Pipenv provides incorrect completion if current shell does not match login shell

It seems as though Pipenv does not correctly detect the fish shell when determining what to output for pipenv --completion. This is a regression that appeared within the last couple of weeks for me.

$ python -m pipenv.help output

Pipenv version: '11.8.0'

Pipenv location: '/home/qasim/.local/lib/python3.6/site-packages/pipenv'

Python location: '/usr/bin/python'

Other Python installations in PATH:

• 2.7: /usr/bin/python2.7

• 2.7: /usr/bin/python2.7

• 3.6: /usr/bin/python3.6m

• 3.6: /usr/bin/python3.6

• 3.6.4: /usr/bin/python

• 2.7.14: /usr/bin/python2

• 3.6.4: /usr/bin/python3

PEP 508 Information:

{'implementation_name': 'cpython',
 'implementation_version': '3.6.4',
 'os_name': 'posix',
 'platform_machine': 'x86_64',
 'platform_python_implementation': 'CPython',
 'platform_release': '4.15.9-1-ARCH',
 'platform_system': 'Linux',
 'platform_version': '#1 SMP PREEMPT Sun Mar 11 17:54:33 UTC 2018',
 'python_full_version': '3.6.4',
 'python_version': '3.6',
 'sys_platform': 'linux'}

System environment variables:

• COLORTERM
• DBUS_SESSION_BUS_ADDRESS
• DESKTOP_AUTOSTART_ID
• DESKTOP_SESSION
• DISPLAY
• GDMSESSION
• GDM_LANG
• GNOME_DESKTOP_SESSION_ID
• HOME
• LANG
• LOGNAME
• MAIL
• MOZ_PLUGIN_PATH
• PATH
• PWD
• SESSION_MANAGER
• SHELL
• SHLVL
• SSH_AUTH_SOCK
• TERM
• USER
• USERNAME
• VTE_VERSION
• WINDOWID
• WINDOWPATH
• XAUTHORITY
• XDG_CURRENT_DESKTOP
• XDG_MENU_PREFIX
• XDG_RUNTIME_DIR
• XDG_SEAT
• XDG_SESSION_DESKTOP
• XDG_SESSION_ID
• XDG_SESSION_TYPE
• XDG_VTNR
• PIP_PYTHON_PATH
• PYTHONUNBUFFERED

Pipenv–specific environment variables:

Debug–specific environment variables:

• PATH: /usr/local/bin:/usr/local/sbin:/usr/bin:/usr/lib/jvm/default/bin:/usr/bin/site_perl:/usr/bin/vendor_perl:/usr/bin/core_perl:/home/qasim/.local/bin
• SHELL: /bin/bash
• LANG: en_GB.UTF-8
• PWD: /home/qasim

---

Expected result

I’d expect the completion to complete without errors 😃

Actual result

When attempting to auto-complete on the first Pipenv command:

qasim@qasim-desktop ~> pipenv - (line 1): Illegal command name “_pipenv_completion()”
begin;  _pipenv_completion() {     local IFS=$'\t'     COMPREPLY=( $( env COMP_WORDS="${COMP_WORDS[*]}" \                    COMP_CWORD=$COMP_CWORD \                    _PIPENV_COMPLETE=complete-bash $1 ) )     return 0 }  complete -F _pipenv_completion -o default pipenv 
        ^
from sourcing file -
	called on line 60 of file /usr/share/fish/functions/eval.fish

in function “eval”
	called on line 1 of file ~/.config/fish/completions/pipenv.fish

from sourcing file ~/.config/fish/completions/pipenv.fish
	called on standard input

in command substitution
	called on standard input

source: Error while reading file “-”

It seems as though this is the bash code being output from <???> rather than the fish code.

Steps to replicate

1. pip install --user --upgrade pipenv
2. Setup fish completions inside ~/.config/fish/completions/pipenv.fish: eval (pipenv --completion)
3. Enter fish and type Pipenv and then try to auto-complete something

[techalchemy]

Thanks for the offer @anowlcalledjosh — our current stance is that we aren’t using psutil in pipenv. We literally just removed it and had to vendor pew and write c extension bindings for windows. PSUtil is slow and it breaks regularly for windows users, we just aren’t going to use it. We handle enough special edge cases, if we can’t trust your environment we aren’t really able to be responsible for maintaining logic to handle edge cases.

If there’s an approach that works around this, I would be ok with that. But in my mind the easiest approach is to set your shell variable

[tsiq-oliver]

How else would you suggest we get the users shell on Linux?

Proof of concept:

>>> import os
>>> import psutil
>>> print(psutil.Process(os.getppid()).exe())
/bin/zsh

Obviously, doesn’t work in scenarios where there’s an intermediate parent.