Module matched 

The DataFusion and sqlparser messages the engine and the CLI recognize by their text, because no typed error says the same (#265).

Each is matched as a substring or a prefix, as its doc says. The tests below trigger each message in DataFusion or sqlparser directly, not through Ursa’s wrappers, and check the text is still there: when an upgrade changes the wording, they fail and name the string to update.

Constants

CSV_FILE_CONTEXT

The start of the context DataFusion puts on an error inferring a CSV file’s schema, followed by the file’s object store location, which has lost its leading / (Users/me/x.csv): super::user_message puts it back (#327, N14).

OPTIMIZER_RULE_FAILED

The start of the context DataFusion’s optimizer puts on a rule’s error (Optimizer rule '…' failed). Which rule failed is DataFusion’s business: super::user_message drops it and keeps the cause.

TIMESTAMP_NS

The nanosecond timestamp type, as arrow names it in a cast error.

TIMESTAMP_OVERFLOW

In arrow’s cast error for a date or timestamp outside the range of a nanosecond timestamp (Cannot cast Date32 value -100000000 to Timestamp(ns): converted value exceeds the representable i64 range), which date_trunc and other timestamp functions meet on an out-of-range date. Matched together with TIMESTAMP_NS.

TOKENIZER_EOF

In sqlparser’s tokenizer error for a quoted identifier or a block comment still open at the end of the input (… before EOF., Unexpected EOF while in a multi-line comment). The shell reads the next line instead (ursa-cli’s input.rs).

TOKENIZER_UNTERMINATED

The start of sqlparser’s tokenizer error for a string literal (or a dollar-quoted or encoded one) still open at the end of the input. The shell reads the next line instead of reporting it (ursa-cli’s input.rs).