fix(nushell): add missing | parse env | update-env for deactivation operations

[jokeyrhyme]

• addresses the unexpected behaviour discussed here: `mise activate nu` emits invalid code #6876
• previously, work was done to have mise reverse the changes it made to the environment variables: fix(activate): reset PATH when activate is called multiple times #6829
• however, for nushell, this resulted in the raw operations CSV text becoming part of the activation script, which is not valid nushell syntax
• this PR wraps the raw operations CSV text in quotes to make it a string literal, and then passes it through | parse env | update-env just as with other places in the nushell script that make use of the raw operations CSV text
• in order for that work, the parse env and update-env functions were re-ordered to always be above the deactivation statements

[theutz]

Thank you for doing this! My love for both Nushell and Mise has made me yearn to learn Rust so I can help out, but I haven't had a chance yet. :D

[jdx]

would you mind adding nushell to either the coverage github action tests or mise.toml so they actually get run in CI? these tests otherwise won't do any good

[jokeyrhyme]

sure thing!

[jokeyrhyme]

huh, any special reason why the Windows end-to-end tests were automated via GitHub Actions, but not the rest of the end-to-end tests?

edit: ignore me, i found them in the "nightly" Action, although they don't seem to log any output to stdout which is intriguing ...

[jokeyrhyme]

mise run e2e --list works on my machine and outputs the full list to stdout

i'm not sure mise run e2e (what is already in "nightly") or even mise run e2e --all actually does anything at the moment ...

[jokeyrhyme]

@jdx well, i've explicitly added the nushell end-to-end tests in "nightly", but i think a follow-up PR is needed to fix the rest of the end-to-end tests, unless you're okay with only the nushell ones running? 🤷

[jdx]

that biome test failure is unrelated. I am going to try to fix it tomorrow

[jokeyrhyme]

oh, hang on, all of the end-to-end tests, including the nushell ones, are spread out across the "coverage" Actions, so explicitly adding the nushell tests to the "nightly" Action wasn't necessary

(but i do need to ensure that nu is available in the "coverage" Actions)

[jokeyrhyme]

yay, for fa24ac2 :

• https://github.com/jdx/mise/actions/runs/19454064247/job/55664635989?pr=6994

E2E shell/test_nushell
  removed 'e2e_nushell_init.Beo2r5iJkQ'
  shell/test_nushell: 0s

• https://github.com/jdx/mise/actions/runs/19454064247/job/55664635978?pr=6994

E2E shell/test_nushell_deactivation
  mise WARN  Failed to deserialize __MISE_DIFF: Invalid input length: 1
  removed 'e2e_nushell_init.DP4HhGPMjm'
  shell/test_nushell_deactivation: 0s

[theutz]

I just downloaded this and tried to use it, but it's giving a "Parse mismatch during operation" error, pointing to the definition for parse vars, and then saying "expected not a pipeline". So, it's actually more broken than before, and the old workaround doesn't seem to fix it.

[jokeyrhyme]

@theutz eek! i'm not getting anything like that on this machine, but it might be worth reverting this PR out of caution ...

which version of nushell do you have?

what is the full output of ^mise activate nu for you?

[theutz]

I'm using 0.108.0.

Error Message

Error: nu::parser::parse_mismatch
× Parse mismatch during operation.

╭─[/Users/michael/.local/share/nushell/vendor/autoload/mise.nu:1:1]

1 │ ╭─▶ def "parse vars" [] {

2 │ │     $in | from csv --noheaders --no-infer | rename 'op' 'name' 'value'

3 │ │   }

4 │ │

5 │ │   def --env "update-env" [] {

6 │ │     for $var in $in {

7 │ │       if $var.op == "set" {

8 │ │         if ($var.name | str upcase) == 'PATH' {

9 │ │           $env.PATH = ($var.value | split row (char esep))

10 │ │         } else {

11 │ │           load-env {($var.name): $var.value}

12 │ │         }

13 │ │       } else if $var.op == "hide" and $var.name in $env {

14 │ │         hide-env $var.name

15 │ │       }

16 │ │     }

17 │ │   }

18 │ │   "set,PATH,/opt/homebrew/sbin:/opt/homebrew/bin:/Users/michael/.local/bin:/Users/michael/bin:/usr/local/bin:/Users/michael/.config/composer/vendor/bin:/Users/michael/.config/carapace/bin:/usr/bin:/bin:/usr/sbin:/sbin

19 │ │   hide,MISE_SHELL,

20 │ │   hide,__MISE_DIFF,

21 │ │   hide,__MISE_DIFF,

22 │ │   " | parse vars | update-env

23 │ │   export-env {

24 │ │

25 │ │     $env.MISE_SHELL = "nu"

26 │ │     let mise_hook = {

27 │ │       condition: { "MISE_SHELL" in $env }

28 │ │       code: { mise_hook }

29 │ │     }

30 │ │     add-hook hooks.pre_prompt $mise_hook

31 │ │     add-hook hooks.env_change.PWD $mise_hook

32 │ │   }

33 │ │

34 │ │   def --env add-hook [field: cell-path new_hook: any] {

35 │ │     let field = $field | split cell-path | update optional true | into cell-path

36 │ │     let old_config = $env.config? | default {}

37 │ │     let old_hooks = $old_config | get $field | default []

38 │ │     $env.config = ($old_config | upsert $field ($old_hooks ++ [$new_hook]))

39 │ │   }

40 │ │

41 │ │   export def --env --wrapped main [command?: string, --help, ...rest: string] {

42 │ │     let commands = ["deactivate", "shell", "sh"]

