Compute a dependency graph between active Python eggs.
The eggdeps tool reports dependencies between eggs in the working set. Dependencies are considered recursively, creating a directed graph. This graph is printed to standard output either as plain text, or as an input file to the graphviz tools.
eggdeps [options] [specifications]
Specifications must follow the usual syntax for specifying distributions of Python packages as defined by pkg_resources.
- If any specifications are given, the corresponding distributions will make up the roots of the dependency graph, and the graph will be restricted to their dependencies.
- If no specifications are given, the graph will map the possible dependencies between all eggs in the working set and its roots will be those distributions that aren’t dependencies of any other distributions.
|-h, --help||show this help message and exit|
|-i IGNORE, --ignore=IGNORE|
|project names to ignore|
|-I RE_IGNORE, --re-ignore=RE_IGNORE|
|regular expression for project names to ignore|
|-e DEAD_ENDS, --dead-end=DEAD_ENDS|
|names of projects whose dependencies to ignore|
|-E RE_DEAD_ENDS, --re-dead-end=RE_DEAD_ENDS|
|regular expression for project names whose dependencies to ignore|
|always omit extra dependencies|
-n, –version-numbers print version numbers of active distributions
|-1, --once||in plain text output, include each distribution only once|
|-t, --terse||in plain text output, omit any hints at unprinted distributions, such as ellipses|
|-d, --dot||produce a dot graph|
|-c, --cluster||in a dot graph, cluster direct dependencies of each root distribution|
|produce a requirements list|
|in a requirements list, print loosest possible version specifications|
The -i, -I, -e, and -E options may occur multiple times.
If both the -d and -r options are given, the one listed last wins. When printing requirements lists, -v wins over -s.
The script entry point recognizes default values for all options, the variable names being the long option names with any dashes replaced by underscores (except for --no-extras, which translates to setting extras=False). This allows for setting defaults using the arguments option of the egg recipe in a buildout configuration, for example.
The goal of eggdeps is to compute a directed dependency graph with nodes that represent egg distributions from the working set, and edges which represent either mandatory or extra dependencies between the eggs.
The working set eggdeps operates on is defined by the egg distributions available to the running Python interpreter. For example, these may be the distributions activated by easy_install or installed in a zc.buildout environment.
If the graph is to be calculated to such specifications that not all required distributions are in the working set, the missing ones will be marked in the output, and their dependencies cannot be determined. The same happens if any distribution that is either specified on the command line or required by any other distribution is available in the working set, but at a version incompatible with the specified requirement.
Graph building strategies
The dependency graph may be built following either of two strategies:
|Analysing the whole working set:|
|Nodes correspond exactly to the distributions in the working set. Edges corresponding to all conceivable dependencies between any active distributions are included, but only if the required distribution is active at the correct version. The roots of the graph correspond to those distributions no other active distributions depend upon.|
|Starting from one or more eggs:|
|Nodes include all packages depended upon by the specified distributions and extras, as well as their deep dependencies. They may cover only part of the working set, as well as include nodes for distributions that are not active at the required versions or not active at all (so their dependencies can not be followed). The roots of the graph correspond to the specified distributions.|
Some information will be lost while building the graph:
- If a dependency occurs both mandatorily and by way of one or more extras, it will be recorded as a plain mandatory dependency.
- If a distribution A with installed extras is a dependency of multiple other distributions, they will all appear to depend on A with all its required extras, even if they individually require none or only a few of them.
Reducing the graph
In order to reduce an otherwise big and tangled dependency graph, certain nodes and edges may be omitted.
|Ignored nodes:||Nodes may be ignored completely by exact name or regular expression matching. This is useful if a very basic distribution is a depedency of a lot of others. An example might be setuptools.|
|Dead ends:||Distributions may be declared dead ends by exact name or regular expression matching. Dead ends are included in the graph but their own dependencies will be ignored. This allows for large subsystems of distributions to be blotted out except for their “entry points”. As an example, one might declare zope.app.* dead ends in the context of zope.* packages.|
|No extras:||Reporting and following extra dependencies may be switched off completely. This will probably make most sense when analysing the working set rather than the dependencies of specified distributions.|
There are two ways eggdeps can output the computed dependency graph: plain text (the default) and a dot file to be fed to the graphviz tools.
Plain text output
The graph is printed to standard output essentially one node per line, indented according to nesting depth, and annotated where appropriate. The dependencies of each node are sorted after the following criteria:
- Mandatory dependencies are printed before extra requirements.
- Dependencies of each set of extras are grouped, the groups being sorted alphabetically by the names of the extras.
- Dependencies which are either all mandatory or by way of the same set of extras are sorted alphabetically by name.
As an illustrating example, the following dependency graph was computed for two Zope packages, one of them required with a “test” extra depending on an uninstalled egg, and some graph reduction applied:
zope.annotation zope.app.container * zope.component zope.deferredimport zope.proxy zope.deprecation zope.event zope.dublincore zope.annotation ... [test] (zope.app.testing) *
If one or more dependencies of a node are due to extra requirements only, the names of those extras are printed in square brackets above their dependencies, half-indented relative to the node which requires them.
If a node with further dependencies occurs at several places in the graph, the subgraph is printed only once, the other occurences being marked by an ellipsis. The place where the subgraph is printed is chosen such that
If a distribution is not in the working set, its name is parenthesised.
Dead ends are marked by an asterisk.
Dot file output
In a dot graphics, nodes and edges are not annotated with text but colored.
These are the color codes for nodes, later ones overriding earlier ones in cases where more than one color is appropriate:
|Green:||Nodes corresponding to the roots of the graph.|
|Yellow:||Direct dependencies of any root nodes, whether mandatory or through extras.|
|Red:||Nodes for eggs installed at a version incompatible with some requirement, or not installed at all.|
Other than being highlighted by color, root nodes and their direct dependencies may be clustered. eggdeps tries to put each root node in its own cluster. However, if two or more root nodes share any direct dependencies, they will share a cluster as well.
All the distributions included in the graph may be output as the Python representation of a list of requirement specifications, either
- listing bare package names,
- including the exact versions as they occur in the working set, or
- specifying complex version requirements that take into account all version requirements made for the distribution in question (but disregard extras completely for the time being). Complex version requirements always require at least the version that occurs in the working set, assuming that we cannot know the version requirements of past versions but reasonably assume that requirements might stay the same for future versions.
The list is sorted alphabetically by distribution name.
|Author:||Thomas Lotze (email@example.com, http://thomas-lotze.de/)|
|Subversion web view:|
|Current change log:|