API Reference

This is the place where the functions and classes in distlib's public API are described.

The distlib package

Classes

class DistlibException[source]

This is the base class for all exceptions raised by this packages, other than lower-level standard Python exceptions.

The distlib.database package

Classes

class DistributionPath[source]

This class represents a set of distributions which are installed on a Python path (like PYTHONPATH / sys.path). Both new-style (distlib) and legacy (egg) distributions are catered for.

Methods:

__init__(path=None, include_egg=False)[source]

Initialise the instance using a particular path.

Parameters:
  • path (list[str]) – The path to use when looking for distributions. If None is specified, sys.path is used.
  • include_egg – If True, legacy distributions (eggs) are included in the search; otherwise, they aren’t.
enable_cache()

Enables a cache, so that metadata information doesn’t have to be fetched from disk. The cache is per instance of the DistributionPath instance and is enabled by default. It can be disabled using disable_cache() and cleared using clear_cache() (disabling won’t automatically clear it).

disable_cache()

Disables the cache, but doesn’t clear it.

clear_cache()[source]

Clears the cache, but doesn’t change its enabled/disabled status. If enabled, the cache will be re-populated when querying for distributions.

get_distributions()[source]

The main querying method if you want to look at all the distributions. It returns an iterator which returns Distribution and, if include_egg was specified as True for the instance, also instances of any EggInfoDistribution for any legacy distributions found.

get_distribution(name)[source]

Looks for a distribution by name. It returns the first one found with that name (there should only be one distribution with a given name on a given search path). Returns None if no distribution was found, or else an instance of Distribution (or, if include_egg was specified as True for the instance, an instance of EggInfoDistribution if a legacy distribution was found with that name).

Parameters:name (str) – The name of the distribution to search for.
get_exported_entries(category, name=None)[source]

Returns an iterator for entries exported by distributions on the path.

Parameters:
  • category (str) – The export category to look in.
  • name (str) – A specific name to search for. If not specified, all entries in the category are returned.
Returns:

An iterator which iterates over exported entries (instances of ExportEntry).

class Distribution[source]

A class representing a distribution, typically one which hasn’t been installed (most likely, one which has been obtained from an index like PyPI).

Properties:

name

The name of the distribution.

version

The version of the distribution.

metadata

The metadata for the distribution. This is a distlib.metadata.Metadata instance.

download_url

The download URL for the distribution. If there are multiple URLs, this will be one of the values in download_urls.

download_urls

A set of known download URLs for the distribution.

New in version 0.2.0: The download_urls attribute was added.

digest

The digest for the source distribution. This is either None or a 2-tuple consisting of the hashing algorithm and the digest using that algorithm, e.g. ('sha256', '01234...').

digests

A dictionary mapping download URLs to digests, if and when digests are available.

New in version 0.2.0: The digests attribute was added.

locator

The locator for an instance which has been retrieved through a locator. This is None for an installed distribution.

class InstalledDistribution(Distribution)[source]

A class representing an installed distribution. This class is not instantiated directly, except by packaging tools. Instances of it are returned from querying a DistributionPath.

Properties:

requested

Whether the distribution was installed by user request (if not, it may have been installed as a dependency of some other distribution).

exports

The distribution’s exports, as described in Exporting things from Distributions. This is a cached property.

Methods:

list_installed_files()[source]

Returns an iterator over all of the individual files installed as part of the distribution, including metadata files. The iterator returns tuples of the form (path, hash, size). The list of files is written by the installer to the RECORD metadata file.

list_distinfo_files()[source]

Similar to list_installed_files(), but only returns metadata files.

check_installed_files()[source]

Runs over all the installed files to check that the size and checksum are unchanged from the values in the RECORD file, written when the distribution was installed. It returns a list of mismatches. If the files in the distribution haven’t been corrupted , an empty list will be returned; otherwise, a list of mismatches will be returned.

Returns:A list which, if non-empty, will contain tuples with the following elements:
  • The path in RECORD which failed to match.
  • One of the strings ‘exists’, ‘size’ or ‘hash’ according to what didn’t match (existence is checked first, then size, then hash).
  • The expected value of what didn’t match (as obtained from RECORD).
  • The actual value of what didn’t match (as obtained from the file system).
read_exports(filename=None)[source]

Read exports information from a file.

Normal access to a distribution’s exports should be through its exports attribute. This method is called from there as needed. If no filename is specified, the EXPORTS file in the .dist-info directory is read (it is expected to be present).

Parameters:filename (str) – The filename to read from, or None to read from the default location.
Returns:The exports read from the file.
Return type:dict
write_exports(exports, filename=None)[source]

Write exports information to a file.

If no filename is specified, the EXPORTS file in the .dist-info directory is written.

Parameters:
  • exports (dict) – A dictionary whose keys are categories and whose values are dictionaries which contain ExportEntry instances keyed on their name.
  • filename (str) – The filename to read from, or None to read from the default location.
class EggInfoDistribution[source]

Analogous to Distribution, but covering legacy distributions. This class is not instantiated directly. Instances of it are returned from querying a DistributionPath.

