Skip to content

import-export.md

Latest commit

 

History

History
29 lines (21 loc) · 2.76 KB

import-export.md

File metadata and controls

29 lines (21 loc) · 2.76 KB
Raw

📥 Import & Export

Load a CSV (or any delimited file) or run a .sql script, from the ⬆ button in the Schema Explorer or a schema's or table's right-click menu.

CSV

  • Delimiter auto-detected (, ; tab |) - the one that splits every line the same way wins, so commas inside quoted values don't fool it; override it if you'd rather
  • Column mapping - each source column shows its first value and the target column it loads into, with a warning when its values may not fit that column's type; blank the target to skip that column entirely
  • Import into an existing table, optionally emptying it first, or create a new one with types inferred from the data (integer, decimal, boolean, date, timestamp, text) - each one editable before you commit to it, and named in your engine's own dialect
  • Headers optional, NULL text configurable (\N, NULL, whatever your exporter writes), leading lines skippable for files with a preamble, whitespace trimmable
  • Empty fields keep their meaning: a bare one is NULL, a quoted "" is an empty string - the same convention Postgres COPY … FORMAT csv uses, and what lets a BarSQL CSV export load straight back in. In a numeric or date column, where "" is never valid, both mean NULL
  • UTF-8 BOMs stripped, ragged rows tolerated, and the preview warns when a file isn't UTF-8 (like Excel's Windows-1252 CSVs), whose unreadable characters would load as �
  • Bad rows are skipped and reported, not fatal - or tick stop at the first error if you'd rather nothing partial lands
  • Rows load in batches with live progress, and Stop ends a long run

SQL scripts

A .sql file runs statement by statement on one connection, so SET, temp tables and scripted transactions behave. You get a progress bar and a count - not thousands of result tabs. It's also how a table backup comes back. The file must be UTF-8 (a BOM is fine); any other encoding is refused rather than loaded with � in place of what it can't read.

Batches are committed as they go, so a stopped or failed import keeps the rows it had already loaded and tells you how many that was. Read-only connections refuse imports outright.

📤 Export

  • Text / CSV / JSON / Markdown / SQL INSERT - SQL is written for the connection's own engine: its identifier quoting, N'' strings and 1/0 booleans on SQL Server, and literals that read the same whatever the server's string settings
  • Export all or selected rows and columns
  • Copy to clipboard or save to file
  • Remembers your last export format
  • CSV and SQL exports import back cleanly - NULL stays apart from an empty string, and cells that spreadsheets would treat as formulas are guarded on the way out and unguarded on the way in