before jumping on the puredis-loader implementation, here is how I see the specs, please tell me if I get what you think:
___________________________________
Strings:

string.csv:
KEY, VALUE
key1, value1
key2, value2

[lpuredis string string.csv]

-> creates a String for each line
___________________________________
List and Sets:

list.csv:
KEY, VALUE1, VALUE2, ...
key1, item1, item2, item3
key2, item1, item2, item3

[lpuredis list list.csv]

-> creates a List for each line

[lpuredis set list.csv]

-> creates a Set for each line
___________________________________
Sorted Sets:

zset.csv:
KEY, VALUE1, SCORE1, VALUE2, SCORE2
key1, item1, 1, item2, 2, item3, 3
key2, item1, 4, item2, 5, item3, 6

[lpuredis zset zset.csv]

-> creates a Sorted Set for each line
____________________________________
Hashes

hash.csv:
KEY, KEY1, VALUE1, KEY2, VALUE2, KEY3, VALUE3
key1, hkey1, value1, hkey2, value2, hkey3, value3
key2, hkey1, value1, hkey2, value2, hkey3, value3

or even:
KEY, FIELD1, FIELD2, FIELD3
key1, item1, item2, item3
key2, item1, item2, item3

[lpuredis hash hash.csv]

-> creates a Hash for each line
_______________________________________________________


For SQL file I am clueless, but I guess it will be similar...  
JSON would be the easiest to translate but would it be of any use to pd users?

L-P