[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: please correct me on my code?

hadi motamedi wrote:
> On my Debian server, I need to separate the individual logs coming
> from various modules concurrently. Please find below a sample of the
> logs:
> ' IPTR >..

That beginning space I think is going to cause problems when using cut
since cut is very literal about its definition of fields.

> SNMP >..

No leading space here.  I think that means you want to use awk instead.

> #cat Edit3 | tr -d "\r" | while read LINE; do

Please see http://www.google.com/search?q=useless+use+of+cat for why
you shouldn't be using cat that way.

You always want to use read -r to avoid backslash interpretation.
Otherwise backslashes in the input will be lost.

To avoid BASH FAQ E4 becoming an issue I wouldn't pipe into while.
Instead I would redirect into it.  Since you are already using bash I
would use a bash-ism for it.

  while read -r LINE; do
  done < <(tr -d "\r" < test.txt)

> >echo "$LINE" | grep -q '>\.\.'
> >if [ $? -eq 0 ]; then

Instead of using $? there you can just use the exit code of grep
directly.  Much cleaner and clearer.

  if echo "$LINE" | grep -q '>\.\.'; then

> >LOGFILE=`echo $LINE | cut -d' ' -f1`.log

Instead of using backticks since you are already using bash you might
as well use the newer and easier $(...) instead.  The quoting rules
are much simpler.  Plus using cut that way doesn't always yield the
results people expect.  Better to use awk splitting instead.  Plus you
should quote the $LINE for the echo just in case it contains
unfortunate characters.

  LOGFILE=$(echo "$LINE" | awk '{print$1}').log

> >else
> >echo "$LINE" >> $LOGFILE

You have LOGFILE set to a data dependent value.  It may be one word or
it may be two words.  You should quote it.  It is the only way to be

  echo "$LINE" >> "$LOGFILE"

> >fi
> >done
> But when I run it, I am receiving the following error:
> -bash: -f1.log : command not found
> -bash: $LOGFILE : ambiguous redirect
> Can you please help me to correct my code?

> >echo "$LINE" >> $LOGFILE

That error coupled with that redirect line means that $LOGFILE must be
two or more words.  The shell is redirecting to something that has
multiple words and is therefore generating an ambiguous redirect
error.  You can test this yourself by this example:

  $ file="one two"
  $ echo hello >$file
  bash: $file: ambiguous redirect

To debug your own code display the result of the LOGFILE.  Something
like this:
  while read -r LINE; do
    if echo "$LINE" | grep -q '>\.\.'; then
      LOGFILE=$(echo "$LINE" | awk '{print$1}').log
      echo "DEBUG: LOGFILE='$LOGFILE'"
      echo "DEBUG: $LINE >> '$LOGFILE'"
  done < <(tr -d "\r" < test.txt)

I am sure that at that time you will find that LOGFILE is not always
the values that you expect.  It is being set to multiple words.  The
test case you provided does not exhibit this behavior.  But I am sure
your actual data case does show it.

Now to solve some other nagging problems.  Looking for the pattern
">.." doesn't mean that there is a field before it "FOO >.." and so I
think your grep should look for the combination of those things.

    if echo "$LINE" | grep -q '^[^[:space:]][^[:space:]]* >\.\.'; then
      LOGFILE=$(echo "$LINE" | awk '{print$1}').log

Doing it that way will guarantee that there is a pattern to be used
for the logfile on the next line.  And then the data would be more
strictly behaved and the 'cut' would be more reliable there but I
would still use awk in that case.

But going further with that I think it is better avoid spreading the
logic across two lines.  Just try to read the line as a log file
changing line and if so then use it.  Using awk for this is a good

  echo "$LINE" | awk '$2 == ">.." { print $1 }'

If $2 (the second field) is exactly ">.." then the pattern matches and
the associated action of printing $1 (the first field) is executed.
If the first field is printed then we have a new logfile name.  If not
then nothing is printed.  This does the matching and the extracting
all at one step.

  NEWLOGNAME=$(echo "$LINE" | awk '$2 == ">.." { print $1 }')

Plus, I don't like using upper case variable names.  They might
collide with the shell's internal variables.  So better to use lower
case variables for your own script.  Here is a complete example with
all of my suggestions.  After debugging the LOGFILE problem having
multiple words in your original then you might want to try something
like this where it is all put together.

  while read -r line; do
    newlogname=$(echo "$line" | awk '$2 == ">.." { print $1 }')
    if [ -n "$newlogname" ]; then
      echo "$line" >> "$logfile"
  done < <(tr -d "\r" < test.txt)


Attachment: signature.asc
Description: Digital signature

Reply to: