30, NULL)");
expect(parsed).not.toBeNull();
expect(parsed?.columnNames).toEqual(["name", "age", "bio"]);
expect(parsed?.rows).toEqual([["Alice", "30", null]]);
});
it("parses a statement without a column list", () => {
const parsed = parseInsertStatementPaste("insert into users values (1, 'bob', NULL)");
expect(parsed?.columnNames).toBeNull();
expect(parsed?.rows).toEqual([["1", "bob", null]]);
});
it("parses multi-row VALUES tuples", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a, b) VALUES (1, 'x'), (2, 'y'), (3, 'z')");
expect(parsed?.rows).toEqual([
["1", "x"],
["2", "y"],
["3", "z"],
]);
});
it("unescapes doubled single quotes inside string literals", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a) VALUES ('it''s ok')");
expect(parsed?.rows).toEqual([["it's ok"]]);
});
it("unescapes MySQL backslash escapes inside string literals", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a, b, c, d) VALUES ('It\\'s', 'line1\\nline2', 'back\\\\slash', 'tab\\there')");
expect(parsed?.rows).toEqual([["It's", "line1\nline2", "back\\slash", "tab\there"]]);
});
it("keeps an escaped quote from merging two values", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a, b) VALUES ('It\\'s', 'ok')");
expect(parsed?.rows).toEqual([["It's", "ok"]]);
});
it("reads double-quoted strings as string values", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a) VALUES (\"plain 'text'\")");
expect(parsed?.rows).toEqual([["plain 'text'"]]);
});
it("strips typed-literal keywords and national prefixes from quoted values", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a, b, c) VALUES (TIMESTAMP '2024-01-01 10:00:00', DATE '2024-01-01', N'nebula')");
expect(parsed?.rows).toEqual([["2024-01-01 10:00:00", "2024-01-01", "nebula"]]);
});
it("supports multiple statements separated by semicolons", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a) VALUES (1); INSERT INTO t (a) VALUES (2);");
expect(parsed?.rows).toEqual([["1"], ["2"]]);
expect(parsed?.columnNames).toEqual(["a"]);
});
it("drops column alignment when statements disagree on column lists", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a) VALUES (1); INSERT INTO t (b) VALUES (2);");
expect(parsed?.rows).toEqual([["1"], ["2"]]);
expect(parsed?.columnNames).toBeNull();
});
it("keeps embedded commas, parens, and newlines inside string literals", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a, b) VALUES ('hello, world', 'line1\nline2 (x)')");
expect(parsed?.rows).toEqual([["hello, world", "line1\nline2 (x)"]]);
});
it("keeps function calls and numeric tokens verbatim for cell coercion", () => {
const parsed = parseInsertStatementPaste("INSERT INTO t (a, b, c) VALUES (NOW(), -12.5, 0xFF)");
expect(parsed?.rows).toEqual([["NOW()", "-12.5", "0xFF"]]);
});
it("skips SQL comments", () => {
const parsed = parseInsertStatementPaste("-- lead comment\nINSERT INTO t (a) /* block */ VALUES (1) -- trailing");
expect(parsed?.rows).toEqual([["1"]]);
});
it("handles quoted identifiers for table and column names", () => {
const parsed = parseInsertStatementPaste('INSERT INTO "my table" ("select", `order`) VALUES (1, 2)');
expect(parsed?.columnNames).toEqual(["select", "order"]);
expect(parsed?.rows).toEqual([["1", "2"]]);
});
it("returns null for INSERT ... SELECT", () => {
expect(parseInsertStatementPaste("INSERT INTO t (a) SELECT a FROM other")).toBeNull();
});
it("returns null when VALUES tuples are missing", () => {
expect(parseInsertStatementPaste("INSERT INTO t (a) VALUES")).toBeNull();
});
});
@@ -0,0 +1,285 @@
// Parser for pasting SQL INSERT statements into blank new rows (#10573).
// Accepts single or multiple statements and converts each VALUES tuple into a
// grid row. String literals keep their unescaped text (doubled quotes and
// MySQL backslash escapes), NULL becomes null, and every other token (numbers,
// functions, hex literals) is kept verbatim so the regular cell coercion
// handles it.