Properties:

name

The name of the distribution.

version

The version of the distribution.

metadata

The metadata for the distribution. This is a distlib.metadata.Metadata instance.

Methods:

list_installed_files()[source]

Returns a list all of the individual files installed as part of the distribution.

class DependencyGraph[source]

This class represents a dependency graph between releases. The nodes are distribution instances; the edges model dependencies. An edge from a to b means that a depends on b.

add_distribution(distribution)[source]

Add distribution to the graph.

add_edge(x, y, label=None)[source]

Add an edge from distribution x to distribution y with the given label (string).

add_missing(distribution, requirement)[source]

Add a missing requirement (string) for the given distribution.

repr_node(dist, level=1)[source]

Print a subgraph starting from dist. level gives the depth of the subgraph.

Direct access to the graph nodes and edges is provided through these attributes:

adjacency_list

Dictionary mapping distributions to a list of (other, label) tuples where other is a distribution and the edge is labelled with label (i.e. the version specifier, if such was provided).

reverse_list

Dictionary mapping distributions to a list of predecessors. This allows efficient traversal.

missing

Dictionary mapping distributions to a list of requirements that were not provided by any distribution.

Functions

make_graph(dists, scheme='default')[source]

Return a dependency graph from the given distributions.

get_dependent_dists(dists, dist)[source]

Recursively generate a list of distributions from dists that are dependent on dist.

get_required_dists(dists, dist)[source]

Recursively generate a list of distributions from dists that are required by dist.

The distlib.resources package

Attributes

cache

An instance of ResourceCache. This can be set after module import, but before calling any functionality which uses it, to ensure that the cache location is entirely under your control.

If you access the file_path property of Resource instance, the cache will be needed, and if not set by you, an instance with a default location will be created. See distlib.util.get_cache_base() for more information.

Functions

finder(package)[source]

Get a finder for the specified package.

If the package hasn’t been imported yet, an attempt will be made to import it. If importing fails, an ImportError will be raised.

Parameters:package (str) – The name of the package for which a finder is desired.
Returns:A finder for the package.
register_finder(loader, finder_maker)[source]

Register a callable which makes finders for a particular type of PEP 302 loader.

Parameters:
  • loader – The loader for which a finder is to be returned.
  • finder_maker – A callable to be registered, which is called when a loader of the specified type is used to load a package. The callable is called with a single argument – the Python module object corresponding to the package – and must return a finder for that package.

Classes

class Resource[source]

A class representing resources. It is never instantiated directly, but always through calling a finder’s find method.

Properties:

is_container

Whether this instance is a container of other resources.

bytes

All of the resource data as a byte string. Raises an exception if accessed on a container resource.

size

The size of the resource data in bytes. Raises an exception if accessed on a container resource.

resources

The relative names of all the contents of this resource. Raises an exception if accessed on a resource which is not a container.

path

This attribute is set by the resource’s finder. It is a textual representation of the path, such that if a PEP 302 loader’s get_data() method is called with the path, the resource’s bytes are returned by the loader. This attribute is analogous to the resource_filename API in setuptools. Note that for resources in zip files, the path will be a pointer to the resource in the zip file, and not directly usable as a filename. While setuptools deals with this by extracting zip entries to cache and returning filenames from the cache, this does not seem an appropriate thing to do in this package, as a resource is already made available to callers either as a stream or a string of bytes.

file_path

This attribute is the same as the path for file-based resource. For resources in a .zip file, the relevant resource is extracted to a file in a cache in the file system, and the name of the cached file is returned. This is for use with APIs that need file names, or need to be able to access data through OS-level file handles. See the ResourceCache documentation for more information about the cache.

Methods:

as_stream()[source]

A binary stream of the resource’s data. This must be closed by the caller when it’s finished with.

Raises an exception if called on a container resource.

class ResourceFinder[source]

A base class for resource finders, which finds resources for packages stored in the file system.

__init__(module)[source]

Initialise the finder for the package specified by module.

Parameters:module – The Python module object representing a package.
find(resource_name)[source]

Find a resource with the name specified by resource_name and return a Resource instance which represents it.

Parameters:resource_name – A fully qualified resource name, with hierarchical components separated by ‘/’.
Returns:A Resource instance, or None if a resource with that name wasn’t found.
iterator(resource_name)[source]

Return a generator which walks through the resources available through resource_name.

Parameters:resource_name – A fully qualified resource name, with hierarchical components separated by ‘/’. You can use ‘’ to mean the ‘root’ resource. If the resource name refers to a non-container resource, only that resource is returned. Otherwise, the named resource is returned, followed by its children, recursively. If there is no resource named resource_name, None is returned.
Returns:A generator to iterate over resources, or None.
is_container(resource)[source]

Return whether a resource is a container of other resources.

Parameters:resource (a Resource instance) – The resource whose status as container is wanted.
Returns:True or False.
get_stream(resource)[source]

Return a binary stream for the specified resource.

Parameters:resource (a Resource instance) – The resource for which a stream is wanted.
Returns:A binary stream for the resource.
get_bytes(resource)[source]

Return the contents of the specified resource as a byte string.

