| | |
- boopak.pload.PackageLoader
-
- PackageCollection
class PackageCollection(boopak.pload.PackageLoader) |
| |
PackageCollection: manages a package collection. This is a subclass
of PackageLoader; it adds the ability to install and delete packages.
This class is intended to be used by a Boodler UI. It should not be
used by a sound-playing process.
Most of the methods defined here (installing, deleting, even examining
packages) invoke the clear_cache() method defined in PackageLoader.
This reflects the fact that any package change can change the
dependency tree, which is part of what PackageLoader caches. Clearing
the cache entirely is somewhat crude, but it's simple and it works.
PackageCollection(basedir=None, coldir=None, dldir=None,
boodler_api_vers=None, importing_ok=False) -- constructor
Create a PackageCollection. It requires two directory names:
coldir is the collection of packages, and dldir is a temporary
workspace for storing downloaded and unzipped files. Or you can just
provide a basedir, in which case these will be basedir/Collection
and basedir/Download (respectively).
The coldir and dldir will be created, if necessary. When the
PackageCollection shuts down, it will delete the files it has created
in the dldir.
If a version is provided (either as a VersionNumber or a string which
converts to one), then this is noted as the Boodler version which
packages must be compatible with. (As noted in the boodler.api_required
metadata field.)
(Note that the boodler.api_required field is not checked until the
package is actually loaded, which is after version selection occurs.
Therefore, it is best if the collection only contains packages which
are compatible with the installed Boodler version. None of this is
likely to be important, since the Boodler API version is not expected
to change much after 2.0.)
If the importing_ok flag is false, this loader will refuse to load
(that is, execute) Python code in packages. This is the default.
It should only be true if the create-package command is to be used.
Publicly readable fields:
collecdir -- the directory containing the package collection
downloaddir -- the directory containing temporary files and directories
currently_creating -- during a create operation, the package which is
being created (at the moment)
Public methods:
find_source() -- load a package from the collection, a .boop file, or
the Internet
install_source() -- install a package into the collection from a file
or the Internet
fetch_source() -- prepare to download a package from the Internet
delete_package() -- delete a package from the collection
delete_group() -- delete all versions of a package from the collection
delete_whole_collection() -- delete all packages in the entire collection
start_import_recording() -- begin noting all bimport() calls
stop_import_recording() -- stop noting bimport() calls, and return them
record_import() -- note a bimport() call by a package being imported
create_temp_dir() -- create a new, empty directory in the temporary dir
create_temp_file() -- create the pathname of a new temporary file
clean_temp() -- clean up the temporary workspace
shut_down() -- shut down the PackageCollection
Internal methods:
rewrite_versions_file() -- write (or overwrite) a new Versions file
|
| |
Methods defined here:
- __init__(self, basedir=None, coldir=None, dldir=None, boodler_api_vers=None, importing_ok=False)
- clean_temp(self)
- clean_temp() -> None
Clean up the temporary download workspace. Delete all files, and
clear the internal caches that referred to them.
- create_temp_dir(self, label='tmp')
- create_temp_dir(label='tmp') -> str
Return the pathname of a new, empty directory in the temporary
workspace. This directory will be deleted by the next clean_temp()
call.
The optional label argument will be used in the directory name.
- create_temp_file(self, label='tmp')
- create_temp_file(label='tmp') -> str
Return the pathname of a new file in the temporary workspace.
The file will not exist, but can be opened for writing. This
file will be deleted by the next clean_temp() call.
The optional label argument will be used in the file name.
- delete_group(self, pkgname)
- delete_group(pkgname) -> None
Delete all versions of a package from the collection.
This method does not generate an error if the package group
is malformed, or even missing.
- delete_package(self, pkgname, vers=None)
- delete_package(pkgname, vers=None) -> None
Delete a package from the collection.
If no second argument is given, the most recent available version
of the package is deleted. If the argument is a VersionNumber,
that version will be deleted. If it is a VersionSpec, the most
recent version that matches the spec will be deleted. A string
value will be converted to a VersionSpec (not a VersionNumber).
If the last version of a package is deleted, the entire group
directory is deleted too.
- delete_whole_collection(self)
- delete_whole_collection() -> None
Delete all packages in the entire collection.
- fetch_source(self, srctype, loc)
- fetch_source(srctype, loc) -> None or Fetcher
Make sure that a package of the form (Source_URL, url) is
downloaded and ready to be passed to find_source() or
install_source(). If it already is (or if srctype is Source_FILE
or Source_PACKAGE), then this function returns None, and the
caller may proceed.
If the URL is *not* downloaded, this returns a Fetcher object.
This should be called in some wise like this:
while (not fetcher.is_done()):
fetcher.work()
Once fetcher.is_done() returns true, the package is downloaded,
and the caller may proceed to find_source() or install_source().
- find_source(self, srctype, loc)
- find_source(srctype, loc) -> PackageInfo
Load a package given (srctype, loc) in one of the following forms:
(Source_PACKAGE, (pkgname, vers))
(Source_FILE, filename)
(Source_URL, url)
Source_PACKAGE will be found and loaded from the collection.
