Strange behaviour of read work file

, ,
#1

Hello all,

I got a csv-file from a foreign system and natural seems to read wrong values out of it. OK, I got a workaround for it. But I didn’t expect that behaviour.

Reproduceable sample :

define data local
1 #tab (A1/1:7)
end-define
define work file 1 '/tmp/test.csv' /* type 'csv'
write work file 1 variable 'A' 'B' 'C' 'D' 'E' 'F'
write work file 1 variable '1' '2' '3'
write work file 1 variable 'z' 'y' 'x' 'w'
write work file 1 variable 'a' 'b' 'c'
close work file 1
read work file 1 #tab(*)
write            #tab(*)
end-work
end

Output:


A B C D E F
1 2 3 D E F
z y x w E F
a b c w E F

… a trap for the unwary. Since there is no “read work … variable” my workaround is:


read work file 1 #TAB(*) giving length #csvfields
if #csvfields <  *UBOUND(#TAB)
add 1 to #csvfields
reset #tab(#csvfields:7)
end-if

Regards

matthias

#2

This may not be the same issue… but you need to reset the array, as you are not reading values into all the fields each time

define data local
1 #tab (A1/1:7)
end-define
define work file 1 ‘/tmp/test.csv’ /* type ‘csv’
write work file 1 variable ‘A’ ‘B’ ‘C’ ‘D’ ‘E’ ‘F’
write work file 1 variable ‘1’ ‘2’ ‘3’
write work file 1 variable ‘z’ ‘y’ ‘x’ ‘w’
write work file 1 variable ‘a’ ‘b’ ‘c’
close work file 1
reset #tab() / reset array… do not wish to have OLD data
read work file 1 #tab()
write #tab()
reset #tab() / reset array… do not wish to have OLD data
end-work
end

#3

The behavior is documented, but I’ve never experienced it because I always using a single WRITE WORK for my detail records, so all detail records have the same number of fields.
The SELECT option is the default for a READ WORK.

#4

Thanks for answering.

@James Gilpin: Yes, it is the same issue and your RESET is a valid Workaround, too. I prefer to do a RESET at the beginning of the loop - just in case there are some ESCAPE TOPs in the following code.

@Ralph Zbrog: In my case, a non-Natural-Application writes the workfile… So I wasn’t aware of the individual length of the records.

#5