Parameters:resource (a Resource instance) – The resource for which the bytes are wanted.
Returns:The data in the resource as a byte string.
get_size(resource)[source]

Return the size of the specified resource in bytes.

Parameters:resource (a Resource instance) – The resource for which the size is wanted.
Returns:The size of the resource in bytes.
get_cache_info(resource)[source]

Return the cache information for the specified resource. This is a two-tuple consisting of an optional prefix (currently used for zip-based resources only, and is otherwise None) and an absolute path.

Parameters:resource (a Resource instance) – The resource for which the cache info is wanted.
Returns:tuple
class ZipResourceFinder[source]

This has the same interface as ResourceFinder.

class ResourceCache[source]

This class implements a cache for resources which must be accessible as files in the file system. It is based on Cache, and adds resource-specific methods.

__init__(base=None)[source]

Initialise a cache instance with a specific directory which holds the cache. If base is not specified, the value resource-cache in the directory returned by get_cache_base() is used.

get(resource)[source]

Ensures that the resource is available as a file in the file system, and returns the name of that file. This method calls the resource’s finder’s get_cache_info() method.

is_stale(resource, path)[source]

Returns whether the data in the resource which is cached in the file system is stale compared to the resource’s current data. The default implementation returns True, causing the resource’s data to be re-written to the file every time.

The distlib.scripts package

Classes

class ScriptMaker[source]

A class used to install scripts based on specifications.

Attributes

source_dir

The directory where script sources are to be found.

target_dir

The directory where scripts are to be created.

add_launchers

Whether to create native executable launchers on Windows.

force

Whether to overwrite scripts even when timestamps show they’re up to date.

clobber

Whether to overwrite existing scripts. The default is False, which means that existing scripts will not be overwritten.

executable

Value to use for the executable to use when constructing a shebang. If specified, it is used in place of any value determined algorithmically.

set_mode

Whether, on POSIX, the scripts should have their execute mode set.

script_template

The text of a template which should contain %(shebang)s, %(module)s and %(func)s in the appropriate places.

The attribute is defined at class level. You can override it at the instance level to customise your scripts.

version_info

A two-tuple of the Python version to be used when generating scripts, where version-specific variants such as foo3 or foo-3.8 are created. This defaults to sys.version_info. The provided tuple can have more elements, but only the first two are used.

New in version 0.3.1.

variant_separator

A string value placed between the root basename and the version information in a variant-specific filename. This defaults to '-', which means that a script with root basename foo and a variant X.Y will have a base filename of foo-3.8 for target Python version 3.8. If you wanted to write foo3.8 instead of foo-3.8, this attribute could be set to ''. If you need more control over filename generation, you can subclass ScriptMaker and override the get_script_filenames() method.

New in version 0.3.2.

Methods

__init__(source_directory, target_directory, add_launchers=True, dry_run=False)[source]

Initialise the instance with options that control its behaviour.

Parameters:
  • source_directory (str) – Where to find scripts to install.
  • target_directory (str) – Where to install scripts to.
  • add_launchers (bool) –

    If true, create executable launchers on Windows. The executables are currently generated from the following project:

    https://bitbucket.org/vinay.sajip/simple_launcher/

  • dry_run – If true, don’t actually install scripts - just pretend to.
make(specification, options=None)[source]

Make a script in the target directory.

Parameters:
  • specification (str) –

    A specification, which can take one of the following forms:

    • A filename, relative to source_directory, such as foo.py or subdir/bar.py.
    • A reference to a callable, given in the form:
      name = some_package.some_module:some_callable [flags]
      

      where the flags part is optional.

      When this form is passed, a Python stub script is created with the appropriate shebang line and with code to load and call the specified callable with no arguments, returning its value as the return code from the script.

      For more information about flags, see Flag formats.

  • options (dict) –

    If specified, a dictionary of options used to control script creation. Currently, the following keys are checked:

    gui: This should be a bool which, if True, indicates that the script is a windowed application. This distinction is only drawn on Windows if add_launchers is True, and results in a windowed native launcher application if options['gui'] is True (otherwise, the native executable launcher is a console application).

    interpreter_args: If specified, this should be a list of strings which are appended to the interpreter executable in the shebang line. If there are values with spaces, you will need to surround them with double quotes.

    Note

    Linux does not handle passing arguments to interpreters particularly well – multiple arguments are bundled up into one when passing to the interpreter – see https://en.wikipedia.org/wiki/Shebang_line#Portability for more information. This may also affect other POSIX platforms – consult the OS documentation for your system if necessary. On Windows, the distlib native executable launchers do parse multiple arguments and pass them to the interpreter.

Returns:

A list of absolute pathnames of files installed (or which would have been installed, but for dry_run being true).

make_multiple(specifications, options)[source]

Make multiple scripts from an iterable.

This method just calls make() once for each value returned by the iterable, but it might be convenient to override this method in some scenarios to do post-processing of the installed files (for example, running 2to3 on them).