(The vers part may be a VersionNumber, VersionSpec, or None.)
Source_FILE will be unzipped (in the temporary work directory)
and then loaded. (However, the temporary directory will not be
kept in the loader's external package list. So you cannot reload
the package without doing another find_source(). To install the
file permanently, use install_source().)
Source_URL will be unzipped and loaded, but only if it has already
been fetched with fetch_source(). This is a blivet in the API;
it would be nice if find_source() could fetch the URL too, but
that is potentially a slow operation, and find_source() is not
set up for slow operations.
- install_source(self, srctype, loc)
- install_source(srctype, loc) -> PackageInfo
Install a package given (srctype, loc) in one of the following forms:
(Source_PACKAGE, (pkgname, vers))
(Source_FILE, filename)
(Source_URL, url)
Source_PACKAGE is not a valid argument for this function, since
it refers to a package which is already installed.
Source_FILE will be unzipped (in the temporary work directory)
and then moved to the collection directory.
Source_URL will be unzipped and installed, but only if it has
already been fetched with fetch_source(). This is a blivet in
the API; it would be nice if install_source() could fetch the
URL too, but that is potentially a slow operation, and
install_source() is not set up for slow operations.
- record_import(self, pkg, name, spec=None)
- record_import(pkg, name, spec=None) -> None
Note a bimport() call by a package being imported. This should
only be called by bimport().
- rewrite_versions_file(self, dirname, versionlist, pkgname='<unknown>')
- rewrite_versions_file(dirname, versionlist,
pkgname='<unknown>') -> None
Write (or overwrite) a new Versions file, in the given directory,
with the given list of VersionNumber objects. The pkgname is stored
in a comment line.
The versionlist may be altered (sorted) by this function, so the
caller should not pass a list object which will be used again.
- shut_down(self)
- shut_down() -> None
Shut down the PackageCollection. It may not be used again.
This cleans up the temporary download workspace.
- start_import_recording(self)
- start_import_recording() -> None
Set the collection to noting all bimport() calls in packages being
imported. (This is used only during package creation, to figure
out dependencies.)
- stop_import_recording(self)
- stop_import_recording() -> dic
Stop noting bimport() calls, and return a list of all such calls
noted. The return dict maps (name,version) pairs to lists
of import requests by that package. Each request in such a list
looks like (name,spec), or (name,version) for an exact request,
for (name,None) for an any-version request.
Data and other attributes defined here:
- instance_count = 0
Methods inherited from boopak.pload.PackageLoader:
- add_external_package(self, dirname, metadata=None, resources=None)
- add_external_package(dirname, metadata=None, resources=None) ->
(str, VersionNumber)
Add an external directory to load a package from. The argument
must be a complete (unpacked) Boodler directory, with a
Metadata file. That package will then be loadable (possibly
hiding a package in the collection directory).
The metadata and resources fields, if supplied, override the
Metadata and Resources files in the directory. (The files do not
even have to exist.) If the metadata and resources fields are
None, the directory's files are checked as usual. (This override
feature is used by the package creation tool. It should not be
used for any other purpose.)
Returns the name and version of the package that was found.
Since this changes the set of what packages are available,
it implicitly invokes clear_cache().
The external package facility is not intended to be how most
packages are loaded. Most packages should be in the collection
directory. You might add an external package in your development
workspace (while developing a soundscape), or in a temporary
download directory (while deciding whether to install a newly-
downloaded package).
- attrify_filename(self, pkg, mod, wholekey, res, filename)
- attrify_filename(pkg, mod, wholekey, res, filename) -> None
Given a filename, create a File representing it, and store the
File in the module at a location defined by wholekey. Submodules
are created as necessary. The res argument is the Resource
object associated with wholekey.
The filename must be in universal format: relative to the package
root, and written with forward slashes, not backslashes.
(If the filename is invalid or unsafe, ValueError is raised.
However, this does not check whether the file exists.)
- clear_cache(self)
- clear_cache() -> None
Erase the PackageLoader's entire knowledge of what groups and
packages are available.
This should be called whenever packages are added to, modified
in, or deleted from the collection directory.
- clear_external_packages(self)
- clear_external_packages() -> None
Remove all external directories.
Since this changes the set of what packages are available,
it implicitly invokes clear_cache().
- discover_all_groups(self)
- discover_all_groups() -> None
Search through the collection directory, and load all the PackageGroups
that are available.
This only searches the on-disk directory the first time you call
it. To force it to re-scan the directory, call clear_cache() first.
- find_all_dependencies(self)
- find_all_dependencies() -> (dict, dict, dict)
Go through the entire collection, and determine exactly which
packages depend on which other packages.
This returns a triple (forward, backward, bad). In the forward
dict, each package key (pkgname,vers) maps to an array of package
keys which the package depends on. In the backward dict, each
package key maps to an array of packages which depend *on* it.
In the bad dict, each package key maps to an array of
(pkgname,spec) of dependencies which could not be loaded.
(In the bad dict tuples, the second element may be None, a
VersionSpec, or a VersionNumber.)
This only searches the on-disk directory the first time you call
it. To force it to re-scan the directory, call clear_cache() first.
