# Nushell core team meeting 2024-04-17
## Attendees
- Filip
- Ian
- Stefan
- Jack
- Michael
- Devyn
- Andy
- Wind
- Jakub
- Darren

## Agenda
- [x] benchmarks
- [x] `echo`
- [x] [#12344](https://github.com/nushell/nushell/pull/12344) use of `--` and how to be consistent, re: [#11435](https://github.com/nushell/nushell/pull/11435)?
- [x] Any final thoughts on [plugin commands](https://hackmd.io/lG8T5nTETaGuHO4GgFKK_w)?
- [ ] Stepped ranges [#12537](https://github.com/nushell/nushell/pull/12537): syntax change or possibly removal?

## Discussed Topics

### Benchmark
- different frameworks available (Filip has several competing PRs)
- [divan](https://github.com/nvzqz/divan)
- fresh
- no file output
- simpler code
- manual work needed to compare changes (Filip built a manual parsing script)
- [criterion](https://github.com/bheisler/criterion.rs)
- old
- statistics/file output
- [tango](https://github.com/bazhenov/tango)
- really young
- A/B testing of two commits/branches in parallel
- (you can still run a single)
- little bit more setup
- supports file output
- doesn't use proc macros :smile:
- not supported on Windows without nightly
- Filip has been in active contact with the dev
- [callgrind](https://valgrind.org/docs/manual/cl-manual.html)
- *no current PR*
- not a library, not a Rust thing
- would need a script
- not timing-sensitive (should be more comparable)

→ Use Tango without git.

## `echo`

change to the `echo` semantics stirring debates

previously trying to emulate bash expectations
- if redirected just yielding a value
- if bare trying to print

Strong muscle memory present for both

edge cases:

# echo and then echo in a closure, redirecting the whole thing
do { echo a; echo b } | lines
# this would have printed a and then returned [b]
# now it just returns [b]

reasons to use `echo`:

echo foo | save foo.txt
# rather than
"foo" | save foo.txt

# it makes lists
echo a b c # ⇒ [a b c]

## `--` arguments when invoking `nu`

- argument quoting challenges
- let's not build our own command injection risk
- `--` as a consistent marker for no more special (flags) parsing?
- Should `nu -c` get special input?
- step by step, devyn taking the lead

## `plugin` family of commands (replacing `register`)
- [Plugin commands (`register` replacement) proposal](/lG8T5nTETaGuHO4GgFKK_w)
- Devyn: probably be able to keep `register` for 1 deprecated release
- `plugin add` would not have to be a hard parser keyword
- packaging/distribution of plugins
- nupm is designed with plugins in mind
- plugins can be written in any language / arbitrary binary
- uses a build script for nupm
- you could hide a download in here as well (no guaranteed signature validation through that)
- Jack?: pre-built distribution
- security/build-infra?
- target matrix to support (tier-1 targets and how to handle the oddbals)
- custom plugin files for:
- development
- overlays/custom virtualenvs

## `nupm`
- General chat on supply chain security/challenges providing sound builds
- worth comparing notes with
- [AUR](https://aur.archlinux.org/)
- cargo folks
- Darren: are we getting ready to include it with the standard library?
- Jakub: registry work to share packages

## `..........` or the ranges

1..2..10
# is going with a step 1
# as the syntax is indicating the **next** element

counting vs striding style

- counting: mathy continuation
- e.g. haskell enumeration syntax
- striding: define the step we take
- e.g. Python's slice operator/`range`

This doesn't currently work:

Tool for slicing

[1 2 3 4 5 6 7] | range 0..2..6
# expect ⇒ [1 3 5 7]