In the `computed column` section of [1], i saw some related doc: ``` On the other hand, computed column can be used to derive event time column because an event time column may need to be derived from existing fields, e.g. the original field is not TIMESTAMP(3) type or is nested in a JSON string. ``` could you please provide a concrete example for this? Thanks Fanbin
[1] https://ci.apache.org/projects/flink/flink-docs-master/dev/table/sql/create.html#create-table On Tue, Nov 10, 2020 at 6:18 PM Fanbin Bu <fanbin...@coinbase.com> wrote: > i also tried: > ts TIMESTAMP WITH LOCAL TIME ZONE > > but it failed with > Rowtime attribute 'ts' must be of type TIMESTAMP but is of type > 'TIMESTAMP(6) WITH LOCAL TIME ZONE'. > > On Tue, Nov 10, 2020 at 5:42 PM Fanbin Bu <fanbin...@coinbase.com> wrote: > >> Hi, >> >> I have source json data like: >> {"ts": "2020-11-09T20:26:10.368123Z", "user_id": "user1", "action": >> "click"} >> ... >> >> my sql is: >> create table t ( >> user_id string, >> action string, >> ts timestamp, >> watermark for ts as ts - interval '5' second >> ) with ( >> 'connector' = 'kafka', >> 'topic' = 'test', >> 'json.timestamp-format.standard' = 'ISO-8601' >> ... >> ) >> >> this does not work since ISO-8601 does not expect `Z` at the end of the >> timestamp. >> It only works for "2020-11-09T20:26:10.368". However, I'm not able to >> change the format. >> >> I checked a few threads and somebody suggested to use udf for unsupported >> timestamp format. what would the create table statement look like? I also >> need watermark working. >> >> I'm thinking about this: >> create table t ( >> user_id string, >> action string, >> ts string, >> transform_ts_format(ts) as new_ts, >> watermark for new_ts as new_ts - interval '5' second >> ) with ( >> ... >> >> q: >> 1. how does watermark know new_ts is a valid timestamp? >> 2. is it possible to reuse ts without introducing a new column? >> >> Thanks, >> Fanbin >> >>
