INSERT INTO FILES fail to parse row/column delimiter like '0x11'

[wxl24life]

Steps to reproduce the behavior (Required)

create database unload;
USE unload;
-- SELECT * FROM sales_records;
CREATE TABLE sales_records(
    record_id     BIGINT,
    seller        STRING,
    store_id      INT,
    sales_time    DATETIME,
    sales_amt     DOUBLE
)
DUPLICATE KEY(record_id)
PARTITION BY date_trunc('day', sales_time)
DISTRIBUTED BY HASH(record_id);

INSERT INTO sales_records
VALUES
    (220313001,"Amy",1,"2022-03-13 12:00:00",8573.25),
    (220314002,"Bob",2,"2022-03-14 12:00:00",6948.99),
    (220314003,"Amy",1,"2022-03-14 12:00:00",4319.01),
    (220315004,"Carl",3,"2022-03-15 12:00:00",8734.26),
    (220316005,"Carl",3,"2022-03-16 12:00:00",4212.69),
    (220317006,"Bob",2,"2022-03-17 12:00:00",9515.88);


INSERT INTO FILES(
    "path" = "oss://cdp-hangzhou/unload/test_03/",
    "format" = "CSV",
    'csv.column_separator' = '0X11'
    -- 'csv.row_delimiter' = '\\X12'
) 
SELECT * FROM unload.sales_records;

Expected behavior (Required)

Real behavior (Required)

StarRocks version (Required)

• main branch

[kevincai]

what's the \X11? is it a single non-printable char or a str "\X11"?

[wxl24life]

what's the \X11? is it a single non-printable char or a str "\X11"?

@kevincai It's hexadecimal representations of some special ASCII control character. The character is supported in other features like load and export. We should try best to make insert into files to align with the behavior there

[kevincai]

@EsoragotoSpirit please take a look, if the relative doc needs an update of this row_delimiter/col_delimiter escape capability?

[EsoragotoSpirit]

And the minor version this will be supported?

[kevincai]

And the minor version this will be supported?

I am able to find the doc describing the column separator here : https://docs.starrocks.io/docs/sql-reference/sql-functions/table-functions/files/#csvcolumn_separator

The rule may not be explicit or clear enough and is only in FILES, not sure if this can be a separate page to cross reference in other places where the column separator is used, and user may be able to jump to the specific chart to know how to set the separator and what's the limitation.

The same applies to the row_delimiter. As far as the minor version, I think we can firstly make the rules clear. This fix may be treated as a bugfix and applies to all versions.