如何在雪花上加载一个 JSON ，它有一个字面量 unicode 转义字符“\\uNo”

Question

I have the following JSON:

{
  "name": "foo \\uNo bar"
}

I'm trying to load this into Snowflake using a STAGE on S3. This is in a CSV file like:

{"name": "foo \\uNo bar"}

However, when I try to load it, Snowflake breaks with an Error parsing JSON message. If I try to load it directly on Snowflake console, as SELECT PARSE_JSON('{"name": "foo \\uNo bar"}') , I get:

Error parsing JSON: hex digit is expected in \u???? escape sequence, pos 17

The problem is that Snowflake is parsing the string, checking for an unicode digit \uNo (which doesn't exist). How can I disable this?

Answer 1

The default FILE FORMAT for parsing CSVs in Snowflake is interpreting the double backslash string '{"name": "foo \\uNo bar"}' as an escape sequence for the character \ which means that the character sequence \uNo is getting passed to PARSE_JSON which then fails because \uNo not a valid escape sequence for a JSON string. You can prevent this by overriding the FILE FORMAT escape sequence settings.

Given this CSV file:

JSON
'{"name": "foo \\uNo bar"}'

And the following CREATE TABLE and COPY INTO statements:

CREATE OR REPLACE TABLE JSON_TEST (JSON TEXT);

COPY INTO JSON_TEST
FROM @my_db.public.my_s3_stage/json.csv
FILE_FORMAT = (TYPE = CSV
               SKIP_HEADER = 1
               FIELD_OPTIONALLY_ENCLOSED_BY = '\''
               ESCAPE = NONE
               ESCAPE_UNENCLOSED_FIELD = NONE);

I am able to parse there result as JSON:

SELECT PARSE_JSON(JSON) FROM JSON_TEST;

Which returns

+-----------------------------+
| JSON                        |
+-----------------------------|
| { "name": "foo \\uNo bar" } |
+-----------------------------+