another real-world-example (with --interactive option)

[felixlohmeier]

Hi Danny,

We had a discussion on catch-all-advanced #199 some time ago and I would now like to report back that your advice to use repeatable args worked well. I used Bashly to create a command line client for OpenRefine: https://github.com/opencultureconsulting/orcli/blob/main/src/bashly.yml. The client has many commands and options and Bashly has helped a lot in keeping the code structured and maintainable.

Of general interest to the Bashly community might be the --interactive option. I'm not sure if my solution works well in all scenarios, but for now it seems to work fine. With ./orcli run --interactive the user gets a shell with autocompletion and alias without prior installation. Maybe this is an idea for another Bashly library?

https://github.com/opencultureconsulting/orcli/blob/6c47e108f7b978a030e57bd05e7421fece5813ab/src/run_command.sh#L84-L96

if [[ ${args[--interactive]} ]]; then
    # case 2: execute scripts and keep shell running
    bash --rcfile <(
        cat ~/.bashrc
        if ! command -v orcli &>/dev/null; then
            echo "alias orcli=${scriptpath}/orcli"
        fi
        for i in "${!files[@]}"; do
            log "executing script ${files[$i]}..."
            awk 1 "${files[$i]}"
        done
        interactive
    ) -i </dev/tty
(...)

https://github.com/opencultureconsulting/orcli/blob/56272613f2791e76aa890516ea84ca2a5c3adc47/src/lib/interactive.sh

function interactive() {
    cat <<'EOF'
PS1="(orcli) [\u@\h \W]\$ "
source <(orcli completions)
echo '================================================================'
echo 'Interactive Bash shell with OpenRefine running in the background'
echo 'Use the "orcli" command and tab completion to control OpenRefine'
echo 'Type "history -a FILE" to write out your session history'
echo 'Type "exit" or CTRL-D to destroy temporary OpenRefine workspace'
echo '================================================================'
EOF
}

Thanks again for your advice!

[DannyBen]

Nice. Thanks for sharing.

1. Can I add your CLI to the real world examples?
2. I will take a deeper look at your interactive implementation. I like interactive terminals. In fact, bashly uses my mister_bin gem, which also comes with an interactive terminal feature.

[DannyBen]

Alright. So here are my thoughts:

Generic interactive terminal

When I think about an interactive terminal, I think of something like the redis CLI or the Postgres CLI: In both, you enter their own "world" with their own commands. It is not a bash subshell.

To create such an experience for any command, we can use rlwrap as the below bash script demonstrates:

#!/usr/bin/env bash
if [[ $# -eq 0 ]]; then
  echo "Interactive shell"
  echo "  Usage:   $0 COMMAND"
  echo "  Example: $0 git"
  exit 1
fi

command="$1"

rlwrap \
  --always-readline \
  --complete-filenames \
  --no-children \
  --prompt-colour=Yellow \
  --substitute-prompt "$command > " \
  -t dumb \
  bash -c "while read -e line; do "$command" \$line || true; done"

running it with any command, will start a shell, where the specified command is pre-populated always.

I called this script it (for interactive terminal), here is how it looks:

I am posting it here, thinking perhaps it would be helpful for your users, using your shell.

Built in bashly add shell

In regards to adding something like this (or like the approach you took) as a standard library - I wonder if it is a common enough use case, and if it can be done in a generic way. Your approach - as I understand it - adds things specific to your use case, and furthermore, requires to have completions in order to provide some value.

I wonder if this could (or should) be generalized.

Adding a bashly shell command

As I mentioned, the command line library that bashly itself uses, supports creating an interactive shell of the command it describes.

So I went ahead and created a bashly shell command (#391), which I think could be useful. In it, you can run any bashly command with these advantages:

1. You do not need to type bashly repeatidly
2. It is faster, since it does not re-invoke Ruby

Here is a quick example of how it works.

Thoughts?

[felixlohmeier]

Can I add your CLI to the real world examples?

Sure, thanks a lot.

Generic interactive terminal
When I think about an interactive terminal, I think of something like the redis CLI or the Postgres CLI: In both, you enter their own "world" with their own commands. It is not a bash subshell.

True, this is probably the much more common use case and user expectation.

To illustrate the orcli use case: A full bash shell should indeed be available so that other unix tools can be used. The flow would be something like this:

1. user enters "orcli run --interactive": orcli starts OpenRefine as a background service with a temporary directory and provides a shell with orcli autocompletion.
2. user plays with orcli and also uses bash for loop, &/wait for parallelization and other unix tools like wc, time etc.
3. user terminates shell: orcli takes care of killing OpenRefine and deleting the temporary directories.

[DannyBen]

• Added orcli to examples.
• I will merge the bashly shell PR.