Parameters:
  • specifications – an iterable giving the specifications to follow.
  • options – As for the make() method.
Returns:

A list of absolute pathnames of files installed (or which would have been installed, but for dry_run being true).

get_script_filenames(name)[source]

Get the names of scripts to be written for the specified base name, based on the variants and version_info for this instance. You can override this if you need to customise the filenames to be written.

Parameters:name (str) – the basename of the script to be written.
Returns:A set of filenames of files to be written as scripts, based on what variants are specified. For example, if the name is foo and the variants are {'X', 'X.Y'} and the version_info is (3, 8), then the result would be {'foo3', 'foo-3.8'}.

New in version 0.3.2.

Functions

enquote_executable(path)[source]

Cover an executable path in quotes. This only applies quotes if the passed path contains any spaces. It’s at least a little careful when doing the quoting - for example, producing e.g. /usr/bin/env "/dir with spaces/bin/jython" instead of "/usr/bin/env /dir with spaces/bin/jython"

Changed in version 0.3.1: This was an internal function _enquote_executable in earlier versions.

The distlib.locators package

Classes

class Locator[source]

The base class for locators. Implements logic common to multiple locators.

matcher

A :class:~distlib.version.VersionMatcher`

__init__(scheme='default')[source]

Initialise an instance of the locator.

Parameters:scheme (str) – The version scheme to use.
_get_project(name)[source]

This method should be implemented in subclasses. It returns a (potentially empty) dictionary whose keys are the versions located for the project named by name, and whose values are instances of distlib.database.Distribution.

get_project(name)[source]

This method calls _get_project() to do the actual work, and provides a caching layer on top.

convert_url_to_download_info(url, project_name)[source]

Extract information from a URL about the name and version of a distribution.

Parameters:
  • url (str) – The URL potentially of an archive (though it needn’t be).
  • project_name (str) – This must match the project name determined from the archive (case-insensitive matching is used).
Returns:

None if the URL does not appear to be that of a distribution archive for the named project. Otherwise, a dictionary is returned with the following keys at a minimum:

  • url – the URL passed in, minus any fragment portion.
  • filename – a suitable filename to use for the archive locally.

Optional keys returned are:

  • md5_digest – the MD5 hash of the archive, for verification after downloading. This is extracted from the fragment portion, if any, of the passed-in URL.
  • sha256_digest – the SHA256 hash of the archive, for verification after downloading. This is extracted from the fragment portion, if any, of the passed-in URL.

Return type:

dict

get_distribution_names()[source]

Get the names of all distributions known to this locator.

The base class raises NotImplementedError; this method should be implemented in a subclass.

Returns:All distributions known to this locator.
Return type:set
locate(requirement, prereleases=False)[source]

This tries to locate the latest version of a potentially downloadable distribution which matches a requirement (name and version constraints). If a potentially downloadable distribution (i.e. one with a download URL) is not found, None is returned – otherwise, an instance of Distribution is returned. The returned instance will have, at a minimum, name, version and source_url populated.

Parameters:
  • requirement (str) – The name and optional version constraints of the distribution to locate, e.g. 'Flask' or 'Flask (>= 0.7, < 0.9)'.
  • prereleases (bool) – If True, prereleases are treated like normal releases. The default behaviour is to not return any prereleases unless they are the only ones which match the requirement.
Returns:

A matching instance of Distribution, or None.

get_errors()[source]

This returns a (possibly empty) list of error messages relating to a recent get_project() or locate() call. Fetching the errors clears the error list.

New in version 0.2.4.

class DirectoryLocator(Locator)[source]

This locator scans the file system under a base directory, looking for distribution archives. The locator scans all subdirectories recursively, unless the recursive flag is set to False.

__init__(base_dir, **kwargs)[source]
Parameters:
  • base_dir (str) – The base directory to scan for distribution archives.
  • kwargs – Passed to base class constructor, apart from the following keyword arguments:
    • recursive (defaults to True) – if False, no recursion into subdirectories occurs.
class PyPIRPCLocator(Locator)[source]

This locator uses the PyPI XML-RPC interface to locate distribution archives and other data about downloads.

__init__(url, **kwargs)[source]
param url:The base URL to use for the XML-RPC service.
type url:str
param kwargs:Passed to base class constructor.
get_project(name)

See Locator.get_project().

class PyPIJSONLocator(Locator)[source]

This locator uses the PyPI JSON interface to locate distribution archives and other data about downloads. It gets the metadata and URL information in a single call, so it should perform better than the XML-RPC locator.

__init__(url, **kwargs)[source]
param url:The base URL to use for the JSON service.
type url:str
param kwargs:Passed to base class constructor.
get_project(name)

See Locator.get_project().

class SimpleScrapingLocator[source]

This locator uses the PyPI ‘simple’ interface – a Web scraping interface – to locate distribution archives.

__init__(url, timeout=None, num_workers=10, **kwargs)[source]
Parameters:
  • url (str) – The base URL to use for the simple service HTML pages.
  • timeout (float) – How long (in seconds) to wait before giving up on a remote resource.
  • num_workers (int) – The number of worker threads created to perform scraping activities.
  • kwargs – Passed to base class constructor.
class DistPathLocator[source]

This locator uses a DistributionPath instance to locate installed distributions.

