Strategy for Answering
Exactly what the conda init
command does and its consequences are shell-specific. Instead of trying to cover all cases, let's walk through a case, noting along the way that one can replicate this analysis by substituting their shell of interest.
Case Study: conda init zsh
Let's look at zsh
as the shell. This is a common shell (default for macOS 10.15+) and very close to bash. Plus, I don't already have it configured.
Probing the Command: Dry Run
Many Conda commands include some form of dry run functionality via a --dry-run, -d
flag, which - combined with verbosity flags - enables seeing what this would do without doing them. For the init
command, dry run alone will only tell us what files it would modify:
$ conda init -d zsh
no change /Users/mfansler/miniconda3/condabin/conda
no change /Users/mfansler/miniconda3/bin/conda
no change /Users/mfansler/miniconda3/bin/conda-env
no change /Users/mfansler/miniconda3/bin/activate
no change /Users/mfansler/miniconda3/bin/deactivate
no change /Users/mfansler/miniconda3/etc/profile.d/conda.sh
no change /Users/mfansler/miniconda3/etc/fish/conf.d/conda.fish
no change /Users/mfansler/miniconda3/shell/condabin/Conda.psm1
no change /Users/mfansler/miniconda3/shell/condabin/conda-hook.ps1
no change /Users/mfansler/miniconda3/lib/python3.7/site-packages/xontrib/conda.xsh
no change /Users/mfansler/miniconda3/etc/profile.d/conda.csh
modified /Users/mfansler/.zshrc
==> For changes to take effect, close and re-open your current shell. <==
Here we can see that it plans to target the user-level resources file for zsh, /Users/mfansler/.zshrc
, but it doesn't tell us how it will modified it. Also, OMG! the UX here is awful, because it in no way reflects the fact that I used the -d
flag. But don't worry: as long as the -d
flag is there it won't actually change things.
Patch Preview
To see what exactly it will do, add a single verbosity flag (-v
) to the command. This will give everything from the previous output, but will now shows us the diff it will use to patch (update) the .zshrc
file.
$ conda init -dv zsh
/Users/mfansler/.zshrc
---
+++
@@ -0,0 +1,16 @@
+
+# >>> conda initialize >>>
+# !! Contents within this block are managed by 'conda init' !!
+__conda_setup="$('/Users/mfansler/miniconda3/bin/conda' 'shell.zsh' 'hook' 2> /dev/null)"
+if [ $? -eq 0 ]; then
+ eval "$__conda_setup"
+else
+ if [ -f "/Users/mfansler/miniconda3/etc/profile.d/conda.sh" ]; then
+ . "/Users/mfansler/miniconda3/etc/profile.d/conda.sh"
+ else
+ export PATH="/Users/mfansler/miniconda3/bin:$PATH"
+ fi
+fi
+unset __conda_setup
+# <<< conda initialize <<<
+
# ...the rest is exactly as above
That is, the plan of action is to add these 16 lines to the .zshrc
file. In this case, I don't have an existing .zshrc
file, so it plans to add it at line 1. If the file had already existed, it would append these lines.
Interpreting the Shell Code
Let's overview this code, before focusing on the details. Essentially, this is a redundant sequence of attempts to set up some shell functionality. They are ordered from most to least functional.
What Conda Hopes To Do
The code
__conda_setup="$('/Users/mfansler/miniconda3/bin/conda' 'shell.zsh' 'hook' 2> /dev/null)"
if [ $? -eq 0 ]; then
eval "$__conda_setup"
gets something from conda
itself, storing the result to a string, and then evaluates that string if the command had a clean exit ($? -eq 0
). The neat engineering here is that the subprocess (technically python -m conda
) passes back a result that can be run within this current process (zsh
), allowing it to define shell functions.
I'll dig deeper into what is going on here in a second.
Fallback 1: Hardcoded Shell Functions
If that strange internal command fails, the devs included a hardcoded version of some essential shell functions (specifically conda activate
). This is in:
miniconda3/etc/profile.d/conda.sh
and they simply check the file exists and source it. Let's hit that last option, then we'll swing back to look at the functionality.
