feat: Dependency tree command #1069
Conversation
|
He @abkfenris, thanks for the PR. This is already a very good start, I gave it a go and it does most of what I expect from a first version. To improve and align with the The It would be nice if you could also write some documentation. I'll do a technical review later! |
|
In the help text I assume this will require a package/multiple packages as another argument, otherwise it will fail? So maybe make it |
|
Having the direct dependencies styled is especially nice for inverted. 
The package that invert is matching against comes from the optional positional argument. I added a note for that, and figured out how to make clap mark it as a dependency, but it doesn't show in quite that form. pixi tree --help
Lock file usage from the CLI
Usage: pixi tree [OPTIONS] [REGEX]
Arguments:
[REGEX] List only packages matching a regular expression
Options:
--platform <PLATFORM> The platform to list packages for. Defaults to the current platform
--manifest-path <MANIFEST_PATH> The path to 'pixi.toml' [env: PIXI_PROJECT_MANIFEST=]
-e, --environment <ENVIRONMENT> The environment to list packages for. Defaults to the default environment
--frozen Don't check or update the lockfile, continue with previously installed environment
[env: PIXI_FROZEN=]
--locked Check if lockfile is up to date, aborts when lockfile isn't up to date with the
manifest file [env: PIXI_LOCKED=]
--no-install Don't install the environment for pypi solving, only update the lock-file if it can
solve without installing
-i, --invert Invert tree and show what depends on given package in the regex argument
-v, --verbose... Increase logging verbosity
-q, --quiet... Decrease logging verbosity
--color <COLOR> Whether the log needs to be colored [env: PIXI_COLOR=] [default: auto] [possible
values: always, never, auto]
-h, --help Print help |
fixes: prefix-dev#238 Adds a tree command for the CLI that can print out the tree of dependencies based on those defined in pixi.toml. Additionally it can filter the regular dependency tree, or show what tree of packages need a given package. It keeps track of visits for the regular tree, but currently does not for the inverted one.
|
Ok, I went and quickly rebased it off of main to split it off from my tinkering with |
Does not fully understand normalized names and markers
|
I taught it how to 
I've also generally cleaned up the tree code, and refactored it to standardize functionality. I'm not however dealing with markers/extras, or always correctly figuring out normalized package names (xpublish-edr is a directly specified dependency, so it should be highlighted). For markers/extras, right now I'm just discarding those dependencies. |
|
I think the error with |
|
see #1074 |
Ah, I missed that. |
Co-authored-by: Ruben Arts <[email protected]>
Co-authored-by: Pavel Zwerschke <[email protected]> Co-authored-by: Bas Zalmstra <[email protected]>
|
Thanks for taking a look everyone.
Is there an example that you can point me too? My messier test environment seems to be getting things right (Xarray from Conda, Xpublish* local PyPI). 
|
|
I've had the chance to already play around with What I noticed is that in some cases it doesn't behave as it should(?) [project]
name = "tmp.yqbLiIBFp9"
channels = ["conda-forge"]
platforms = ["osx-arm64", "linux-64"]
[tasks]
[dependencies]
mlflow = "1.27.0"For some reason, pandas and pandas-profiling are not part of the tree 🤔 In |
|
Hmm, I'd have to try with that environment to be sure, but have you tried the latest commit? I had some issues with normalization for a few which meant that things were getting dropped. |
|
i thought that i used the latest commit but i'm not sure anymore... i'll try again after lunch 🙌🏻 |
|
I can reproduce it on my machine @pavelzw |
|
I also just tested again with |
|
All dependencies without a version in the "dependencies" of the package are missing from the tree. So also |
Thanks! I think I know right where that would be coming from. I'll tweak that and test it after I grab breakfast. |
|
Ah sorry I now read this, but I hope you don't mind me fixing it. I also renamed parts of the function as it was hard to grasp the variables at first. Let me know what you think. |
That works for me. I don't have that great of a handle on all the possible iterators, so I was gonna slap an |
Talking about real-world scenarious, I ended up using |
|
|
|
Another thing I stumbled upon: when running The error message is good but there might be use cases where you want to only have a sub-tree of your whole tree. not a deal-breaker though and i'm really happy how his turned out 🚀 |
|
@pavelzw that is a good point, lets get this in and fix that in a following PR, as I also can't wait for this PR. |
|
I took a quick swing at matching transitive dependencies, but I can push that to a new PR rather than hold this one on another round of debugging. 
