sqlpane

Drop in a file. Query it with SQL. Nothing leaves your browser.

sqlpane is an offline-first SQL workbench that runs entirely in the browser. Drag in CSV, JSON, Excel, Parquet or SQLite files, or a MySQL or SQL Server script export, and query them with SQL. It is built on DuckDB-WASM and works without a server or an account.

線上試用 / Try it online: https://zzoraemon043.github.io/sqlpane/

繁體中文 | English

拖進檔案，用 SQL 查詢，資料不離開瀏覽器。

sqlpane 是完全在瀏覽器中執行、離線優先的 SQL 工作台。把 CSV、JSON、Excel、Parquet、SQLite 檔案，或 MySQL、SQL Server 匯出的 .sql 指令碼拖進來，就能直接用 SQL 查詢。底層使用 DuckDB-WASM，不需要伺服器，也不需要帳號。

想看一個資料檔的內容，通常得先安裝資料庫、寫匯入腳本，或把檔案上傳到雲端服務。sqlpane 在瀏覽器分頁中直接打開檔案、轉成資料表，馬上就能查詢，資料始終留在你的電腦上。

• 拖放或按「開啟檔案」開啟 CSV、TSV、JSON、NDJSON、Parquet、Excel（.xlsx）與 SQLite 資料庫檔
• 匯入 MySQL（mysqldump）與 SQL Server（產生指令碼）匯出的 .sql 檔，並產生匯入報告
• SQL 編輯器：語法上色、依目前結構自動補全資料表與欄位、格式化、執行游標所在或選取的陳述式
• 查詢紀錄與已存查詢保存在瀏覽器的 IndexedDB
• 結構側欄：資料表、欄位、型別與前 100 筆預覽
• 虛擬捲動的結果表格，百萬筆資料也流暢，可排序、篩選、調整欄寬，並顯示筆數與查詢時間
• 長條圖、折線圖、散佈圖、圓餅圖
• 結果匯出為 CSV、JSON、Parquet 與 Excel
• 可安裝為 PWA，第一次載入後完全離線使用
• 繁體中文與英文介面，預設依瀏覽器語言

把檔案拖到頁面任何位置，或按右上角「開啟檔案」，可以一次選多個檔案。每個檔案會變成資料表：

名稱已存在時會自動加上 _2、_3。在左側「資料表」展開資料表，可以「預覽」、「重新命名」或「移除」。

CSV 與 TSV 會自動偵測分隔符號、標題列與欄位型別。偵測不正確時，按「CSV 選項」指定分隔符號或是否有標題列，再重新開啟檔案。

• MySQL：mysqldump --databases 資料庫名稱 > dump.sql
• SQL Server：在 SQL Server Management Studio 對資料庫按右鍵，選「工作 > 產生指令碼」，在「進階」把「要編寫指令碼的資料類型」設為「結構描述和資料」。UTF-16（Unicode）與 UTF-8 檔案都能讀取。

匯入器會處理 CREATE TABLE（欄位、型別、可否為 NULL、主索引鍵）、單列與多列 INSERT、反引號與方括號識別字、N'...' 字串、十六進位與位元常值、NULL、CAST(... AS ...)，以及 GO 批次分隔。型別會對應到 DuckDB，例如 TINYINT(1) 對應 BOOLEAN、NVARCHAR(MAX) 對應 VARCHAR、UNIQUEIDENTIFIER 對應 UUID、MONEY 對應 DECIMAL(19,4)。

SET、USE、LOCK TABLES、DROP TABLE、ALTER TABLE、GO、註解這類不影響資料的陳述式會略過。檔案以串流方式讀取，匯入時會顯示進度；完成後的匯入報告會列出略過的陳述式、無法匯入的陳述式，以及失敗資料列的行號與原因。

SQL Server 的 .bak 備份檔與 MySQL 的資料目錄無法直接開啟，請依上面的方式改匯出 .sql 指令碼。

每次執行都會記在「查詢紀錄」，點一下即可帶回編輯器。瀏覽器不允許使用 IndexedDB 時（例如部分無痕模式），程式仍可正常使用，只是紀錄在關閉頁面後就會消失。

• 點欄位名稱依該欄排序（遞增、遞減、取消）；欄位名稱下方的輸入框可篩選，輸入 null 只顯示空值
• 拖曳欄位名稱右側邊界可調整欄寬
• 「圖表」分頁可選擇圖表類型、類別或 X 軸欄位與數值欄位
• 「匯出」會匯出目前篩選與排序後的資料列，Excel 最多 1,048,575 列

第一次開啟網站後，右上角顯示「已可離線使用」時，網站與 SQL 引擎都已存在瀏覽器中，之後即使沒有網路也能開啟與查詢。在 Chrome 或 Edge 中可從網址列的安裝按鈕把 sqlpane 安裝成應用程式。

• 所有檔案都在你的瀏覽器中讀取與查詢，不會上傳
• DuckDB 的 WebAssembly 檔、SQLite 讀取程式與其他程式碼都由網站本身提供，執行時不會從 CDN 載入任何東西
• 網站設定了 Content Security Policy，只允許連線到網站本身；DuckDB 自動下載擴充套件的功能已關閉

