在 postgres 中将列数据类型从 VARCHAR 转换为 ARRAY（使用 Psycopg2）

Question

I have a pandas dataframe df, which I want to store as a schema in a database. By default, it is taking text as the default datatype. I used dtype=sqlalchemy.types.VARCHAR argument to save the columns as VARCHAR.

df.to_sql('data', con, schema='schema', index= False, dtype=sqlalchemy.types.VARCHAR)

However, the datatype of one of my column is a multidimensional "array". I tried following code(s), but I am not getting expected output.

cursor.execute("
alter table data alter col2 drop default;
alter table data alter col2 type text[][] using array[col2];
alter table data alter col2 set default '{}'");

The above code is converting the desired column to text array but the array is empty.

My data looks like this:

col1     col2
A1       A1:DEF, Human; X2:XYZ, Mouse;  Y1:RST, Rat
B1       B1:GHI, Human; Y2:ZXY, Mouse
C1       C1:JKL, Human; Z2:USC, Mouse

I would like to store col1 as VARCHAR and col2 as multidimensional array of size n, storing each part separated by ';'as one element.

For A1:
array[1]: A1:DEF, Human
array[2]: X2:XYZ, Mouse
array[3]: Y1:RST, Rat

Any suggestion, how to make it work?

Thanks

Answer 1

Use the function string_to_array() :

alter table data alter col2 drop default;
alter table data alter col2 type text[] using string_to_array(col2, '; ');
alter table data alter col2 set default '{}';

Next update the table if you really want to get multidimensional arrays:

update data d
set col2 = s.col2
from (
    select col1, array_agg(string_to_array(elem, ', ')) as col2
    from data
    cross join unnest(col2) elem
    group by col1) s
where d.col1 = s.col1;

Db<>Fiddle.