export interface ParsedInsertStatementPaste {
rows: Array<Array<string | null>>;
columnNames: string[] | null;
}
interface InsertStatementPart {
columnNames: string[] | null;
tuples: Array<Array<string | null>>;
}
type CharClass = "code" | "single-quote" | "double-quote" | "backtick" | "bracket-ident";
interface ScanState {
class: CharClass;
text: string;
}
function appendRaw(state: ScanState, char: string): void {
state.text += char;
}
function splitTopLevelStatements(text: string): string[] {
const statements: string[] = [];
const state: ScanState = { class: "code", text: "" };
let depth = 0;
for (let index = 0; index < text.length; index++) {
const char = text[index]!;
if (state.class === "code") {
if (char === "'" || char === '"' || char === "`") {
state.class = char === "'" ? "single-quote" : char === '"' ? "double-quote" : "backtick";
appendRaw(state, char);
continue;
}
if (char === "[" && depth === 0) {
state.class = "bracket-ident";
appendRaw(state, char);
continue;
}
if (char === "-" && text[index + 1] === "-") {
while (index < text.length && text[index] !== "\n") index++;
if (index < text.length) index--;
continue;
}
if (char === "/" && text[index + 1] === "*") {
index = text.indexOf("*/", index + 2);
index = index === -1 ? text.length : index + 1;
continue;
}
if (char === "(") depth++;
if (char === ")") depth = Math.max(0, depth - 1);
if (char === ";" && depth === 0) {
if (state.text.trim()) statements.push(state.text);
state.text = "";
continue;
}
appendRaw(state, char);
continue;
}
if (char === "\\" && (state.class === "single-quote" || state.class === "double-quote")) {
// MySQL string escapes: \' must not end the literal (backticks keep
// backslashes literal).
appendRaw(state, char);
if (index + 1 < text.length) appendRaw(state, text[++index]!);
continue;
}
appendRaw(state, char);
const closing = state.class === "single-quote" ? "'" : state.class === "double-quote" ? '"' : state.class === "backtick" ? "`" : "]";
if (char === closing) {
// Doubled quotes ('' / "" / ``) stay inside the literal; a doubled ]
// inside a bracket identifier is not standard, so only quotes double.
if (closing !== "]" && text[index + 1] === closing) {
appendRaw(state, text[++index]!);
continue;
}
state.class = "code";
}
}
if (state.text.trim()) statements.push(state.text);
return statements;
}
function normalizeIdentifier(raw: string): string {
const trimmed = raw.trim();
if (trimmed.length >= 2) {
const first = trimmed[0]!;
const last = trimmed[trimmed.length - 1]!;
if ((first === "'" && last === "'") || (first === '"' && last === '"') || (first === "`" && last === "`")) {
return trimmed.slice(1, -1).replaceAll(first + first, first);
}
if (first === "[" && last === "]") return trimmed.slice(1, -1).replaceAll("]]", "]");
}
return trimmed;
}
interface Lexer {
source: string;
pos: number;
}
function skipWhitespace(lexer: Lexer): void {
while (lexer.pos < lexer.source.length && /\s/.test(lexer.source[lexer.pos]!)) lexer.pos++;
}
function matchKeyword(lexer: Lexer, keyword: string): boolean {
skipWhitespace(lexer);
const slice = lexer.source.slice(lexer.pos, lexer.pos + keyword.length);
if (slice.toUpperCase() !== keyword) return false;
const next = lexer.source[lexer.pos + keyword.length];
if (next && !/[\s(]/.test(next)) return false;
lexer.pos += keyword.length;
return true;
}
function readBracketedGroup(lexer: Lexer): string[] | null {
skipWhitespace(lexer);
if (lexer.source[lexer.pos] !== "(") return null;
lexer.pos++;
const items: string[] = [];
let current = "";
let depth = 0;
let classState: CharClass = "code";
while (lexer.pos < lexer.source.length) {
const char = lexer.source[lexer.pos++]!;
if (classState === "code") {
if (char === "'" || char === '"' || char === "`") {
classState = char === "'" ? "single-quote" : char === '"' ? "double-quote" : "backtick";
} else if (char === "[" && depth === 0) {
classState = "bracket-ident";
} else if (char === "(") {
depth++;
} else if (char === ")") {
if (depth === 0) {
items.push(current);
return items;
}
depth--;
} else if (char === "," && depth === 0) {
items.push(current);
current = "";
continue;
}
} else if (char === "\\" && (classState === "single-quote" || classState === "double-quote")) {
current += char;
if (lexer.pos < lexer.source.length) current += lexer.source[lexer.pos++]!;
continue;
} else if (char === (classState === "bracket-ident" ? "]" : classState === "single-quote" ? "'" : classState === "double-quote" ? '"' : "`")) {
classState = "code";
}
current += char;
}
return null;
}
function readValueToken(lexer: Lexer): string | null {
skipWhitespace(lexer);
if (lexer.pos >= lexer.source.length) return null;
const start = lexer.pos;
const char = lexer.source[lexer.pos]!;
if (char === "'" || char === '"' || char === "`") {
const closing = char;
lexer.pos++;
while (lexer.pos < lexer.source.length) {
const inner = lexer.source[lexer.pos++]!;
if (inner === closing) {
if (lexer.source[lexer.pos] === closing) lexer.pos++;
else break;
} else if (inner === "\\" && closing !== "`") {
if (lexer.pos < lexer.source.length) lexer.pos++;
}
}
return lexer.source.slice(start, lexer.pos);
}
while (lexer.pos < lexer.source.length) {
const inner = lexer.source[lexer.pos]!;
// A table reference never contains commas, brackets, or whitespace, so the
// first one of those ends the token (and leaves "(" for the column list).
