pipdeptree

pipdeptree is a command line utility for displaying the installed python packages in form of a dependency tree. It works for packages installed globally on a machine as well as in a virtualenv. Since pip freeze shows all dependencies as a flat list, finding out which are the top level packages and which packages do they depend on requires some effort. It's also tedious to resolve conflicting dependencies that could have been installed because older version of pip didn't have true dependency resolution [1]. pipdeptree can help here by identifying conflicting dependencies installed in the environment.

To some extent, pipdeptree is inspired by the lein deps :tree command of Leiningen.

Installation

$ pip install pipdeptree

pipdeptree has been tested with Python versions 2.7, 3.5, 3.6, 3.7, 3.8, 3.9 as well as pypy2 and pypy3.

Python 2.6 is way past it's end of life but if you ever find yourself stuck on a legacy environment, version 0.9.0 might work.

Running in virtualenvs

New in ver. 2.0.0

If you want to run pipdeptree in the context of a particular virtualenv, you can specify the --python option. Note that this capability has been recently added in version 2.0.0.

Alternately, you may also install pipdeptree inside the virtualenv and then run it from there.

Usage and examples

To give you a brief idea, here is the output of pipdeptree compared with pip freeze:

$ pip freeze
Flask==0.10.1
itsdangerous==0.24
Jinja2==2.11.2
-e git+git@github.com:naiquevin/lookupy.git@cdbe30c160e1c29802df75e145ea4ad903c05386#egg=Lookupy
MarkupSafe==0.22
pipdeptree @ file:///private/tmp/pipdeptree-2.0.0b1-py3-none-any.whl
Werkzeug==0.11.2

And now see what pipdeptree outputs,

$ pipdeptree
Warning!!! Possibly conflicting dependencies found:
* Jinja2==2.11.2
 - MarkupSafe [required: >=0.23, installed: 0.22]
------------------------------------------------------------------------
Flask==0.10.1
  - itsdangerous [required: >=0.21, installed: 0.24]
  - Jinja2 [required: >=2.4, installed: 2.11.2]
    - MarkupSafe [required: >=0.23, installed: 0.22]
  - Werkzeug [required: >=0.7, installed: 0.11.2]
Lookupy==0.1
pipdeptree==2.0.0b1
  - pip [required: >=6.0.0, installed: 20.1.1]
setuptools==47.1.1
wheel==0.34.2

Is it possible to find out why a particular package is installed?

New in ver. 0.5.0

Yes, there's a --reverse (or simply -r) flag for this. To find out which packages depend on a particular package(s), it can be combined with --packages option as follows:

$ pipdeptree --reverse --packages itsdangerous,MarkupSafe
Warning!!! Possibly conflicting dependencies found:
* Jinja2==2.11.2
 - MarkupSafe [required: >=0.23, installed: 0.22]
------------------------------------------------------------------------
itsdangerous==0.24
  - Flask==0.10.1 [requires: itsdangerous>=0.21]
MarkupSafe==0.22
  - Jinja2==2.11.2 [requires: MarkupSafe>=0.23]
    - Flask==0.10.1 [requires: Jinja2>=2.4]

What's with the warning about conflicting dependencies?

As seen in the above output, pipdeptree by default warns about possible conflicting dependencies. Any package that's specified as a dependency of multiple packages with different versions is considered as a conflicting dependency. Conflicting dependencies are possible if older version of pip<=20.2 (without the new resolver [1]) was ever used to install dependencies at some point. The warning is printed to stderr instead of stdout and it can be completely silenced by specifying the -w silence or --warn silence option. On the other hand, it can be made mode strict with --warn fail, in which case the command will not only print the warnings to stderr but also exit with a non-zero status code. This is useful if you want to fit this tool into your CI pipeline.

Note: The --warn option is added in version 0.6.0. If you are using an older version, use --nowarn flag to silence the warnings.

Warnings about circular dependencies

In case any of the packages have circular dependencies (eg. package A depends on package B and package B depends on package A), then pipdeptree will print warnings about that as well.

