Hi

Use a tFilterRow component to filter the rows.
source file:
Org Id,f1,f2
ORG001,abc,3
ORG001,ghi,5
ORG001,mno,7
     ,(bla),(bla)
Org Id,f1,f2
ORG002,abc,8
ORG002,ghi,9
ORG002,mno,4
     ,(bla),(bla)

Result:

Starting job test at 10:43 21/02/2008.
.------+---+--.
|  tLogRow_3  |
|=-----+---+-=|
|org_id|f1 |f2|
|=-----+---+-=|
|ORG001|abc|3 |
|ORG001|ghi|5 |
|ORG001|mno|7 |
|ORG002|abc|8 |
|ORG002|ghi|9 |
|ORG002|mno|4 |
'------+---+--'

Job test ended at 10:43 21/02/2008. [exit code=0]

Best regards

          shong

Thank you both for your responses - both methods have their pros/cons.

In my original post I over-simplified the example data - there are actually about 50 columns beyond the "ORGnnn" start field - so both the manual type conversions or regex may be a pain (and I guess not maintainable using metadata?)

Is there an equivalent to a fileDelimited component that works downstream from the start component - i.e. applies a schema to each row (string) that's passed to it?

I imagined something like:

[text line reader] -> [filter (startswith "ORG")] -> [rowDelimited] -> [tMap] -> etc

Is this possible?

The attributes are the same for each "ORGnnn" row - I have defined a schema in the repository for this and I wanted to use that rather than hard-code the type conversion functions.

Looks like the cleanest solution is to filter the lines first into a temporary file and then use the standard delimited reader with the repository schema.

Many thanks for your help - I've learned two useful TOS methods!