Ask HN: Why aren't Interactive command-line sessions more common in CLI apps?

[fouc]

I'm especially talking about interfaces that include their own query prompt & internal menu system like the fdisk command on linux, or like the recently mentioned GameStonk Terminal [0]

This is one of my absolute favorite types of interfaces because all the commands are easily discoverable, and it's easy to memorize combinations of commands that take you through the different levels of menus.

Often I find when I see an exciting new command-line app mentioned on HN, I'm disappointed when it ends up being primarily operated by parameters only.

For command-line apps that are intended to be used on a daily basis & used with different parameters almost every time, these types of apps could probably benefit hugely from offering a menu & query interface.

[0] https://news.ycombinator.com/item?id=27651435

[asymptosis]

It comes from a constellation of causes:

* Most command line programs are FOSS, which means people are scratching their own itch instead of trying to develop a product intended for consumption by others.

* Adding any kind of interactivity increases the complexity of any program, increasing the difficulty of development and maintenance.

* People working on the command line often want to call the program from other scripts, which is harder for interactive programs.

* Interactivity is not good for composing operations except manually. Non-interactive commands are much more easily composed with each other in sophisticated and customisable pipelines.

[theamk]

Because you need a lot of work to make command-line session to be as good as CLI. Especially my CLI which I have customized just the way I like.

To take fdisk for example: I hate it and try to avoid it as much as possible. For example, after fdisk session, I have no record of the commands I executed in my history file. Copying the command to put into wiki is pretty hard. And I cannot type $((128 * 1024 * 1024)) in place of the argument. And if I want to execute the same sequence of operations multiple times on a different disks, there is no way to do so.

If you want to make a complex command more discoverable, start from a shell completion rule, or a good manual. Don’t bother with interactive session unless you are sure you can make it better than an existing shell.

[fouc → theamk]

That's a good point about the lack of history & difficulty of sharing the commands.

Perhaps there's a middle ground between the shell & fdisk, where they can meet in the middle and produce the same interface, but maintain all the advantages of the shell history and parsing.

[elviejo]

on a related question: Do you know a CLI tool that would allow me to explore a Rest API interactively?

[theamk → elviejo]

For session-based approach, using your favorite scripting language works pretty well. I use Python, so it would be IPython for CLI or Jupyter Notebook for web-based.

If you want individual commands in your shell, then I generally use “curl ..| jq” plus a lot of history editing. A terminal which handles middle mouse button properly helps a lot. If endpoints are rate-limited or non-idempotent, then there is an occasional “| tee .. |” as well.