$ pipdeptree --exclude pip,pipdeptree,setuptools,wheel
Warning!!! Cyclic dependencies found:
- CircularDependencyA => CircularDependencyB => CircularDependencyA
- CircularDependencyB => CircularDependencyA => CircularDependencyB
------------------------------------------------------------------------
wsgiref==0.1.2
argparse==1.2.1

Similar to the warnings about conflicting dependencies, these too are printed to stderr and can be controlled using the --warn option.

In the above example, you can also see --exclude option which is the opposite of --packages ie. these packages will be excluded from the output.

Using pipdeptree to write requirements.txt file

If you wish to track only top level packages in your requirements.txt file, it's possible by grep-ing [2]. only the top-level lines from the output,

$ pipdeptree --warn silence | grep -E '^\w+'
Flask==0.10.1
gnureadline==8.0.0
Lookupy==0.1
pipdeptree==2.0.0b1
setuptools==47.1.1
wheel==0.34.2

There is a problem here though - The output doesn't mention anything about Lookupy being installed as an editable package (refer to the output of pip freeze above) and information about its source is lost. To fix this, pipdeptree must be run with a -f or --freeze flag.

$ pipdeptree -f --warn silence | grep -E '^[a-zA-Z0-9\-]+'
Flask==0.10.1
gnureadline==8.0.0
-e git+git@github.com:naiquevin/lookupy.git@cdbe30c160e1c29802df75e145ea4ad903c05386#egg=Lookupy
pipdeptree @ file:///private/tmp/pipdeptree-2.0.0b1-py3-none-any.whl
setuptools==47.1.1
wheel==0.34.2

$ pipdeptree -f --warn silence | grep -E '^[a-zA-Z0-9\-]+' > requirements.txt

The freeze flag will not prefix child dependencies with hyphens, so you could dump the entire output of pipdeptree -f to the requirements.txt file thus making it human-friendly (due to indentations) as well as pip-friendly.

$ pipdeptree -f | tee locked-requirements.txt
Flask==0.10.1
  itsdangerous==0.24
  Jinja2==2.11.2
    MarkupSafe==0.23
  Werkzeug==0.11.2
gnureadline==8.0.0
-e git+git@github.com:naiquevin/lookupy.git@cdbe30c160e1c29802df75e145ea4ad903c05386#egg=Lookupy
pipdeptree @ file:///private/tmp/pipdeptree-2.0.0b1-py3-none-any.whl
  pip==20.1.1
setuptools==47.1.1
wheel==0.34.2

On confirming that there are no conflicting dependencies, you can even treat this as a "lock file" where all packages, including the transient dependencies will be pinned to their currently installed versions. Note that the locked-requirements.txt file could end up with duplicate entries. Although pip install wouldn't complain about that, you can avoid duplicate lines (at the cost of losing indentation) as follows,

$ pipdeptree -f | sed 's/ //g' | sort -u > locked-requirements.txt

Using pipdeptree with external tools

New in ver. 0.5.0

It's also possible to have pipdeptree output json representation of the dependency tree so that it may be used as input to other external tools.

$ pipdeptree --json

Note that --json will output a flat list of all packages with their immediate dependencies. This is not very useful in itself. To obtain nested json, use --json-tree

New in ver. 0.11.0

$ pipdeptree --json-tree

Visualizing the dependency graph

The dependency graph can also be visualized using GraphViz:

$ pipdeptree --graph-output dot > dependencies.dot
$ pipdeptree --graph-output pdf > dependencies.pdf
$ pipdeptree --graph-output png > dependencies.png
$ pipdeptree --graph-output svg > dependencies.svg

Note that graphviz is an optional dependency ie. required only if you want to use --graph-output. If the version of graphviz installed in the env is older than 0.18.1, then a warning will be displayed about upgrading graphviz. Support for older versions of graphviz will be dropped soon.

Since version 2.0.0b1, --package and --reverse flags are supported for all output formats ie. text, json, json-tree and graph.

In earlier versions, --json, --json-tree and --graph-output options override --package and --reverse.

