# QUERY_FAILED

> DuckDB could not bind or run a query

**Error** · category **project** (the project has a mistake in an asset file, croft.json or the command line) · exit **2** (invalid project or usage)

DuckDB could not bind or run a query.

The SQL parsed and passed croft's checks, but DuckDB failed while binding or running it: a type that does not
convert (Conversion Error), a function called with the wrong arguments (Binder Error), a value out of range,
text that is not valid JSON read as JSON, and the like. details.duckdb is DuckDB's
message and details.duckdbErrorType its kind. A missing table or column is UNKNOWN_TABLE or UNKNOWN_COLUMN, and
SQL that does not parse is SQL_SYNTAX.

It comes from croft query, from the read server and @zabaca/croft/read, and from ctx.query in asset code, where
it also covers a query run before the asset's table exists (its first load) or after its step ended. A preview that
built an asset but could not compare it with the live table reports it as a warning.

Nothing was written.

What to do: fix the SQL; DuckDB's message says what went wrong. Look at the columns and their types first:
croft describe <asset> shows them, and croft query "DESCRIBE <asset>" too. Cast explicitly (TRY_CAST(x AS BIGINT)
gives NULL instead of failing). In asset code, handle the first run, when there is nothing saved to read yet, and
await every ctx.query inside rows().

The same page is in the CLI: `croft docs QUERY_FAILED`.

Source: https://croft.zabaca.com/docs/reference/errors/query-failed/