A quick redis consumer in bash (using a custom silex console in this case, but you get the point)

consume.bash

#!/bin/bash

POP_FROM="jobs"
PUSH_TO="jobs_running"

REDIS_CLI=$(which redis-cli)
CONSOLE_DIR=$(cd $(dirname "$0"); pwd)/..;

while job=$($REDIS_CLI brpoplpush $POP_FROM $PUSH_TO 0); do
    eval $(echo $CONSOLE_DIR/console $job);
done;

eval $(echo ...) ? why not simply use $CONSOLE_DIR/console $job)

Good question, thanks for asking! The reason is because of the way bash handles arguments, quoting and parameter expansion.

Say you have a job that has an argument that can contain space, you'll want to describe the job using a string like the following:

foo:bar arg1 "arg2 with spaces"

So from a purely "stringy" point of view, that's what $job would contain. From a bash point of view, $job contains 5 words:

1. 'foo:bar'
2. 'arg1'
3. '"arg2'
4. 'with'
5. 'spaces"'

You can check this in your terminal by running set -v and set -x (if you use bash), they tell bash to show you how it expands parameters before executing a command:

$ set -v
$ set -x
$ job='arg1 "arg2 with spaces"'
$ echo $job

(type set +v and set +x to disable them)

And your console call will (most likely) end up failing with a "too many arguments" error message. That's why we eval $job instead of carelessly using it!