Add support for `conda init --condabin`, which adds `condabin/` to PATH #965

jaimergp · 2025-03-24T18:29:05Z

Description

Closes #960.

Based on #943 to avoid conflicts down the line because this adds new options to construct.yaml. Excuse the diff while 943 gets reviewed.

Windows:

macOS:

Checklist - did you ...

Add a file to the news directory (using the template) for the next release's release notes?
Add / update necessary tests?
Add / update outdated documentation?

…o pydantic-schema

jaimergp · 2025-06-16T10:52:28Z

@marcoesters, I think this is ready! It'd be nice to check if it behaves as intended locally too. Maybe I can extend the tests to actually check for PATH values but we are already testing that in the conda test suite 🤷

Also, I'd like to hear what you think about the proposed UI/text changes. I'd be happy to change whatever there, not feeling particularly strongly, but I do want to make it less confusing for end users.

marcoesters · 2025-06-16T16:53:10Z

@marcoesters, I think this is ready! It'd be nice to check if it behaves as intended locally too. Maybe I can extend the tests to actually check for PATH values but we are already testing that in the conda test suite 🤷

It would still be good to test that the installers implement this correctly.

Also, I'd like to hear what you think about the proposed UI/text changes. I'd be happy to change whatever there, not feeling particularly strongly, but I do want to make it less confusing for end users.

On Windows, I'm not sure it's obvious what the benefit of one option over the other is, so having both available is not a great idea. So, I propose to either have installer providers a) choose what they want /AddToPath to add to the path ($INSTDIR or condabin), or b) to allow /AddToPath to be disabled. If we choose a), we also need to decide on what the default value should be.

soapy1

Looks great! Just a couple of questions

soapy1 · 2025-06-16T17:43:07Z

CONSTRUCT.md

+### `add_condabin_to_path_default`
+
+Default value for the option added by `add_condabin_to_path`. The default
+is true for GUI installers (EXE, PKG) and false for shell installers. The user


what is the motivation behind making this default value different for different types of installers?

Consistency with previous UX. I assume the notion is that GUI users don't want to mess too much with the terminal, while shell installer users would be more ok with some tinkering. It's a bad proxy though.

I think consistency with previous UX is to not make it the default because in past UX, that option did not exist. If I updated constructor without making changes to the construct.yaml file, my installer would behave differently.

An installer built with an updated constructor wouldn't see any changes in behaviour because add_condabin_to_path defaults to false. This is the default value once add_condabin_to_path is true.

constructor/nsis/_nsis.py

jaimergp · 2025-06-17T08:14:26Z

On Windows, I'm not sure it's obvious what the benefit of one option over the other is, so having both available is not a great idea.

It's on the installer creator to offer both. By default, the condabin option is disabled. The installer creator has to opt-in. I would assume that at that point they will know what to do about the classic initialization option (disable it). I tried to discuss this in the docs.

a) choose what they want /AddToPath to add to the path ($INSTDIR or condabin)

They have two flags, which can be used together if needed, but it's on the installer creator to actually provide them.

b) to allow /AddToPath to be disabled

This will need a different PR, and it actually opened a can of worms. I'll open an issue to discuss it and then a potential fix.

jaimergp · 2025-06-17T09:18:43Z

I ended up opening these issues:

marcoesters · 2025-06-17T19:44:54Z

constructor/header.sh

+{%-   if add_condabin_to_path %}
+{%-     if add_condabin_to_path_default %}
+-z           do not add PREFIX/condabin to PATH; ignored in interactive mode
+{%-     else %}
+-c           add PREFIX/condabin to PATH; ignored in interactive mode
+{%-     endif %}
+{%-   endif %}


I am concerned about adding these additional, complementary flags here. #982 already tried to implement an extra flag for another option. If we add #1004 to the mix, I am concerned about maintainability.

The order of the flags is also getting a little chaotic. I wonder if we should keep the existing behavior of adding to path and initializing should be off in non-interactive mode unless opted in.

We still need the flags to opt-in so which ones do you recommend? We can also make it a single flag option like -c OPTION Whether to initialize conda in some way: shell, condabin, none. This would make initialize_conda and add_condabin_to_path mutually exclusive though. Do you think we should turn initialize_conda into an enum type instead of a bool directly?

Yes, we need flags to opt in. My concern is more to have two different flags for the opt-in/opt-out behavior. It would be ideal if we used long options, which are more explicit than -c. What do you think?

constructor/main.py

constructor/nsis/main.nsi.tmpl

constructor/shar.py

marcoesters · 2025-06-17T20:27:37Z

docs/source/howto.md

+inject some shell functions so you can run `conda` from anywhere. That is, the `conda` executable
+you usually run is not the Python entry point, but a shell function that ends up calling that entry
+point. This wrapper is needed by `conda (de)activate` to operate correctly, since it needs to


The sentence starting with "That is, ..." was a little confusing for me to read. Is it needed at all?

I think it's an important distinction to make, but happy to hear alternative wordings to make it clearer.

