When scripting in bash or any other shell in *NIX, while running a command that will take more than a few seconds, a progress bar is needed.
For example, copying a b
I did a pure shell version for an embedded system taking advantage of:
/usr/bin/dd's SIGUSR1 signal handling feature.
Basically, if you send a 'kill SIGUSR1 $(pid_of_running_dd_process)', it'll output a summary of throughput speed and amount transferred.
backgrounding dd and then querying it regularly for updates, and generating hash ticks like old-school ftp clients used to.
Using /dev/stdout as the destination for non-stdout friendly programs like scp
The end result allows you to take any file transfer operation and get progress update that looks like old-school FTP 'hash' output where you'd just get a hash mark for every X bytes.
This is hardly production quality code, but you get the idea. I think it's cute.
For what it's worth, the actual byte-count might not be reflected correctly in the number of hashes - you may have one more or less depending on rounding issues. Don't use this as part of a test script, it's just eye-candy. And, yes, I'm aware this is terribly inefficient - it's a shell script and I make no apologies for it.
Examples with wget, scp and tftp provided at the end. It should work with anything that has emits data. Make sure to use /dev/stdout for programs that aren't stdout friendly.
#!/bin/sh
#
# Copyright (C) Nathan Ramella (nar+progress-script@remix.net) 2010
# LGPLv2 license
# If you use this, send me an email to say thanks and let me know what your product
# is so I can tell all my friends I'm a big man on the internet!
progress_filter() {
local START=$(date +"%s")
local SIZE=1
local DURATION=1
local BLKSZ=51200
local TMPFILE=/tmp/tmpfile
local PROGRESS=/tmp/tftp.progress
local BYTES_LAST_CYCLE=0
local BYTES_THIS_CYCLE=0
rm -f ${PROGRESS}
dd bs=$BLKSZ of=${TMPFILE} 2>&1 \
| grep --line-buffered -E '[[:digit:]]* bytes' \
| awk '{ print $1 }' >> ${PROGRESS} &
# Loop while the 'dd' exists. It would be 'more better' if we
# actually looked for the specific child ID of the running
# process by identifying which child process it was. If someone
# else is running dd, it will mess things up.
# My PID handling is dumb, it assumes you only have one running dd on
# the system, this should be fixed to just get the PID of the child
# process from the shell.
while [ $(pidof dd) -gt 1 ]; do
# PROTIP: You can sleep partial seconds (at least on linux)
sleep .5
# Force dd to update us on it's progress (which gets
# redirected to $PROGRESS file.
#
# dumb pid handling again
pkill -USR1 dd
local BYTES_THIS_CYCLE=$(tail -1 $PROGRESS)
local XFER_BLKS=$(((BYTES_THIS_CYCLE-BYTES_LAST_CYCLE)/BLKSZ))
# Don't print anything unless we've got 1 block or more.
# This allows for stdin/stderr interactions to occur
# without printing a hash erroneously.
# Also makes it possible for you to background 'scp',
# but still use the /dev/stdout trick _even_ if scp
# (inevitably) asks for a password.
#
# Fancy!
if [ $XFER_BLKS -gt 0 ]; then
printf "#%0.s" $(seq 0 $XFER_BLKS)
BYTES_LAST_CYCLE=$BYTES_THIS_CYCLE
fi
done
local SIZE=$(stat -c"%s" $TMPFILE)
local NOW=$(date +"%s")
if [ $NOW -eq 0 ]; then
NOW=1
fi
local DURATION=$(($NOW-$START))
local BYTES_PER_SECOND=$(( SIZE / DURATION ))
local KBPS=$((SIZE/DURATION/1024))
local MD5=$(md5sum $TMPFILE | awk '{ print $1 }')
# This function prints out ugly stuff suitable for eval()
# rather than a pretty string. This makes it a bit more
# flexible if you have a custom format (or dare I say, locale?)
printf "\nDURATION=%d\nBYTES=%d\nKBPS=%f\nMD5=%s\n" \
$DURATION \
$SIZE \
$KBPS \
$MD5
}
Examples:
echo "wget"
wget -q -O /dev/stdout http://www.blah.com/somefile.zip | progress_filter
echo "tftp"
tftp -l /dev/stdout -g -r something/firmware.bin 192.168.1.1 | progress_filter
echo "scp"
scp user@192.168.1.1:~/myfile.tar /dev/stdout | progress_filter