Fix identifier parsing not breaking on the |> pipe operator

[alexander-beedie]

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

• Updated the pipe operator tests with some new coverage for the above, and implemented a test_pipe! helper macro that additionally validates pipe queries starting "FROM <tbl>" (in addition to "SELECT * FROM <tbl>").

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

[alexander-beedie]

Rebased / addressed test format👌

[iffyio]

can we drop the test_pipe macro entirely and write the test cases explicitly like dialects.verified_stmt("SELECT * FROM users |> WHERE id = 1");. And likely we should drop the pipe_dialects() method and construct the dialects explicitly too. For similar reason mentioned previously in that its not obvious without reading a bit of custom code to figure out what the tests are doing and I don't think its worth it in this case.