__init__(url, distpath, **kwargs)[source]
Parameters:
  • distpath (DistributionPath) – The distribution path to use.
  • kwargs – Passed to base class constructor.
class AggregatingLocator(Locator)[source]

This locator uses a list of other aggregators and delegates finding projects to them. It can either return the first result found (i.e. from the first aggregator in the list provided which returns a non-empty result), or a merged result from all the aggregators in the list.

__init__(*locators, **kwargs)[source]
Parameters:
  • locators (list[Locator]) – A list of aggregators to delegate finding projects to.
  • merge (bool) – If this kwarg is True, each aggregator in the list is asked to provide results, which are aggregated into a results dictionary. If False, the first non-empty return value from the list of aggregators is returned. The locators are consulted in the order in which they’re passed in.
class DependencyFinder[source]

This class allows you to recursively find all the distributions which a particular distribution depends on.

__init__(locator)[source]

Initialise an instance with the locator to be used for locating distributions.

find(requirement, metas_extras=None, prereleases=False)[source]

Find all the distributions needed to fulfill requirement.

Parameters:
  • requirement – A string of the from name (version) where version can include an inequality constraint, or an instance of Distribution (e.g. representing a distribution on the local hard disk).
  • meta_extras – A list of meta extras such as :test:, :build: and so on, to be included in the dependencies.
  • prereleases – If True, allow pre-release versions to be returned - otherwise, don’t return prereleases unless they’re all that’s available.
Returns:

A 2-tuple. The first element is a set of Distribution instances. The second element is a set of problems encountered during dependency resolution. Currently, if this set is non- empty, it will contain 2-tuples whose first element is the string ‘unsatisfied’ and whose second element is a requirement which couldn’t be satisfied.

In the set of Distribution instances returned, some attributes will be set:

  • The instance representing the passed-in requirement will have the requested attribute set to True.
  • All requirements which are not installation requirements (in other words, are needed only for build and test) will have the build_time_dependency attribute set to True.

Functions

get_all_distribution_names(url=None)[source]

Retrieve the names of all distributions registered on an index.

Parameters:url (str) – The XML-RPC service URL of the node to query. If not specified, The main PyPI index is queried.
Returns:A list of the names of distributions registered on the index. Note that some of the names may be Unicode.
Return type:list
locate(requirement, prereleases=False)

This convenience function returns the latest version of a potentially downloadable distribution which matches a requirement (name and version constraints). If a potentially downloadable distribution (i.e. one with a download URL) is not found, None is returned – otherwise, an instance of Distribution is returned. The returned instance will have, at a minimum, name, version, download_url and download_urls.

Parameters:
  • requirement (str) – The name and optional version constraints of the distribution to locate, e.g. 'Flask' or 'Flask (>= 0.7, < 0.9)'.
  • prereleases (bool) – If True, prereleases are treated like normal releases. The default behaviour is to not return any prereleases unless they are the only ones which match the requirement.
Returns:

A matching instance of Distribution, or None.

Variables

default_locator

This attribute holds a locator which is used by locate() to locate distributions.

The distlib.index package

Classes

class PackageIndex[source]

This class represents a package index which is compatible with PyPI, the Python Package Index. It allows you to register projects, upload source and binary distributions (with support for digital signatures), upload documentation, verify signatures and get a list of hosts which are mirrors for the index.

Methods:

__init__(url=None, mirror_host=None)[source]

Initialise an instance, setting instance attributes named from the keyword arguments.

Parameters:
  • url – The root URL for the index. If not specified, the URL for PyPI is used (’http://pypi.org/pypi’).
  • mirror_host – The DNS name for a host which can be used to determine available mirror hosts for the index. If not specified, the value ‘last.pypi.python.org’ is used.
register(metadata)[source]

Register a project with the index.

Parameters:metadata – A Metadata instance. This should have at least the Name and Version fields set, and ideally as much metadata as possible about this distribution. Though it might seem odd to have to specify a version when you are initially registering a project, this is required by PyPI. You can see this in PyPI’s Web UI when you click the “Package submission” link in the left-hand side menu.
Returns:An urllib HTTP response returned by the index. If an error occurs, an HTTPError exception will be raised.
upload_file(metadata, filename, signer=None, sign_password=None, filetype='sdist', pyversion='source', keystore=None)[source]

Upload a distribution to the index.

Parameters:
  • metadata – A Metadata instance. This should have at least the Name and Version fields set, and ideally as much metadata as possible about this distribution.
  • file_name – The path to the file which is to be uploaded.
  • signer – If specified, this needs to be a string identifying the GnuPG private key which is to be used for signing the distribution.
  • sign_password – The passphrase which allows access to the private key used for the signature.
  • filetype – The type of the file being uploaded. This would have values such as sdist (for a source distribution), bdist_wininst for a Windows installer, and so on. Consult the distutils documentation for the full set of possible values.
  • pyversion – The Python version this distribution is compatible with. If it’s a pure-Python distribution, the value to use would be source - for distributions which are for specific Python versions, you would use the Python version in the form X.Y.
  • keystore – The path to a directory which contains the keys used in signing. If not specified, the instance’s gpg_home attribute is used instead. This parameter is not used unless a signer is specified.
Returns:

An urllib HTTP response returned by the index. If an error occurs, an HTTPError exception will be raised.

