Hive table with dynamic number of columns - sql

Question

TestTable
inputsCOLUMN
3-300-150-150-R
3-200-100-100-A
5-500-00-500-A
output
3_open 3_spent 3_closing 3_type 5_open 5_spent 5_closing 5_type
-------- --------- ----------- -------- -------- --------- ----------- --------
300 150 150 R 500 00 500 A
200 100 100 A
Above is the input table called TestTable. It has two columns that contains rows of data(strings)
And there is a desired output table of which the column names are based on the input string.
the column name is the first number on the string + another string name, like CONCAT(split(inputsCOLUMN,'\\-')[0],'-','type')
so that output is the desired output. and the below query is not working as desired because of that part when i am trying to concatenate an alias i think is not allowed. so help me if there is a way i can find that desired output.
SELECT split(inputsCOLUMN,'\\-')[1] as CONCAT(split(inputsCOLUMN,'\\-')[0],'-','open'),
split(inputsCOLUMN,'\\-')[2] as CONCAT(split(inputsCOLUMN,'\\-')[0],'-','spent'),
split(inputsCOLUMN,'\\-')[3] as CONCAT(split(inputsCOLUMN,'\\-')[0],'-','closing'),
split(inputsCOLUMN,'\\-')[4] as CONCAT(split(inputsCOLUMN,'\\-')[0],'-','type')

Answer 1

Hive cannot have a dynamic number of columns, and it cannot have dynamic column names. It must be able to determine the entire schema (column count, types, and names) at query planning time, without looking at any data.
It's also not clear to me how exactly you're matching up input records into a single row. For example, how do you know which "3" record corresponds to which "5" record.
If you knew that, for example, there would always be a "3" record and a "5" record and you could commit to those being the only column names, and if you had a consistent way of matching up records to "flatten" this data, then it is possible, but difficult. I've done almost this exact operation before, and it involved a custom UDTF and a custom UDAF, and some code to auto-generate the actual query, which ended up being hundreds of lines long in some cases. I would re-evaluate why you want to do this in the first place and see if you can come up with another approach.