43 │ │

44 │ │     if ($command == null) {

45 │ │       ^"/Users/michael/.local/bin/mise"

46 │ │     } else if ($command == "activate") {

47 │ │       $env.MISE_SHELL = "nu"

48 │ │     } else if ($command in $commands) {

49 │ │       ^"/Users/michael/.local/bin/mise" $command ...$rest

50 │ │       | parse vars

51 │ │       | update-env

52 │ │     } else {

53 │ │       ^"/Users/michael/.local/bin/mise" $command ...$rest

54 │ │     }

55 │ │   }

56 │ │

57 │ │   def --env mise_hook [] {

58 │ │     ^"/Users/michael/.local/bin/mise" hook-env -s nu

59 │ │       | parse vars

60 │ │       | update-env

61 │ │   }

62 │ ├─▶

· ╰──── expected not a pipeline

╰────

Output from ^mise activate nu

def "parse vars" [] {

$in | from csv --noheaders --no-infer | rename 'op' 'name' 'value'

}
def --env "update-env" [] {

for $var in $in {

if $var.op == "set" {

if ($var.name | str upcase) == 'PATH' {

$env.PATH = ($var.value | split row (char esep))

} else {

load-env {($var.name): $var.value}

}

} else if $var.op == "hide" and $var.name in $env {

hide-env $var.name

}

}

}

"set,PATH,/opt/homebrew/sbin:/opt/homebrew/bin:/Users/michael/.local/bin:/Users/michael/bin:/usr/local/bin:/Users/michael/.config/composer/vendor/bin:/Users/michael/.config/carapace/bin:/usr/bin:/bin:/usr/sbin:/sbin

hide,MISE_SHELL,

hide,__MISE_DIFF,

hide,__MISE_DIFF,

" | parse vars | update-env

export-env {
$env.MISE_SHELL = "nu"

let mise_hook = {

condition: { "MISE_SHELL" in $env }

code: { mise_hook }

}

add-hook hooks.pre_prompt $mise_hook

add-hook hooks.env_change.PWD $mise_hook

}
def --env add-hook [field: cell-path new_hook: any] {

let field = $field | split cell-path | update optional true | into cell-path

let old_config = $env.config? | default {}

let old_hooks = $old_config | get $field | default []

$env.config = ($old_config | upsert $field ($old_hooks ++ [$new_hook]))

}
export def --env --wrapped main [command?: string, --help, ...rest: string] {

let commands = ["deactivate", "shell", "sh"]
if ($command == null) {

^"/Users/michael/.local/bin/mise"

} else if ($command == "activate") {

$env.MISE_SHELL = "nu"

} else if ($command in $commands) {

^"/Users/michael/.local/bin/mise" $command ...$rest

| parse vars

| update-env

} else {

^"/Users/michael/.local/bin/mise" $command ...$rest

}

}
def --env mise_hook [] {

^"/Users/michael/.local/bin/mise" hook-env -s nu

| parse vars

| update-env

}

[jokeyrhyme]

very mysterious

I'm not having any of that on my machine

and nushell in the coverage tests does source this and doesn't complain, so what is different here ... ?

[kdar]

I'm also having this same issue with the latest nushell and mise on Windows.

nushell:

╭─────────┬─────────╮
│ channel │ release │
│ current │ true    │
│ latest  │ 0.108.0 │
╰─────────┴─────────╯

mise:

2025.11.6 windows-x64 (2025-11-18)

[jokeyrhyme]

wait, Windows? @theutz are you also on Windows?
I wonder if this is a line ending thing ...
i personally tested this change during development on macOS and Linux, but not Windows ...

[theutz]

I'm not using Windows, no. I'm on macOS Tahoe.

[theutz]

Yeah, I don't know where those errors would be coming from. I've gone over the nushell code, and it doesn't make any sense. I'm not an expert in nu, but nothing seems too out of place to me.

I guess the only thing that might not make sense to me is the presence of these lines in the activate output:

"set,PATH,/opt/homebrew/sbin:/opt/homebrew/bin:/Users/michael/.local/bin:/Users/michael/bin:/usr/local/bin:/Users/michael/.config/composer/vendor/bin:/Users/michael/.config/carapace/bin:/usr/bin:/bin:/usr/sbin:/sbin
hide,MISE_SHELL,
hide,__MISE_DIFF,
hide,__MISE_DIFF,
" | parse vars | update-env

If we're treating the output of this command as a module, should that code be beneath the export-env line?

[jokeyrhyme]

i find nushell can occasionally be wildly confused about the origin coordinates of an error

there was only one export-env { ... } in the original output of mise active nu before this PR, and this PR doesn't change what does and doesn't happen there

it'd be interesting to selectively delete/change parts of your output from mise active nu to try to narrow down exactly which part is making it explode when it gets sourced

[theutz]

Good idea! So, I tried commenting everything out, and the culprit seems to be fact that line. If the script is used instead of sourced, I think it might not be proper semantics to write a pipeline in the top-level of the module like that. I think it has to go in the export-env. And, indeed, when I manually move it there, it works.

[jokeyrhyme]

oh! i use source instead of use, even though the mise documentation says to use: https://mise.jdx.dev/installing-mise.html#nushell

that's the difference! if i replace source ~/.cache/init/mise.nu with use ~/.cache/init/mise.nu then i can reproduce the error

i think because the atuin documentation says source ( https://docs.atuin.sh/guide/installation/#installing-the-shell-plugin ) i guess maybe copied that approach for mise in my own settings

okay, excellent, i can reproduce locally and hence test locally and hence raise a PR to fix this confidently :)

[jokeyrhyme]

here we go: #7013

[theutz]

Thank you, sir!