if (inner === "," || inner === "(" || inner === ")" || /\s/.test(inner)) break;
lexer.pos++;
}
return lexer.source.slice(start, lexer.pos).trim() || null;
}
function unescapeSqlString(text: string): string {
return text.replace(/\\(.)/gs, (_match, char: string) => {
switch (char) {
case "n":
return "\n";
case "r":
return "\r";
case "t":
return "\t";
case "0":
return "\0";
default:
// \\ \' \" and unrecognized escapes keep the escaped character itself,
// matching MySQL string-literal semantics.
return char;
}
});
}
function convertValue(raw: string | null): string | null {
if (raw === null) return null;
let trimmed = raw.trim();
if (trimmed.toUpperCase() === "NULL") return null;
// Typed literals keep only the quoted payload: TIMESTAMP '...', DATE '...',
// or a national-character prefix N'...'.
const typed = trimmed.match(/^(?:N|DATE|TIME|TIMESTAMP|DATETIME)(?=\s*['"])/i);
if (typed) trimmed = trimmed.slice(typed[0].length).trimStart();
if (trimmed.length >= 2 && trimmed[0] === "'" && trimmed[trimmed.length - 1] === "'") {
return unescapeSqlString(trimmed.slice(1, -1).replaceAll("''", "'"));
}
if (trimmed.length >= 2 && trimmed[0] === '"' && trimmed[trimmed.length - 1] === '"') {
return unescapeSqlString(trimmed.slice(1, -1).replaceAll('""', '"'));
}
return trimmed;
}
function parseSingleInsertStatement(statement: string): InsertStatementPart | null {
const lexer: Lexer = { source: statement, pos: 0 };
if (!matchKeyword(lexer, "INSERT")) return null;
if (!matchKeyword(lexer, "INTO")) return null;
// Skip the table reference (possibly db.schema."table" or `db`.`table`).
if (!readValueToken(lexer)) return null;
// Optional column-name list: the next bracketed group before VALUES/VALUE.
const save = lexer.pos;
let columnNames: string[] | null = null;
const group = readBracketedGroup(lexer);
if (group) {
columnNames = group.map(normalizeIdentifier);
} else {
lexer.pos = save;
}
if (!matchKeyword(lexer, "VALUES") && !matchKeyword(lexer, "VALUE")) return null;
const tuples: Array<Array<string | null>> = [];
for (;;) {
skipWhitespace(lexer);
if (lexer.source[lexer.pos] !== "(") break;
const rawValues = readBracketedGroup(lexer);
if (rawValues === null) return null;
// Values inside a tuple: split on top-level commas again (readBracketedGroup
// already did that) and convert each token.
tuples.push(rawValues.map(convertValue));
skipWhitespace(lexer);
if (lexer.source[lexer.pos] === ",") {
lexer.pos++;
continue;
}
break;
}
if (tuples.length === 0) return null;
return { columnNames, tuples };
}
export function parseInsertStatementPaste(text: string): ParsedInsertStatementPaste | null {
const statements = splitTopLevelStatements(text);
// The first statement decides the paste's semantics; leading comments were
// already stripped by the statement splitter.
if (statements.length === 0 || !/^\s*insert\s+into\s/i.test(statements[0]!)) return null;
const rows: Array<Array<string | null>> = [];
let columnNames: string[] | null = null;
let sawColumnNames = false;
for (const statement of statements) {
const parsed = parseSingleInsertStatement(statement);
if (!parsed) return null;
if (parsed.columnNames) {
if (!sawColumnNames) {
columnNames = parsed.columnNames;
sawColumnNames = true;
} else if (columnNames && parsed.columnNames.join("\u0000") !== columnNames.join("\u0000")) {
columnNames = null;
}
}
rows.push(...parsed.tuples);
}
if (rows.length === 0) return null;
return { rows, columnNames };
}