I am converting an old dataset into a newly structured data开发者_C百科base. Currently they have data in this format:
[quantity int]~[var1 string]|[var2 string optional]|[var3 string optional];(etc);
[quantity]~[var1]|[var2]|[var3]
where various patterns are delimited first by semicolons, then by tildes, and finally by pipe characters within the same field (no brackets) within the existing database. For instance:
3~S|Red|Top;1~S|Blue|Top; ... etc ... ;20~XL|Green|Left
I want to take that string in each record and do the following:
- Create a new record in an existing table based off each semicolon split
- Split each token again by the tilde, putting the first piece into one field and the second into a separate field (I don't care about splitting by pipe) in the same record
I don't know if it's even possible, but I'm hoping that it is. I am also hoping to only have to do this once, as the new system will take care of creating all this crap automatically. My sproc-fu is quite awful in MySQL, so any help is much appreciated.
Thanks so much!
This seems like a pretty gross thing to do in SQL, but here's a proof-of-concept script.
If the output looks right to you, replace the SELECT with an appropriate INSERT statement and it should get you what you want.
delimiter ;;
drop procedure if exists load_crazy_stuff;;
create procedure load_crazy_stuff(in s longtext)
begin
declare pos int;
declare record longtext;
declare leftpart int;
declare rightpart longtext;
set s = concat(s,';');
while length(s)>0 DO
set pos = instr(s,';');
set record = left(s,pos-1);
set s = substr(s,pos+1);
set pos = instr(record,'~');
set leftpart = left(record,pos-1);
set rightpart = substr(record,pos+1);
select leftpart, rightpart;
end while;
end;;
call load_crazy_stuff('3~S|Red|Top;1~S|Blue|Top;20~XL|Green|Left');;
The non-gross way to do it is this:
load data infile '/tmp/your-data-file' into table yourtable fields terminated by '~' lines terminated by ';';
精彩评论