Skip to content
QuerybaraQuerybara

Import

  • PostgreSQL
  • MySQL
  • MariaDB

The import wizard loads a file into a table. It detects how the file is written, shows a live preview, matches the file’s columns to the table’s, and runs the load as a job in its own process, so you can keep working while it runs.

Import wizard for shop.products with a supplier price list CSV: format, encoding, delimiter and quote detected, the header row recognised, and a live preview of the rows with an inferred type for each column.Import wizard for shop.products with a supplier price list CSV: format, encoding, delimiter and quote detected, the header row recognised, and a live preview of the rows with an inferred type for each column.
Import detects the format, delimiter and types, and previews the file before anything is written.

Formats

Section titled “Formats”
Format Notes
CSV, TSV Delimiter, quote, header and encoding detected
JSON An array of objects
JSON Lines One object per line
Excel .xlsx workbooks, read as a stream; pick the worksheet and the header row
XML Rows are the elements at a path, detected from a sample or chosen; attributes and child elements become columns
Parquet Typed columns from the file’s schema; lists, maps and structs become JSON text

Text formats can be gzip-compressed. Legacy .xls and password-protected workbooks are refused.

Parquet

Section titled “Parquet”

A Parquet file is recognised by its .parquet, .parq or .pq extension, or by its contents. Its columns keep their types: decimals stay exact, dates keep BC and five-digit years, times and timestamps keep every microsecond or nanosecond, and UUIDs come in as their text. Binary values come in as \x hex. Nested columns (lists, maps, structs) come in as JSON text, one value per row.

Querybara reads Snappy, GZIP, ZSTD, Brotli and LZ4 pages, and uncompressed ones. Files compressed with LZO, and encrypted files, are refused with a message saying so. The file is read one row group at a time, so a large file does not need to fit in memory.

Import into a table

Section titled “Import into a table”

  1. In the sidebar or the Objects tab, right-click a table and choose Import data…. To create a table, right-click a schema (a database on MySQL and MariaDB) and choose Import into new table….

  2. File: choose Choose file… and pick the file.

  3. Preview: check the detected settings and the sample rows, with each column’s inferred type under its name. Change what was detected if needed:

    • Format, and Encoding (shown as Detected: …)
    • CSV and TSV: Delimiter, Quote, NULL is written as and First row holds the column names
    • Excel: Worksheet and Header row (0: none)
    • XML: Rows are the elements at, or Or type a path
    • Parquet: Parquet file shows the row count, the row groups, the codecs and the program that wrote the file. There is no encoding to choose.

    The preview updates as you change them.

  4. Columns: file columns are matched to table columns by name; case, spaces, _ and - do not count. Pick a different Table column for a file column, or — skip — it. Unmapped table columns get their defaults.

  5. Options: pick the mode and the batch and error settings (see below).

  6. Review: check the file, format, target, column pairs, mode and options, then choose Import.

Import pipelineAn import detects the file's format, delimiter, header and encoding, previews it, matches columns, applies the chosen mode, and streams rows into the table in the job runner process, reporting failed rows by row, line and column.products.xlsxworksheet 2026Detectformat · header · encodingPreviewsku · name · price · qtyLW-TAB-OAK-1004 · Oak dining table · 889.99 · 1LW-SEA-ASH-1051 · Ash bench · 279.99 · 2LW-STO-ELM-1077 · Elm stool · 154.99 · 4↳ sku → sku · name → name · price → price · qty → quantityappendupdateupsertdeletereplaceJob runnerits own processshop.productsPostgreSQLstreaming · 82%Row 418 failedline 419 · column pricebytesrowsrowrowrow 418

Import pipeline

  1. You pick a file: CSV, TSV, JSON, JSON Lines (gzipped too), Excel or XML.
  2. Querybara detects the format, delimiter, header and encoding, and shows a live preview.
  3. Columns are matched to the table automatically; for a new table, types are inferred.
  4. You choose a mode: append, update, upsert, delete or replace.
  5. The job runner process streams the rows in, with progress, while you keep working.
  6. A row that fails is reported by row, line and column. Cancel rolls the import back.

Import into a new table

Section titled “Import into a new table”

When you start from Import into new table…, the Columns step plans the table instead. Querybara proposes a Table name from the file and a column name and type for each file column, from the types it inferred. For a Parquet file, the types come from the file’s schema rather than from the sample values. For each column you can untick Import, rename it, change its Type, and set Nullable and Primary key. The CREATE TABLE statement is shown below the table.

Modes

Section titled “Modes”

The mode applies when importing into an existing table:

Mode What it does
Append Inserts every row
Update Updates rows that match on the key columns
Upsert Inserts new rows and updates the ones that match
Delete matching Deletes rows that match on the keys
Replace Empties the table, then inserts every row

For Update, Upsert and Delete matching, tick the Key columns (rows match on these).

Options

Section titled “Options”
Option Choices
Rows per batch How many rows each statement batch holds
Transaction One for the whole file or One per batch
When a row fails Stop and roll back, or Skip it and continue
Turn off foreign key checks during the load Existing tables. On PostgreSQL it also skips triggers and needs a superuser

With Skip it and continue, each failed row is listed in the Jobs panel with its row, line and column. Cancel in the Jobs panel stops the import and rolls back its transaction.

Saved settings keeps the wizard’s choices under a name (Save settings…) for the next import; file paths are never saved.

From the command line

Section titled “From the command line”

querybara import takes the same formats, modes and options. It exits 0 when imported, 1 when rows were skipped, 2 on failure and 130 when interrupted.

Terminal window
querybara import dev --table shop.customers --file customers.csv
querybara import dev --table shop.products --file products.json.gz --mode upsert --key id
querybara import dev --table shop.orders --file q3.xlsx --sheet July --header-row 3
querybara import dev --table shop.orders --file orders.xml --row-path /export/table/row
querybara import dev --table shop.order_items --file items.tsv --create --on-error skip
querybara import dev --table shop.orders_archive --file orders.parquet --create
Section titled “Related”

Documents Querybara 0.1.1 · built frombc9f5aa