• 開啟的檔案會載入到瀏覽器記憶體，實際可處理的大小取決於裝置記憶體與瀏覽器限制
• Excel 只支援 .xlsx，不支援舊版 .xls
• 無法直接連線到 MySQL 或 SQL Server 伺服器
• 查詢中無法直接用 read_parquet、read_json 讀取檔案，請先開啟檔案再查詢資料表

需要 Node.js 22。

npm ci
npm run dev          # 開發伺服器
npm test             # 單元測試
npm run lint         # 程式碼檢查
npm run build        # 建置到 dist/
npm run test:e2e     # 端對端測試（需先建置）
npm run screenshots  # 重新產生 README 截圖

測試資料都是手寫或由測試程式產生，不含真實個人資料。推送到 main 後，GitHub Actions 會建置並部署到 GitHub Pages。使用的第三方套件與授權列在 THIRD_PARTY_NOTICES.md。

MIT

Drop in a file. Query it with SQL. Nothing leaves your browser.

Looking inside a data file usually means installing a database, writing an import script or uploading the file to a cloud service. sqlpane opens the file in a browser tab, turns it into a table and lets you query it right away. Your data stays on your machine.

• Open CSV, TSV, JSON, NDJSON, Parquet, Excel (.xlsx) and SQLite database files by drag and drop or with Open files
• Import MySQL (mysqldump) and SQL Server (Generate Scripts) .sql exports, with an import report
• SQL editor with syntax highlighting, table and column completion from the current schema, formatting, and running the statement under the cursor or the selection
• Query history and saved queries stored in the browser's IndexedDB
• Schema sidebar with tables, columns, types and a preview of the first 100 rows
• Virtualized result grid that stays fast with millions of rows, with sorting, filtering, column resizing, row count and query time
• Bar, line, scatter and pie charts
• Export results to CSV, JSON, Parquet and Excel
• Installable as a PWA and fully usable offline after the first visit
• Interface in Traditional Chinese and English, following the browser language by default

Drop files anywhere on the page or use Open files in the top right corner. You can open several files at once. Each file becomes tables:

When a name is taken, _2, _3 and so on is added. Expand a table under Tables to preview, rename or remove it.

CSV and TSV files are read with automatic detection of the delimiter, header row and column types. If the detection is wrong, set the delimiter or header row under CSV options and open the file again.

• MySQL: mysqldump --databases your_database > dump.sql
• SQL Server: in SQL Server Management Studio, right-click the database, choose Tasks > Generate Scripts, and under Advanced set "Types of data to script" to "Schema and data". Both UTF-16 (Unicode) and UTF-8 files work.

The importer handles CREATE TABLE (columns, types, nullability and primary keys), single-row and multi-row INSERT, backtick and square bracket identifiers, N'...' strings, hex and bit literals, NULL, CAST(... AS ...) and GO batch separators. Types are mapped to DuckDB, for example TINYINT(1) to BOOLEAN, NVARCHAR(MAX) to VARCHAR, UNIQUEIDENTIFIER to UUID and MONEY to DECIMAL(19,4).

Statements that do not affect data, such as SET, USE, LOCK TABLES, DROP TABLE, ALTER TABLE, GO and comments, are skipped. The file is read as a stream and the import shows its progress. The import report lists skipped statements, statements that could not be imported, and the line number and reason for every failed row.

SQL Server .bak backups and MySQL data directories cannot be opened. Export a .sql script as described above instead.

Every run is added to History, and clicking an entry puts it back in the editor. When the browser does not allow IndexedDB, for example in some private windows, the app keeps working and the history lasts until the page is closed.

• Click a column name to sort by it (ascending, descending, off). Type in the box under a column name to filter; type null to show only empty values
• Drag the right edge of a column header to resize the column
• The Chart tab lets you pick the chart type, the category or X axis column and the value column
• Export saves the rows currently shown after filtering and sorting. Excel holds at most 1,048,575 rows

After the first visit, once the header shows "Ready for offline use", the site and the SQL engine are stored in the browser and work without a network connection. In Chrome or Edge you can install sqlpane as an app from the install button in the address bar.

• Files are read and queried in your browser and are never uploaded
• The DuckDB WebAssembly files, the SQLite reader and all other code are served by the site itself. Nothing is loaded from a CDN at runtime
• The site sets a Content Security Policy that only allows connections to the site itself, and DuckDB's automatic extension download is turned off

• Opened files are loaded into browser memory, so the practical size limit depends on the device and the browser
• Excel support covers .xlsx, not the older .xls
• Connecting to a live MySQL or SQL Server is not possible
• read_parquet and read_json cannot be used inside queries; open the file first and query its table

Requires Node.js 22.

npm ci
npm run dev          # development server
npm test             # unit tests
npm run lint         # lint
npm run build        # build into dist/
npm run test:e2e     # end-to-end tests (build first)
npm run screenshots  # regenerate the README screenshots

Test fixtures are hand-written or generated by the tests and contain no real personal data. Pushing to main builds the site and deploys it to GitHub Pages with GitHub Actions. Third-party packages and their licenses are listed in THIRD_PARTY_NOTICES.md.

MIT