Fallback 2: The Last Resort
The absolute last resort is to literally violate the standing recommendation since Conda v4.4, which is to simply put the base environment's bin
directory on PATH
. In this case, there is no conda activate
functionality; this only ensures that Conda is on your PATH.
Details: Shell Functionality
Coming back to the intended case, we can inspect exactly what it would evaluate by simply getting that string result:
$ conda shell.zsh hook
__add_sys_prefix_to_path() {
# In dev-mode CONDA_EXE is python.exe and on Windows
# it is in a different relative location to condabin.
if [ -n "${_CE_CONDA}" ] && [ -n "${WINDIR+x}" ]; then
SYSP=$(\dirname "${CONDA_EXE}")
else
SYSP=$(\dirname "${CONDA_EXE}")
SYSP=$(\dirname "${SYSP}")
fi
if [ -n "${WINDIR+x}" ]; then
PATH="${SYSP}/bin:${PATH}"
PATH="${SYSP}/Scripts:${PATH}"
PATH="${SYSP}/Library/bin:${PATH}"
PATH="${SYSP}/Library/usr/bin:${PATH}"
PATH="${SYSP}/Library/mingw-w64/bin:${PATH}"
PATH="${SYSP}:${PATH}"
else
PATH="${SYSP}/bin:${PATH}"
fi
\export PATH
}
__conda_exe() (
__add_sys_prefix_to_path
"$CONDA_EXE" $_CE_M $_CE_CONDA "$@"
)
__conda_hashr() {
if [ -n "${ZSH_VERSION:+x}" ]; then
\rehash
elif [ -n "${POSH_VERSION:+x}" ]; then
: # pass
else
\hash -r
fi
}
__conda_activate() {
if [ -n "${CONDA_PS1_BACKUP:+x}" ]; then
# Handle transition from shell activated with conda <= 4.3 to a subsequent activation
# after conda updated to >= 4.4. See issue #6173.
PS1="$CONDA_PS1_BACKUP"
\unset CONDA_PS1_BACKUP
fi
\local ask_conda
ask_conda="$(PS1="${PS1:-}" __conda_exe shell.posix "$@")" || \return
\eval "$ask_conda"
__conda_hashr
}
__conda_reactivate() {
\local ask_conda
ask_conda="$(PS1="${PS1:-}" __conda_exe shell.posix reactivate)" || \return
\eval "$ask_conda"
__conda_hashr
}
conda() {
\local cmd="${1-__missing__}"
case "$cmd" in
activate|deactivate)
__conda_activate "$@"
;;
install|update|upgrade|remove|uninstall)
__conda_exe "$@" || \return
__conda_reactivate
;;
*)
__conda_exe "$@"
;;
esac
}
if [ -z "${CONDA_SHLVL+x}" ]; then
\export CONDA_SHLVL=0
# In dev-mode CONDA_EXE is python.exe and on Windows
# it is in a different relative location to condabin.
if [ -n "${_CE_CONDA:+x}" ] && [ -n "${WINDIR+x}" ]; then
PATH="$(\dirname "$CONDA_EXE")/condabin${PATH:+":${PATH}"}"
else
PATH="$(\dirname "$(\dirname "$CONDA_EXE")")/condabin${PATH:+":${PATH}"}"
fi
\export PATH
# We're not allowing PS1 to be unbound. It must at least be set.
# However, we're not exporting it, which can cause problems when starting a second shell
# via a first shell (i.e. starting zsh from bash).
if [ -z "${PS1+x}" ]; then
PS1=
fi
fi
conda activate base
I'm not going to walk through all this, but the main part is that instead of directly putting bin
on PATH, it defines a shell function called conda
and this serves as a wrapper for the condabin/conda
entrypoint. This also defines a new functionality conda activate
, which uses a shell function, __conda_activate()
, behind the scenes. At the final step, it then activates the base environment.
Why do it this way?
This is engineered like this in order to be responsive to the configuration settings. Configuration options like auto_activate_base
and change_ps1
affect how Conda manipulates the shell, and so that changes what functionality Conda includes in its shell functions.
Does Conda "Pollute the Environment"?
Not really. The main behavioral things like auto-activation and prompt modification can be disabled through configuration settings, so that conda init
ultimately just adds the conda activate
function to the shell, enabling clean switching between environments without ever having to manually manipulate PATH.