|
|
Go ahead @abkfenris 👍 |
When a regex pattern is given to `pixi tree` and a top level dependency isn't matched, it currently errors rather than showing any matching transitive dependencies. This prints transitive dependencies in separate trees, similar to `pixi tree -i <package_name>`. Builds on prefix-dev#1069
When a regex pattern is given to `pixi tree` and a top level dependency isn't matched, it currently errors rather than showing any matching transitive dependencies. This prints transitive dependencies in separate trees, similar to `pixi tree -i <package_name>`. Builds on prefix-dev#1069
|
Ok, draft of transitive matching over on #1111 |
fixes: #238
A really rough pass at a
pixi treecommand, based off ofcargo treeandmamba repoquery.I had an idea that I didn't actually expect to work that I started tinkering around while working on another branch ( #1055 ), so it isn't very clean, but since folks are interested in playing with it.
It does:
It does not or is not:
Know how to start with PyPI dependenciesShow versions on an inverted treepixi listand kept tweaking till something worked.Full tree:
Filtered tree
pixi tree pre-commit └── pre-commit v3.3.3 ├── cfgv v3.3.1 │ └── python v3.12.2 │ ├── bzip2 v1.0.8 │ ├── libexpat v2.6.2 │ ├── libffi v3.4.2 │ ├── libsqlite v3.45.2 │ │ └── libzlib v1.2.13 │ ├── libzlib v1.2.13 (*) │ ├── ncurses v6.4.20240210 │ ├── openssl v3.2.1 │ ├── readline v8.2 │ │ └── ncurses v6.4.20240210 (*) │ ├── tk v8.6.13 │ │ └── libzlib v1.2.13 (*) │ └── xz v5.2.6 ├── identify v2.5.35 │ └── python v3.12.2 (*) ├── nodeenv v1.8.0 │ └── python v3.12.2 (*) ├── python v3.12.2 (*) ├── pyyaml v6.0.1 │ ├── python v3.12.2 (*) │ ├── python v3.12.2 (*) │ ├── python_abi v3.12 │ └── yaml v0.2.5 └── virtualenv v20.25.1 ├── distlib v0.3.8 │ └── python v3.12.2 (*) ├── filelock v3.13.1 │ └── python v3.12.2 (*) ├── platformdirs v4.2.0 │ └── python v3.12.2 (*) └── python v3.12.2 (*)Inverted
pixi tree -i yaml ruamel.yaml ├── cffconvert └── pykwalify └── cffconvert ruamel.yaml.clib └── ruamel.yaml ├── cffconvert └── pykwalify └── cffconvert pyyaml └── pre-commit yaml └── pyyaml └── pre-commitHelp
pixi tree --help Lock file usage from the CLI Usage: pixi tree [OPTIONS] [REGEX] Arguments: [REGEX] List only packages matching a regular expression Options: --platform <PLATFORM> The platform to list packages for. Defaults to the current platform --manifest-path <MANIFEST_PATH> The path to 'pixi.toml' [env: PIXI_PROJECT_MANIFEST=] -e, --environment <ENVIRONMENT> The environment to list packages for. Defaults to the default environment --frozen Don't check or update the lockfile, continue with previously installed environment [env: PIXI_FROZEN=] --locked Check if lockfile is up to date, aborts when lockfile isn't up to date with the manifest file [env: PIXI_LOCKED=] --no-install Don't install the environment for pypi solving, only update the lock-file if it can solve without installing -i, --invert Invert tree and show what depends on given package -v, --verbose... Increase logging verbosity -q, --quiet... Decrease logging verbosity --color <COLOR> Whether the log needs to be colored [env: PIXI_COLOR=] [default: auto] [possible values: always, never, auto] -h, --help Print help