Changed in version 0.1.9: The keystore argument was added.

upload_documentation(metadata, doc_dir)[source]

Upload HTML documentation to the index. The contents of the specified directory tree will be packed into a .zip file which is then uploaded to the index.

Parameters:
  • metadata – A Metadata instance. This should have at least the Name and Version fields set.
  • doc_dir – The path to the root directory for the HTML documentation. This directory should be the one that contains index.html.
Returns:

An urllib HTTP response returned by the index. If an error occurs, an HTTPError exception will be raised.

verify_signature(self, signature_filename, data_filename, keystore=None)[source]

Verify a digital signature against a downloaded distribution.

Parameters:
  • signature_filename – The path to the file which contains the digital signature.
  • data_filename – The path to the file which was supposedly signed to obtain the signature in signature_filename.
  • keystore – The path to a directory which contains the keys used in verification. If not specified, the instance’s gpg_home attribute is used instead.
Returns:

True if the signature can be verified, else False. If an error occurs (e.g. unable to locate the public key used to verify the signature), a ValueError is raised.

Changed in version 0.1.9: The keystore argument was added.

read_configuration()[source]

Read the PyPI access configuration.

save_configuration()[source]

Save the PyPI access configuration. You must have set username and password attributes before calling this method.

search(query, operation=None)[source]

Search the index for distributions matching a search query.

Parameters:
  • query – The query, either as a string or a dictionary. If a string 'foo' is passed, it will be treated equivalently to passing the dictionary {'name': 'foo'}. The dictionary can have the following keys:
    • name
    • version
    • stable_version
    • author
    • author_email
    • maintainer
    • maintainer_email
    • home_page
    • license
    • summary
    • description
    • keywords
    • platform
    • download_url
    • classifiers (list of classifier strings)
    • project_url
    • docs_url (URL of the pythonhosted.org docs if they’ve been supplied)
  • operation – If specified, it should be either 'and' or 'or'. If not specified, 'and' is assumed. This is only used if a passed dictionary has multiple keys. It determines whether the intersection or the union of matches is returned.
Returns:

A (possibly empty) list of the distributions matching the query. Each entry in the list will be a dictionary with the following keys:

  • _pypi_ordering – the internal ordering value (an integer)
  • name –The name of the distribution
  • version – the version of the distribution
  • summary – the summary for that version

New in version 0.1.8.

Additional attributes:

username

The username to use when authenticating with the index.

password

The password to use when authenticating with the index.

gpg

The path to the signing and verification program.

gpg_home

The location of the key database for the signing and verification program.

mirrors

The list of hosts which are mirrors for this index.

boundary

The boundary value to use when MIME-encoding requests to be sent to the index. This should be a byte-string.

The distlib.util package

Classes

class Cache[source]

This base class implements common operations for distlib caches.

__init__(base)[source]

Initialise a cache instance with a specific directory which holds the cache.

Warning

If base is specified and exists, it should exist and its permissions (relevant on POSIX only) should be set to 0700 - i.e. only the user of the running process has any rights over the directory. If this is not done, the application using this functionality may be vulnerable to security breaches as a result of other processes being able to interfere with the cache.

prefix_to_dir(prefix)[source]

Converts a prefix (e.g. the name of a resource’s containing .zip, or a wheel pathname) into a directory name in the cache. This implementation delegates the work to path_to_cache_dir().

class ExportEntry[source]

Attributes:

A class holding information about a exports entry.

name

The name of the entry.

prefix

The prefix part of the entry. For a callable or data item in a module, this is the name of the package or module containing the item.

suffix

The suffix part of the entry. For a callable or data item in a module, this is a dotted path which points to the item in the module.

flags

A list of flags. See Flag formats for more information.

value[source]

The actual value of the entry (a callable or data item in a module, or perhaps just a module). This is a cached property of the instance, and is determined by calling resolve() with the prefix and suffix properties.

dist

The distribution which exports this entry. This is normally an instance of InstalledDistribution.

class HTTPSHandler[source]

A request handler inheriting from urllib.request.HTTPSHandler which does certificate validation.

class HTTPSOnlyHandler[source]

A request handler inheriting from urllib.request.HTTPSHandler which raises an exception if an attempt is made to open an HTTP (as opposed to HTTPS) connection.

Functions

get_cache_base()[source]

Return the base directory which will hold distlib caches. If the directory does not exist, it is created.

On Windows, if LOCALAPPDATA is defined in the environment, then it is assumed to be a directory, and will be the parent directory of the result. On POSIX, and on Windows if LOCALAPPDATA is not defined, the user’s home directory – as determined using os.expanduser('~') – will be the parent directory of the result.