- find_item_resources(self, obj)
- find_item_resources(obj) -> (PackageInfo, Resource)
Given an object in a package module, try to find the package,
and the Resource that represents the object. If it has no metadata
defined, this returns a blank Resource object.
This tries to use object attributes such as __module__ and __name__
to identify it. This will work on classes (but not instances)
defined in a module; this covers the normal case of Agent classes.
It will also work on File objects defined in a module. Beyond
that, results are questionable.
- generate_package_path(self, pkgname, vers=None)
- generate_package_path(pkgname, vers=None) -> str
Return the pathname in the collection directory which contains the
given package group (if vers is None), or the given package
(otherwise).
This does not create the directory or check for its existence.
This is guaranteed to create a distinct pathname for every
pkgname or (pkgname, vers) pair. This is true even on case-
insensitive filesystems, which is a wrinkle, since version
strings are case-sensitive. We work around this by putting
a "^" character before each capital letter.
- import_package_content(self, pkg)
- import_package_content(pkg) -> None
Import the package's content, if it hasn't already been imported.
Warning: this method imports Python source code from the package
directory, which means it *executes* Python source code from the
package directory. Do not call this on untrusted packages.
A sound-player will have to call this, but a package manager
should not.
- list_all_current_packages(self)
- list_all_current_packages() -> list of (str, VersionNumber)
Search through the collection directory, and return a list of all the
available packages. If multiple versions of a package are available,
only the most recent will be listed.
Returns a list of (packagename, version) tuples (in no particular
order).
This only searches the on-disk directory the first time you call
it. To force it to re-scan the directory, call clear_cache() first.
- list_all_packages(self)
- list_all_packages() -> list of (str, list of VersionNumber)
Search through the collection directory, and return a list of all the
available packages.
Returns a list of (packagename, list of version) tuples. The top
list is in no particular order, but the version list will be sorted
newest-to-oldest.
This only searches the on-disk directory the first time you call
it. To force it to re-scan the directory, call clear_cache() first.
- load(self, pkgname, versionspec=None)
- load(pkgname, versionspec=None) -> PackageInfo
Load a package, given its name and a version spec.
If no second argument is given, the most recent available version
of the package is loaded. If the argument is a VersionNumber,
that version will be loaded. If it is a VersionSpec, the most
recent version that matches the spec will be loaded. A string
value will be converted to a VersionSpec (not a VersionNumber).
This generates a PackageNotFoundError if no matching package
is available. It generates PackageLoadError if the package
was malformed in some way which prevented loading.
- load_group(self, pkgname)
- load_group(self, pkgname) -> PackageGroup
Load a PackageGroup, given its package name.
This is not tremendously useful for outside users, although you
can call it if you want. A PackageGroup represents all the
available versions of a particular package.
- load_item_by_name(self, name, package=None)
- load_item_by_name(name, package=None) -> value
Given a string that names a resource -- for example,
'com.eblong.example/reptile.Hiss' -- import the module and
return the resource object.
If the string begins with a slash ('/boodle.builtin.NullAgent')
then the regular Python modules are searched. No importing
is done in this case; it is really intended only for the
contents of boodle.agent.
If the string ends with a slash ('com.eblong.example/'), then
the module itself is returned.
If the package argument is supplied (a PackageInfo object), it
becomes the package to look in for unqualified resource names
('reptile.Hiss'). If no package argument is supplied, then
an unqualified resource name raises ValueError.
- load_package_dependencies(self, pkg)
- load_package_dependencies(pkg) -> (set, dict, int)
Attempt to load all the packages which the given package depends on.
The argument should be a PackageInfo object.
This returns a triple (good, bad, count):
- good is a set containing (packagename, version) pairs for every
package that was loaded successfully. (This will include the original
package.)
- bad is a dict. The keys are packagenames which did not load
successfully. Each maps to a (nonempty) list of version requests
for that package, which could not be fulfilled. (The list contains
None, VersionSpecs, and VersionNumbers. Values may occur more than
once.)
- count is an int, representing how many actual errors occurred.
This describes package format problems and read errors. It does not
include packages that were simply not available. (The bad dict
includes both errors and not-availables; so len(bad) >= count.)
If bad is empty, then all dependencies are available.
Note that the good list may include more than one version of a
package.
- load_specific(self, pkgname, vers)
- load_specific(pkgname, vers) -> PackageInfo
Load a package, given its name and a specific version number.
(The version number may be a VersionNumber or a string that
can be converted to one.)
This is an internal call; external callers should use load().
(load_specific() is intended to be called by a package
function which has already consulted package_groups and a
PackageGroup object. If you bypass those, you might load
a package which is not part of any PackageGroup. That would
leave the cache in a confusing state.)
- remove_external_package(self, dirname)
- remove_external_package(dirname) -> None
Remove an external directory. The argument should be one that was
previously passed to add_external_package(). (If it is not a
current external directory, this silently does nothing.)
Since this changes the set of what packages are available,
it implicitly invokes clear_cache().
Data and other attributes inherited from boopak.pload.PackageLoader:
- global_loader = None
- import_recorder = None
| |