CSV bulk loader

The Data card on a database’s Overview tab has a Load CSV button. Pick a .csv file, choose a target table, and PerSQL infers column types from the first 200 rows and inserts the whole file inside one transaction.

Modes

• Create (default) — creates the table if it doesn’t exist; fails if it does.
• Append — inserts into an existing table. Column count and order must match the CSV.
• Replace — drops the table (if any) and re-creates it from the CSV’s inferred schema. Destructive.

Type inference

Each column’s type is picked by walking the first 200 data rows:

All non-empty values are…	Type
Integer literals	INTEGER
Numeric (any finite number)	REAL
Anything else	TEXT

Empty cells become NULL. If a row outside the sample window has a value that doesn’t fit (e.g. "oops" in an INTEGER column), the INSERT for that row fails and the whole transaction rolls back.

Headers

By default, the first row is treated as a header and used as the column names. Names get sanitized — non-alphanumeric chars become underscores, duplicates get a suffix, empty cells fall back to col_<n>. Uncheck First row is a header in the dialog if your CSV starts with data.

Limits

• 10 MB file size, 100 000 rows.
• Multipart upload only (multipart/form-data).
• Manage permission required.

REST

curl --cookie cookies.txt \
  -F "file=@orders.csv" \
  -F "table=orders" \
  -F "mode=create" \
  -F "header=true" \
  https://api.persql.com/api/namespaces/acme/databases/orders/load-csv

Response:

{
  "success": true,
  "data": {
    "table": "orders",
    "rowsInserted": 12500,
    "columns": [
      { "name": "id", "type": "INTEGER" },
      { "name": "customer_id", "type": "INTEGER" },
      { "name": "total", "type": "REAL" },
      { "name": "status", "type": "TEXT" }
    ]
  }
}

Tips

• For very large datasets, split into multiple files of <100 k rows and use mode=append for the rest.
• Quoting follows RFC-4180 — "a,b" is one field, "" inside a quoted field is an escaped ".
• BOM at the start of the file is stripped automatically.