The result is just the directory '.distlib' in the parent directory as determined above.

If a home directory is unavailable (no such directory, or if it’s write- protected), a parent directory for the cache is determined using tempfile.mkdtemp(). This returns a directory to which only the running process has access (permission mask 0700 on POSIX), meaning that the cache should be isolated from possible malicious interference by other processes.

Note

This cache is used for the following purposes:

  • As a place to cache package resources which need to be in the file system, because they are used by APIs which either expect filesystem paths, or to be able to use OS-level file handles. An example of the former is the load_verify_locations() method in Python’s ssl module. The subdirectory resource-cache is used for this purpose.
  • As a place to cache shared libraries which are extracted as a result of calling the mount() method of the Wheel class. The subdirectory dylib-cache is used for this purpose.

The application using this cache functionality, whether through the above mechanisms or through using the value returned from here directly, is responsible for any cache cleanup that is desired. Note that on Windows, you may not be able to do cache cleanup if any of the cached files are open (this will generally be the case with shared libraries, i.e. DLLs). The best way to do cache cleanup in this scenario may be on application startup, before any resources have been cached or wheels mounted.

path_to_cache_dir(path)[source]

Converts a path (e.g. the name of an archive) into a directory name suitable for use in a cache. The following algorithm is used:

  1. On Windows, any ':' in the drive is replaced with '---'.
  2. Any occurrence of os.sep is replaced with '--'.
  3. '.cache' is appended.
get_export_entry(specification)[source]

Return a export entry from a specification, if it matches the expected format, or else None.

Parameters:specification (str) – A specification, as documented for the distlib.scripts.ScriptMaker.make() method.
Returns:None if the specification didn’t match the expected form for an entry, or else an instance of ExportEntry holding information about the entry.
resolve(module_name, dotted_path)[source]

Given a module name and a dotted_path representing an object in that module, resolve the passed parameters to an object and return that object.

If the module has not already been imported, this function attempts to import it, then access the object represented by dotted_path in the module’s namespace. If dotted_path is None, the module is returned. If import or attribute access fails, an ImportError or AttributeError will be raised.

Parameters:
  • module_name (str) – The name of a Python module or package, e.g. os or os.path.
  • dotted_path (str) – The path of an object expected to be in the module’s namespace, e.g. 'environ', 'sep' or 'path.supports_unicode_filenames'.

The distlib.wheel package

This package has functionality which allows you to work with wheels (see PEP 427).

Attributes

cache

An instance of distlib.util.Cache. This can be set after module import, but before calling any functionality which uses it, to ensure that the cache location is entirely under your control.

If you call the mount method of a Wheel instance, and the wheel is successfully mounted and contains C extensions, the cache will be needed, and if not set by you, an instance with a default location will be created. See distlib.util.get_cache_base() for more information.

COMPATIBLE_TAGS

A set of (pyver, abi, arch) tags which are compatible with this Python implementation.

Classes

class Wheel[source]

This class represents wheels – either existing wheels, or wheels to be built.

__init__(spec)[source]

Initialise an instance from a specification.

Parameters:spec (str) – This can either be a valid filename for a wheel (for when you want to work with an existing wheel), or just the name-version-buildver portion of a wheel’s filename (for when you’re going to build a wheel for a known version and build of a named project).
build(paths, tags=None, wheel_version=None)[source]

Build a wheel. The name, version and buildver should already have been set correctly.

Parameters:
  • paths – This should be a dictionary with keys 'prefix', 'scripts', 'headers', 'data' and one of 'purelib' or 'platlib'. These must point to valid paths if they are to be included in the wheel.
  • tags – If specified, this should be a dictionary with optional keys 'pyver', 'abi' and 'arch' indicating lists of tags which indicate environments with which the wheel is compatible.
  • wheel_version – If specified, this is written to the wheel’s “Wheel-Version” metadata. If not specified, the implementation’s latest supported wheel version is used.
install(self, paths, maker, **kwargs)[source]

Install from a wheel.

Parameters:
  • paths – This should be a dictionary with keys 'prefix', 'scripts', 'headers', 'data', 'purelib' and 'platlib'. These must point to valid paths to which files may be written if they are in the wheel. Only one of the 'purelib' and 'platlib' paths will be used (in the case where they are different), depending on whether the wheel is for a pure-Python distribution.
  • maker – This should be set to a suitably configured instance of ScriptMaker. The source_dir and target_dir arguments can be set to None when creating the instance - these will be set to appropriate values inside this method.
  • warner – If specified, should be a callable that will be called with (software_wheel_ver, file_wheel_ver) if they differ. They will both be in the form of tuples (major_ver, minor_ver).
  • lib_only – It’s conceivable that one might want to install only the library portion of a package – not installing scripts, headers, data and so on. If lib_only is specified as True, only the site-packages contents will be installed.
is_compatible()[source]

Determine whether this wheel instance is compatible with the running Python.

Returns:True if compatible, else False.
is_mountable()[source]

Determine whether this wheel instance is indicated suitable for mounting in its metadata.

