Very Computer

by **Brad Slavi** » Fri, 03 May 1996 04:00:00

I have a question based on 'unix etiquette as it were'

question as follows

I am involved in the development of a text parsing utility for a large
product, (unix platform)... I have several options when dealing with standard
out and standard error...

1) output both standard responses and errors to standard out and errors to
standard error

2) output normal responses to standard out and errors to standard error giving
the line number of the input file that the error occured on...

which is the accepted solution?

-Brad Slavin

Unix QA Engineer
Asymetrix Corp

by **Lou** » Sun, 05 May 1996 04:00:00

As a systems admin, I prefer programs that send errors to
std err, and ONLY regular output to stdout. This helps
when piping commands, running cmds in wrapper scripts that
do things with the output, create error logs, etc. At
least DOCUMENT what you decide to do, so that I know what
to expect. For example, grep returns nothing when there is no
match, so I can do `grep string myfile > junk`. I can now process junk,
because I know what the -input data looks like- (programming 101).
It either is empty, or contains "string" I can wc -l,or if -f junk,
etc. I would get really screwed up if junk contained
"there were no occurances of 'string' in file 'myfile'"

Regards,
Lou

Very Computer

standard out and standard error

standard out and standard error

standard out and standard error