使用 Apache Spark 在 Databricks 中使用 SQL 查询进行 CASTING 问题

Question

I am trying to use the CAST function on a field in SQL to cast a Time field as a string, but I'm getting unexpected errors

The code is as follows:

select DATE_ADD(cast(LAST_MOD_TIME as STRING) % 100 + CAST(LAST_MOD_TIME AS STRING) / 100 % 100 * 60 + CAST(LAST_MOD_TIME AS STRING) / 10000 * 3600, CAST(LAST_MOD_DATE AS STRING)) from DmWo

But I'm getting the error:

Error in SQL statement: AnalysisException: cannot resolve 'date_add((((CAST(CAST(dmwo.`LAST_MOD_TIME` AS STRING) AS DOUBLE) % CAST(100 AS DOUBLE)) + (((CAST(CAST(dmwo.`LAST_MOD_TIME` AS STRING) AS DOUBLE) / CAST(100 AS DOUBLE)) % CAST(100 AS DOUBLE)) * CAST(60 AS DOUBLE))) + ((CAST(CAST(dmwo.`LAST_MOD_TIME` AS STRING) AS DOUBLE) / CAST(10000 AS DOUBLE)) * CAST(3600 AS DOUBLE))), CAST(dmwo.`LAST_MOD_DATE` AS STRING))' due to data type mismatch: argument 1 requires date type, however, '(((CAST(CAST(dmwo.`LAST_MOD_TIME` AS STRING) AS DOUBLE) % CAST(100 AS DOUBLE)) + (((CAST(CAST(dmwo.`LAST_MOD_TIME` AS STRING) AS DOUBLE) / CAST(100 AS DOUBLE)) % CAST(100 AS DOUBLE)) * CAST(60 AS DOUBLE))) + ((CAST(CAST(dmwo.`LAST_MOD_TIME` AS STRING) AS DOUBLE) / CAST(10000 AS DOUBLE)) * CAST(3600 AS DOUBLE)))' is of double type. argument 2 requires (int or smallint or tinyint) type, however, 'CAST(dmwo.`LAST_MOD_DATE` AS STRING)' is of string type.; line 1 pos 7;
'Project [unresolvedalias(date_add((((cast(cast(LAST_MOD_TIME#88591 as string) as double) % cast(100 as double)) + (((cast(cast(LAST_MOD_TIME#88591 as string) as double) / cast(100 as double)) % cast(100 as double)) * cast(60 as double))) + ((cast(cast(LAST_MOD_TIME#88591 as string) as double) / cast(10000 as double)) * cast(3600 as double))), cast(LAST_MOD_DATE#88590 as string)), None)]

Any thoughts on where I might be going wrong with the CAST function?

Answer 1

There are several mistakes here:

1. The first argument to DATE_ADD function is the start_date , and needs to be a date type, but you're passing in the time of the day in seconds based on the logic, and the type is double due to the division operator.
2. The second argument to DATE_ADD function is the num_days after the starting date and needs to be int type. You're passing in the starting date itself as a string type.
3. DATE_ADD function returns the date that is num_days after start_date , but you are adding the date and the time of day counted in seconds. That's not what this function does for you.

A simpler way is to use the to_timestamp() function:

select to_timestamp(concat(LAST_MOD_DATE, LAST_MOD_TIME), 'yyyyMMddHHmmss')
from DmWo