Returns:True if mountable, else False.
mount(append=False)[source]

Mount the wheel so that its contents can be imported directly, without the need to install the wheel. If the wheel contains C extensions and has metadata about these extensions, the extensions are also available for import.

If the wheel tags indicate it is not compatible with the running Python, a DistlibException is raised. (The is_compatible() method is used to determine compatibility.)

If the wheel is indicated as not suitable for mounting, a distlib.DistlibException is raised. (The is_mountable() method is used to determine mountability.)

param append:If True, the wheel’s pathname is added to the end of sys.path. By default, it is added to the beginning.

Note

Wheels may state in their metadata that they are not intended to be mountable, in which case this method will raise a distlib.DistlibException with a suitable message. If C extensions are extracted, the location for extraction will be under the directory dylib-cache in the directory returned by get_cache_base().

Wheels may be marked by their publisher as unmountable to indicate that running directly from a zip is not supported by the packaged software.

unmount()[source]

Unmount the wheel so that its contents can no longer be imported directly. If the wheel contains C extensions and has metadata about these extensions, the extensions are also made unavailable for import.

Note

Unmounting does not automatically clean up any extracted C extensions, as that may not be desired (and not possible, on Windows, because the files will be open). See the get_cache_base() documentation for suggested cleanup scenarios.

verify()[source]

Verify sizes and hashes of the wheel’s contents against the sizes and hashes declared in the wheel’s RECORD. Raise a distlib.DistlibException if a size or digest mismatch is detected.

New in version 0.1.8.

update(modifier, dest_dir=None, **kwargs)[source]

Allows a user-defined callable access to the contents of a wheel. The callable can modify the contents of the wheel, add new entries or remove entries. The method first extracts the wheel’s contents to a temporary location, and then calls the modifier like this:

modified = modifier(path_map, **kwargs)

where path_map is a dictionary mapping archive paths to the location of the corresponding extracted archive entry, and kwargs is whatever was passed to the update method. If the modifier returns True, a new wheel is built from the (possibly updated) contents of path_map and, as a final step, copied to the location of the original wheel (hence effectively modifying it in-place). The passed path_map will contain all of the wheel’s entries other than the RECORD entry (which will be recreated if a new wheel is built).

New in version 0.1.8.

name

The name of the distribution.

version

The version of the distribution

buildver

The build tag for the distribution.

pyver

A list of Python versions with which the wheel is compatible. See PEP 427 and PEP 425 for details.

abi

A list of application binary interfaces (ABIs) with which the wheel is compatible. See PEP 427 and PEP 425 for details.

arch

A list of architectures with which the wheel is compatible. See PEP 427 and PEP 425 for details.

dirname

The directory in which a wheel file is found/to be created.

filename

The filename of the wheel (computed from the other attributes)

metadata

The metadata for the distribution in the wheel, as a Metadata instance.

info

The wheel metadata (contents of the WHEEL metadata file) as a dictionary.

exists

Whether the wheel file exists.

New in version 0.1.8.

Functions

is_compatible(wheel, tags=None)[source]

Indicate if a wheel is compatible with a set of tags. If any combination of the tags of wheel is found in tags, then the wheel is considered to be compatible.

Parameters:
  • wheel – A Wheel instance or the filename of a wheel.
  • tags – A set of tags to check for compatibility. If not specified, it defaults to the set of tags which are compatible with this Python implementation.
Returns:

True if compatible, else False.

The distlib.versions package

This package has functionality which allows you to work with standard versions (see PEP 440) but also legacy versions and semantic versions.

Classes

class Version[source]

This base class represents a version.

class Matcher[source]

This base class represents a version matcher. It’s used for parsing and comparing versions.

class VersionScheme[source]

This class represents a version scheme (e.g. legacy, semantic or standard).

class NormalizedVersion[source]

This class represents PEP 440-compatible versions.

class NormalizedMatcher[source]

This class represents a matcher for PEP 440-compatible versions.

Functions

get_scheme(name)[source]

Return a VersionScheme instance corresponding to the specified name.

The distlib.manifest package

This package has functionality which allows you to maintain a manifest, i.e. a list of files which makes up a source distribution.

Classes

class Manifest[source]

This base class represents a manifest. You can explore files under a directory, add and remove files, and process directives to e.g. include or exclude files based on patterns.

files

The list of paths representing files included in the manifest.

allfiles

The list of paths representing all files in the directory tree which the manifest is based on.

base

The base directory on which this manifest instance is based.

process_directive(directive)[source]
Process a directive which either adds some files from allfiles to
files, or removes some files from files.

The distlib.metadata package

This package has functionality which allows you to manage a distribution’s metadata.

Classes

class Metadata[source]

This class represents a distribution’s metadata.

todict()[source]

Returns the metadata as a dictionary.

The distlib.markers package

This package has functionality for interpreting environment markers.

Functions

interpret(marker, execution_context=None)[source]

This function returns the result of interpreting an environment marker in optional execution context, which is used for name lookups.

Next steps

You might find it helpful to look at the mailing list.