- Keep it simple and explicit
- Log stdout and stderr output to files
- Keep stderr and stdout separate. Apps or programs calling your script may want stderr output
- Prepend a time stamp to each line of output
- Add a separator to demarc each new run of the script
Example:
LOG_FILE=log
# Prepend a timestamp to each line, send it to log.err and backout to stdout (via tee)
exec 2> >(while read line; do echo "$(date +%Y-%m-%d::%H:%M:%S) :: $line"; done | tee -a ${LOG_FILE}.err)
# Prepend a timestamp to each line, send it to log and backout to stdout (via tee)
exec > >(while read line; do echo "$(date +%Y-%m-%d::%H:%M:%S) :: $line"; done | tee -a ${LOG_FILE})
# Another alternative (But it combines both stderr and stdout)
# exec &> >(while read line; do echo "$(date +%Y-%m-%d::%H:%M:%S) :: $line"; done | tee -a ${LOG_FILE})
# Add a demarcation point to show a new run of the script
echo -e "###### NEW LOG ###### \n\n" >> ${LOG_FILE}.err
echo -e "###### NEW LOG ###### \n\n" >> ${LOG_FILE}
Create a function to carry out cleanup tasks, then use the trap builtin to call the function in the event that users hits control+c or the script exits early for some reason.
Example:
#
# Add to the top of your script so you can catch things like:
# * control + c
# * script exiting unintentionally
#
function cleanup() {
echo "Cleaning up"
rm /file/with/sensitiveinfo
}
trap cleanup SIGHUP SIGINT SIGTERM EXIT
- I have been using the following standards in my functions:
- Use the keyword
localfor locally defined function variables. - prefix local variables with an underscore: _varname <- python influenced
- Function names: make them descriptive and use underscores between words
- Define a function header, this has been good for standardizing output and having a clear definition of where the script exited or where you are in it as it runs.
- Define an error function to be called on any errors
- Use the keyword
function function_header() {
# Where ever this is included it will write out the function header
# and any message passed in. The Header name will be parsed and any
# underscores will be replaced with a space
local _message=$1
echo ""
echo "Function: ${FUNCNAME[1]//_/ }"
echo "================================================================================"
echo ""
echo -e " ${_message}"
echo ""
}
Leverage a function to capture errors and print a standardized message followed by an exit.
function error() {
local _return_code=$1
local _message=$2
if [ ${_return_code} -ne 0 ]; then
echo ""
echo " ----------------------------------------------------------------------------------------"
echo ""
echo " Error in: ${FUNCNAME[1]}"
echo " Return Code: ${_return_code}"
echo -e " Message: ${_message}"
echo ""
echo " ----------------------------------------------------------------------------------------"
echo ""
exit 1
fi
}
Examples of using the error call with return code of the previous command:
# The redirect at the end is optional, if this is customer facing script
# you may want to hide stderr and present the user with another message.
ls /a/file/that/does/not/exist 2>/dev/null
error $? "Unable to list the file"
Output:
----------------------------------------------------------------------------------------
Error: main
Return Code: 1
Message: Unable to list the file
----------------------------------------------------------------------------------------
Examples of using the error call when the command returns 0 but result still constitutes an error:
# In cases where the return code returned is 0 but the result is not
# the expected result you can use the error function like this:
_http_return_code=$(curl -s -I http://www.example.org/doesnotexist | head -n 1| cut -d$' ' -f2)
if [[ ${_return_code} -ne 200 ]]; then
error 1 "Unable to download file"
fi
Output:
----------------------------------------------------------------------------------------
Error: main
Return Code: 1
Message: Unable to download file
----------------------------------------------------------------------------------------
- Add a description at the top (But inside the function)
- Uses
localto scope variables locally and predefine variables as needed at the top - calls the function header
- calls the error function
function example__pre_install() {
# Example description if needed
local _msg="Carrying out preinstall steps..."
local _system=$(python -c 'import platform; print(platform.system());')
local _dist=''
function_header "${_msg}"
if [[ ${_system} == "Darwin" ]]; then
_dist=$(python -c "import platform; print(platform.mac_ver());")
else
_dist=$(python -c "import platform; print(platform.dist());")
fi
echo " * ${_system}: ${_dist}"
echo " * Verifying system..."
# Do Verification stuff.
echo " * Installing packages..."
# Do Install stuff
_http_return_code=$(curl -s -I http://www.example.org/exampleinstaller | head -n 1| cut -d$' ' -f2)
if [[ ${_return_code} -ne 200 ]]; then
error 1 "Unable to download file"
fi
}
Leverage python, it's generally available on most systems, has builtins to obtain the OS and distribution details, and short snippets can easily be pulled into your script.
- cross platform way to get the OS:
$ python -c 'import platform; print(platform.system());'
Darwin
- cross platform way to get the distribution:
$ python -c "import platform; print(platform.dist());"
('Ubuntu', '14.04', 'trusty')
printf "Waiting for command to complete"
for i in $(seq 1 10); do
printf "."
sleep 1
done
printf "\n"