Hi John,
Tried out Ted's suggestion. Works fine in Drill 1.13. Here is my "test2.json"
file:
{ "$f1": "fred", "f$2": "wilma", "f3$": "barney" }
Then ran two queries:
SELECT * FROM `test2.json`;+-------+--------+---------+| $f1 | f$2 | f3$
|+-------+--------+---------+| fred | wilma | barney
|+-------+--------+---------+
SELECT `$f1`, `f$2`, `f3$` FROM `test2.json`;
+-------+--------+---------+| $f1 | f$2 | f3$
|+-------+--------+---------+| fred | wilma | barney
|+-------+--------+---------+
So, the trick you are looking for is to use the back-tick quote for names.
Back-ticks are need to quote names that are SQL reserved words or that contain
non-symbol characters. Notice how the queries above also quote the file name to
avoid issues.
The error you received, however, suggests that something is amiss with your
JSON syntax. As Ted suggested, can you show us a few example lines?
Thanks,
- Paul
On Monday, August 27, 2018, 4:38:07 PM PDT, Ted Dunning
<[email protected]> wrote:
Can you post a sample file with, say, 5-10 lines?
Is it the file names?
Or the data values that are giving you fits?
On Mon, Aug 27, 2018, 12:51 John Folkers <[email protected]> wrote:
> Hello, I downloaded Drill over the weekend, and I love it.
>
>
> Problem: $ string token in a key.
>
>
> Question: How can I get Drill to not trip on the $ string token when it
> sees it inside the keyname?
>
>
> Error Message
> Error: DATA_READ ERROR: Failure while reading ExtendedJSON typed value.
> Expected a VALUE_NUMBER_INT but received a token of type VALUE_STRING
>
>
> Solution is to remove $ string token from key
> sed -i 's/\$key/key/g' <file>
>
>
> Once I do the above, Drill can now query it. I have no control over
> source. I have TBs of this streaming NDJSON I'd love to sort through with
> Drill. I can do this if I can solve this. Please help.
>
> Thanks,
> John
>
>
>