Suggested change

inject some shell functions so you can run `conda` from anywhere. That is, the `conda` executable

you usually run is not the Python entry point, but a shell function that ends up calling that entry

point. This wrapper is needed by `conda (de)activate` to operate correctly, since it needs to

inject some shell functions so you can run `conda` from anywhere. That is, the `conda` "executable"

you usually run is not the `conda` entry point, but a shell function that ends up calling the actual entry

point. This wrapper is needed by `conda (de)activate` to operate correctly, since it needs to

I think the "Python entry point", while technically correct, is coming a little out of nowhere. Maybe this?

docs/source/howto.md

marcoesters · 2025-06-17T20:33:05Z

CONSTRUCT.md

+### `add_condabin_to_path_default`
+
+Default value for the option added by `add_condabin_to_path`. The default
+is true for GUI installers (EXE, PKG) and false for shell installers. The user


I think consistency with previous UX is to not make it the default because in past UX, that option did not exist. If I updated constructor without making changes to the construct.yaml file, my installer would behave differently.

Co-authored-by: Marco Esters <[email protected]>

jaimergp added 30 commits February 28, 2025 16:16

Crude schema migration to Pydantic model and JSON Schema

800206f

add $schema path in examples

500f97c

Add pydantic to dependencies, small fixes

e3aa238

Refine types

1ecaa7b

Generate documentation from Pydantic model

4d870ec

Adjust documentation generation pipeline

0014436

Implement JSON Schema validation

4316140

Split Pydantic model to _schema.py and check up-to-date

8e995d5

Merge branch 'main' of github.com:conda/constructor into pydantic-schema

9dce9ed

Adjust hash description again

4285932

pre-commit

e479ccf

cleanup

3f59e6e

Check docs and schema with Python 3.12 to use StrEnum

1178425

Fix schema for extra_files

44ca08b

fix extra_envs schema

0e15add

Restore help message

5ff3828

add news

e4aadbf

Fix make_docs

3ea4e2e

Restore description for exclude

dc40fca

fix deprecations

da0dac7

pre-commit

d50560e

fully restore --help-construct

2c64a9d

pre-commit

4beb978

Update scripts/make_docs.py

f1ddc96

Use an enum for installer_type

ed88121

Merge branch 'pydantic-schema' of github.com:jaimergp/constructor int…

035ac75

…o pydantic-schema

docs sync

e894627

Add pattern for virtual_specs str

97df99b

Define schema for pkg_domains

ac29a2e

Allow several extra_files_* mappings per entry

7693368

jaimergp added 7 commits June 16, 2025 10:16

fix example

b2e80b4

check conda version

0aadf4b

false by default

69db7a9

Add choice box to PKG

6eba227

update schema

43e7352

Adjust options UI in Windows installers

7bdc7b4

add docs

3cbd090

jaimergp changed the title ~~(WIP) Add option to add condabin/ to PATH~~ Add support for conda init --condabin, which adds condabin/ to PATH Jun 16, 2025

jaimergp marked this pull request as ready for review June 16, 2025 10:50

jaimergp requested a review from a team as a code owner June 16, 2025 10:50

soapy1 reviewed Jun 16, 2025

View reviewed changes

jaimergp mentioned this pull request Jun 17, 2025

Make sure constructor's _system_path is also included conda/conda-standalone#148

Open

3 tasks

jaimergp added 2 commits June 17, 2025 10:26

prevent PATH manipulation in All Users mode here too

87bb9db

Only show CLI options if enabled

0c1e065

jaimergp added 7 commits June 17, 2025 13:09

update test

89f186b

add -c/-z condabin option, enable it in batch mode

53028b1

pre-commit

4059d10

posix

3d494be

fix if block

3b4c789

delayed expansion for windows

e58d0b5

fix if block again

1d6fd1c

marcoesters mentioned this pull request Jun 17, 2025

RFC: Undoing constructor / conda-standalone cross "bundling" #549

Open

marcoesters reviewed Jun 17, 2025

View reviewed changes

jaimergp and others added 2 commits June 18, 2025 10:23

Apply suggestions from code review

ee781f4

Co-authored-by: Marco Esters <[email protected]>

use default in help message

7900ba5

jaimergp mentioned this pull request Jun 18, 2025

Inconsistencies in initialize_conda / initialize_by_default handling across installers #1004

Open

2 tasks

Add support for conda init --condabin, which adds condabin/ to PATH #965

Are you sure you want to change the base?

Add support for conda init --condabin, which adds condabin/ to PATH #965

Uh oh!

Conversation

jaimergp commented Mar 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Checklist - did you ...

Uh oh!

jaimergp commented Jun 16, 2025

Uh oh!

marcoesters commented Jun 16, 2025

Uh oh!

soapy1 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jaimergp commented Jun 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jaimergp commented Jun 17, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jaimergp Jun 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Add support for `conda init --condabin`, which adds `condabin/` to PATH #965

Add support for `conda init --condabin`, which adds `condabin/` to PATH #965

jaimergp commented Mar 24, 2025 •

edited

Loading

jaimergp commented Jun 17, 2025 •

edited

Loading

jaimergp Jun 18, 2025 •

edited

Loading