There's a bug in Pipe SQL parsing with identifiers followed by the |> operator.

Fix

parse_identifiers needs to break on Token::VerticalBarRightAngleBracket as well, otherwise it continues to consume the following tokens as Ident (until one of the other break conditions is encountered).

Example

FROM t |> DROP c |> RENAME a AS x

Without the fix the DROP section consumes the RENAME tokens as identifiers:

Drop { columns: [
    Ident { value: "c" },
    Ident { value: "RENAME" },
    Ident { value: "a" },
    Ident { value: "AS" },
    Ident { value: "x" }]
}

After the fix the identifier boundary is respected, and the operations parse correctly:

Drop { columns: [Ident { value: "c" }] }
Rename { mappings: [
    IdentWithAlias { ident: Ident { value: "a" }, alias: Ident { value: "x" } }]
}

Also

• Split the monolithic test into several more manageable tests for clarity
  (one per pipe-operator).