author / shell Goto Github PK

Having a useExcept() method (similar to useWith) would help simplify configurations.

The main use case is when all commands except one (or a small minority) require the same middleware. As new commands are created, it becomes increasingly burdensome to modify middleware without this.

Can u show me an example of using a shell command in raw html for browser?

As I've used this library to build applications, I've recognized a need for experimental commands, beta commands, and even hidden commands (hidden from help).

It would be nice to have a "tag" or some other identifier to identify which commands are experimental, which should be hidden, etc.

One possible implementation is:

new Command({
  name: 'cmd',
  tags: ['beta'],
  async handler () {...}
})

Special tags may have special documentation/help features:

• hidden: available, but does not show up
• ignore: completely ignore the command (won't run)
• experimental: prefixes description with EXPERIMENTAL.
• deprecated: prefixes description with DEPRECATED.

Tags should be exposed to command handlers and directly through the shell, making it possible for developers to create custom functionality for tagged commands.

Please use the reactions to cast a vote in favor of or against this feature suggestion »

Instead of requiring users to choose node-shell or browser-shell, it would be better to have a single cross-runtime module.

Note to self: Use the same strategy from NGN to make this happen.

Is your feature request related to a problem? Please describe.
I often find myself wanting to be able to return a value when I run a command.

Describe the solution you'd like
Right now, I have to pass in a callback function to do this:

shell.exec('my-command', value => {...})

It would be nice to be able to do this:

const value = await shell.exec('my-command')

...doStuffWithValue...

This would allow me to simply return a value from my command handler function without having to fire a callback.

Please use the reactions to cast a vote in favor of or against this feature suggestion »

I want a transformer to process arguments/flags passed to the CLI utility. For example:

my cmd -i ./local/path

flags: {
  input: {
    alias: 'i',
    description: 'The input source directory.',
    type: 'string',
    transform (value) {
      return path.resolve(value)
    }
  }
}

In the case above, the resulting "input" flag would have a value of /users/cbutler/local/path instead of ./local/path because the transformer modifies the data before it is used within a command.

This could be used for sanitizing inputs of any kind, or normalizing data (such as case sensitivity, typos, etc). It would be better to do this where the flag is defined, instead of requiring each and every command handler to do this. Currently, middleware is the only way to handle this, but that's less intuitive than a transformer.

Issue:
Middleware assigned to the shell is executing after middleware assigned to specific commands.

How To Reproduce:

let status = []
const cmd = new Command({
  name: 'run',
  async handler(meta) {
    return meta.input
  }
})

cmd.use(async function(meta, next) {
  status.push('2')
  next()
})

const sh = new Shell({
  name: 'test',
  use: [
    (meta, next) => {
      status.push('1')
      next()
    }
  ],
  commands: [cmd]
})

sh.exec(['run', 'other'])

Expected Behavior:
The value of status should be 1,2, but is actually 2,1.

Middleware should be executed according to the hierarchy of the command, i.e.:

graph LR
sh[1. shell middleware]
c[2. command middleware]
s[3. subcommand middleware]

sh-->c
c-->s

It currently runs in the opposite order.

Similar to issue #3, sometimes there are "nearly universal" flags. In other words, there are flags which apply to all but one or two commands within a shell. Having an option to ignore a set of common flag attributes or exclude them from being applied to a command would reduce configuration footprints and simplify maintenance of shells.

I could use some help cleaning up the docs. These were written pretty quickly and there are clearly some errors and typos.

How to use a shell command to run "ls -l" and also "mkdir test" for node

Consider adding a plugin capable of generating documentation in Markdown format (JSON is already supported).