Usage

usage: pipdeptree.py [-h] [-v] [-f] [--python PYTHON] [-a] [-l] [-u]
                     [-w [{silence,suppress,fail}]] [-r] [-p PACKAGES]
                     [-e PACKAGES] [-j] [--json-tree]
                     [--graph-output OUTPUT_FORMAT]

Dependency tree of the installed python packages

optional arguments:
  -h, --help            show this help message and exit
  -v, --version         show program's version number and exit
  -f, --freeze          Print names so as to write freeze files
  --python PYTHON       Python to use to look for packages in it (default:
                        where installed)
  -a, --all             list all deps at top level
  -l, --local-only      If in a virtualenv that has global access do not show
                        globally installed packages
  -u, --user-only       Only show installations in the user site dir
  -w [{silence,suppress,fail}], --warn [{silence,suppress,fail}]
                        Warning control. "suppress" will show warnings but
                        return 0 whether or not they are present. "silence"
                        will not show warnings at all and always return 0.
                        "fail" will show warnings and return 1 if any are
                        present. The default is "suppress".
  -r, --reverse         Shows the dependency tree in the reverse fashion ie.
                        the sub-dependencies are listed with the list of
                        packages that need them under them.
  -p PACKAGES, --packages PACKAGES
                        Comma separated list of select packages to show in the
                        output. If set, --all will be ignored.
  -e PACKAGES, --exclude PACKAGES
                        Comma separated list of select packages to exclude
                        from the output. If set, --all will be ignored.
  -j, --json            Display dependency tree as json. This will yield "raw"
                        output that may be used by external tools. This option
                        overrides all other options.
  --json-tree           Display dependency tree as json which is nested the
                        same way as the plain text output printed by default.
                        This option overrides all other options (except
                        --json).
  --graph-output OUTPUT_FORMAT
                        Print a dependency graph in the specified output
                        format. Available are all formats supported by
                        GraphViz, e.g.: dot, jpeg, pdf, png, svg

Known issues

1. pipdeptree relies on the internal API of pip. I fully understand that it's a bad idea but it mostly works! On rare occasions, it breaks when a new version of pip is out with backward incompatible changes in internal API. So beware if you are using this tool in environments in which pip version is unpinned, specially automation or CD/CI pipelines.

Limitations & Alternatives

pipdeptree merely looks at the installed packages in the current environment using pip, constructs the tree, then outputs it in the specified format. If you want to generate the dependency tree without installing the packages, then you need a dependency resolver. You might want to check alternatives such as pipgrip or poetry.

Runing Tests (for contributors)

There are 2 test suites in this repo:

1. Unit tests that use mock objects. These are configured to run on every push to the repo and on every PR thanks to Github Actions.
2. End-to-end tests that are run against actual packages installed in virtualenvs

Unit tests can be run against all version of python using tox as follows:

$ make test-tox-all

This assumes that you have python versions specified in the tox.ini file.

If you don't want to install all the versions of python but want to run tests quickly against Python3.6 only:

$ make test

Unit tests are written using pytest and you can also run the tests with code coverage as follows,

$ make test-cov

On the other hand, end-to-end tests actually create virtualenvs, install packages and then run tests against them. These tests are more reliable in the sense that they also test pipdeptree with the latest version of pip and setuptools.

The downside is that when new versions of pip or setuptools are released, these need to be updated. At present the process is manual but I have plans to setup nightly builds for these for faster feedback.

The end-to-end tests can be run as follows,

$ make test-e2e  # starts with a clean virtualenvs

$ # or

$ make test-e2e-quick # reuses existing virtualenvs

By default the e2e tests uses python executable python3.6. To use an alternate version set the environment var E2E_PYTHON_EXE.

$ E2E_PYTHON_EXE=python2.7 make test-e2e

Release checklist

1. Make sure that tests pass on Github Actions.
2. Create a commit with following changes and push it to github
3. Update the __version__ in the pipdeptree.py file.
   1. Add Changelog in CHANGES.md file.
   2. Also update README.md if required.
4. Create an annotated tag on the above commit and push the tag to github
5. Upload new version to PyPI.

License

MIT (See LICENSE)

Footnotes

[1]	(1, 2) pip version 20.3 has been released in Nov 2020 with the dependency resolver <https://blog.python.org/2020/11/pip-20-3-release-new-resolver.html>_

[2]	If you are on windows (powershell) you can run pipdeptree --warn silence | Select-String -Pattern '^\w+' instead of grep