I think you can use something along these lines (only some ideas, I made a lot of implicite assumptions):
LOAD trim(@1) as tmpString, if(not wildmatch(@1,'*$*') and not previous(wildmatch(@1,'*$*')), previous(@1), peek(Header1)) as Header1,
if(not wildmatch(@1,'*$*'), @1, peek(Header2)) as Header2, if(not wildmatch(@1,'*$*'),1) as isHeader
(txt, codepage is 1252, no labels, delimiter is '\t', header is 17 lines);
LOAD LineNo, Header1, Header2, Product, [Product Description], left(tmpString3,index(tmpString3,' ')) as [Service Category], money#(trim(mid(tmpString3,index(tmpString3,' '))),'$#','.',',') as [Price in USD];
LOAD LineNo, Header1, Header2, Product, left(tmpString2,index(tmpString2,' ')) as [Product Description], trim(mid(tmpString2,index(tmpString2,' '))) as tmpString3;
LOAD recno() as LineNo, Header1, Header2, left(tmpString,index(tmpString,' ')) as Product, trim(mid(tmpString,index(tmpString,' '))) as tmpString2
resident TMP where not isHeader=1;
drop table TMP;
TMP table just handles the two headers I identified (If you have more then two levels, well, this will already break it). I skipped the first lines, if you need to load in effective date, it should not be too difficult to add this as an additional date field.
RESULT table loads are handling the separation of the different fields, mostly string operations.
See also attached,
comm51927.qvw 160.8 K
Sorry, no wizard involved.
If you want to parse in a file like that, you need to define the structure of the data, so I started with Header1, Header2 and detail lines. How to distinguish between them etc.
Then try to structure the detail lines (your actual fields) and try to parse them (I used a trim & crop technique).
There are several loads involved, if you stop after each load and look at the created table, it should be quite clear how it works.