如何将 json 字符串数据类型列转换为 hive 中的映射数据类型列？

我需要从所有行中获取所有唯一的键值. 每行都有不同的键和值请找到上面的专栏图片。

例如：一排看起来像

{"START_TIME":1549002807568,"PARSING.QUERY_FORMED":1549002807586,"CUBES_WITH_PERMISSIONS":1549002807568,"PARSING.CUBE_MATCH_SELECTED":1549002807586,"POTENTIAL_COMPLETIONS_ADDED":1549002807587,"QUERY_PARSED":1549002807586,"SUGGESTIONS_FORMED":1549002807606,"PARSING.SEQUENCES_GENERATED":1549002807568,"PARSING.NGRAM_MATCHES_CACHED":1549002807585}

使用两行数据进行测试，所有键值对都是相同的，除了第二个 JSON 中有一个额外的NEW_KEY and PARSING.NGRAM_MATCHES_CACHED价值观不同。

with data as
(
select stack(2, --data example
'{"START_TIME":1549002807568,"PARSING.QUERY_FORMED":1549002807586,"CUBES_WITH_PERMISSIONS":1549002807568,"PARSING.CUBE_MATCH_SELECTED":1549002807586,"POTENTIAL_COMPLETIONS_ADDED":1549002807587,"QUERY_PARSED":1549002807586,"SUGGESTIONS_FORMED":1549002807606,"PARSING.SEQUENCES_GENERATED":1549002807568,"PARSING.NGRAM_MATCHES_CACHED":1549002807585}',
'{"NEW_KEY":12345,"START_TIME":1549002807568,"PARSING.QUERY_FORMED":1549002807586,"CUBES_WITH_PERMISSIONS":1549002807568,"PARSING.CUBE_MATCH_SELECTED":1549002807586,"POTENTIAL_COMPLETIONS_ADDED":1549002807587,"QUERY_PARSED":1549002807586,"SUGGESTIONS_FORMED":1549002807606,"PARSING.SEQUENCES_GENERATED":1549002807568,"PARSING.NGRAM_MATCHES_CACHED":154900280758}'
) as str
)

select str_to_map(concat_ws(',',collect_set(key_value)),',',':') --collect set, concatenate and convert to map
from
(
select explode(split(regexp_replace (str,'[{}"]',''),',')) key_value from data --remove JSON delimiters, split and explode pairs
)s;

Result:

OK
{"START_TIME":"1549002807568","PARSING.QUERY_FORMED":"1549002807586","CUBES_WITH_PERMISSIONS":"1549002807568","PARSING.CUBE_MATCH_SELECTED":"1549002807586","POTENTIAL_COMPLETIONS_ADDED":"1549002807587","QUERY_PARSED":"1549002807586","SUGGESTIONS_FORMED":"1549002807606","PARSING.SEQUENCES_GENERATED":"1549002807568","PARSING.NGRAM_MATCHES_CACHED":"154900280758","NEW_KEY":"12345"}
Time taken: 158.414 seconds, Fetched: 1 row(s)

当然，"PARSING.NGRAM_MATCHES_CACHED"结果中只存在一次，因为map不允许相同的key出现两次。所有的 key_values 都是唯一的。 请阅读代码中的注释。