表格转 JSON / SQL 转换器 - 在线处理 CSV、TSV、Markdown

将 CSV 粘贴（或拖入 .csv 文件）到输入面板，工具会自动检测分隔符，以表头行作为键将每一行解析为一个 JSON 对象，并即时显示结果。将输出格式切换为 JSON 数组、JSONL 或 SQL INSERT，即可获得你所需的确切结构。

可以。解析器遵循 RFC 4180——用双引号包裹的字段可以包含逗号、换行符甚至字面引号（转义为 ""）而不会破坏解析。因此像 "Smith, John" 这样的值仍然保持为单个单元格，而非两个。

该工具会统计第一行非空内容中的制表符、逗号、分号和竖线，并选取出现最多的那个。如果您的文件较为特殊或含义不明，可以覆盖自动检测，手动选择制表符（TSV）、逗号、分号、竖线或 Markdown 表格。

将输出格式切换为 SQL INSERT 并输入表名。工具会把列名清理为合法标识符（将空格、连字符等替换为下划线），并为每一行生成一条 INSERT 语句，其中字符串字面量已正确转义、数值保持原样、空单元格则填入 NULL。非常适合初始化数据库或迁移数据。

JSONL（JSON Lines）每行放置一个 JSON 对象，外层不带数组。它是将数据流式传输到 BigQuery、Snowflake、Logstash、Elasticsearch 以及大多数机器学习管道的标准格式。当每一行将被独立摄取时，请选择 JSONL。

可以。符合整数或小数格式的值会输出为 JSON 数字，字面量「true」/「false」会变为布尔值，空单元格则变为 null。如果你需要将所有内容都作为字符串处理，请在转换前相应地准备好数值。

可以。选择「Markdown 表格」作为分隔符，然后粘贴标准的以竖线分隔的 Markdown 表格（表头行、分隔行、正文行）。分隔行会被自动跳过，单元格中的空白会被去除。

不会——解析和转换完全在你的浏览器中运行。这让该工具可安全用于敏感的电子表格，如工资单导出、客户名单或绝不应接触第三方服务器的财务记录。

When checked, the tool treats your table's first row as column names and uses them as the JSON object keys (or generates col_1, col_2... for SQL if you leave it blank). Uncheck it if your data has no header row — a plain list of rows — and the tool will instead output raw JSON arrays or auto-name columns col_1, col_2, etc. for SQL INSERT statements.

Switch Output to SQL INSERT and a Table name field appears where you can type your target table's name, such as customers or order_items. The tool automatically sanitizes whatever you enter — replacing spaces and symbols with underscores and prefixing a leading digit — so the generated SQL always has a valid identifier even if your input is messy.

Uncheck 'First row is header' if your export has no column names, set Output to SQL INSERT, and type your real table name in the Table name field. The tool will auto-generate col_1, col_2... column names, coerce numbers and booleans automatically, and produce one properly-escaped INSERT statement per row that you can paste straight into your database client.

JSON (arrays) drops the field names and outputs each row as a plain list of values in column order — for example ["Alice", 34, "Berlin"] instead of {"name":"Alice","age":34,"city":"Berlin"}. It still respects the 'First row is header' checkbox, using that row only to know how many columns to expect rather than as keys. Choose it when the receiving code (a charting library, a positional database driver, or a spreadsheet import) reads values by position rather than by field name; otherwise stick with JSON (objects).

Try sample fills the input panel with a short three-row name/age/city table (tab-separated) so you can immediately see how the delimiter detection, header row and each output format behave before you paste real data. Once there's content in the box, a Clear button appears — next to the Input label in the editor view, and again as a full-width button in the Settings panel — that wipes the textarea back to empty in one click instead of selecting and deleting the text manually.