[Subversion] / PEAK / CHANGES.txt  

View of /PEAK/CHANGES.txt

Parent Directory | Revision Log
Revision: 1345 - (download)
Tue Aug 19 16:21:08 2003 UTC (20 years, 7 months ago) by pje
File size: 25248 byte(s)
Fix problem found by Yaroslav Samchuk: The logging system would raise an
error if a log message with no arguments contained a '%' sign.
Fixes and Enhancements since Version 0.5 alpha 2

 Changed, Enhanced, or Newly Deprecated Features

 - There's a new 'peak.storage.files' module, with handy classes like
   'EditableFile'.  'EditableFile' is a class that lets you edit the contents
   of a file "in place", with atomic replacement of the original at transaction
   commit.  If the transaction is aborted, the original file is left unchanged.

 - 'peak.running.scheduler.UntwistedReactor' now supports a configuration
   property ('peak.running.reactor.checkInterval') to determine how long it
   should run 'select()' calls for, when there are no scheduled tasks.

 - 'peak.running.scheduler.UntwistedReactor' now supports using a "signal
   manager" component (via the 'peak.running.signalManager' property) to
   process signals while a 'run()' loop is in progress.  Signal managers can
   also be added or removed at any time via the new 'peak.util.signal_stack'
   module.

 - 'peak.running.commands.AbstractCommand' now offers a '_run()' method that
   can be overridden in subclasses, instead of 'run()'.  If you override the
   new '_run()' method instead, you get the advantage of automatic handling
   for invocation errors and 'SystemExit' exceptions, provided for you by the
   'run()' method.

 - There is now a 'storage.DMFor(class)' function that returns a configuration
   key for registering or looking up data managers by class.  You can use the
   returned key in a binding's 'offerAs' list, or as the target of a 'bindTo()'
   or 'lookupComponent()'.  The key is constructed using
   'config.ProviderOf(storage.IDataManager,class)'.

 - You can now register utilities that are keyed by the combination of an
   interface and one or more classes, using 'config.ProviderOf(iface,*classes)'
   as a configuration key.  Providers are registered under an '(iface,klass)'
   tuple for each specified class, and looked up using the MRO order of the
   class specified for lookup.  In other words, searching for a provider will
   find a provider for the requested class, or one of its base classes, with
   precedence given to the more-specific provider.  This is primarily intended
   for services like data managers and Specialists.

 - It's now possible to extend .ini file parsing with custom section types,
   and PEAK defines its own built-in section types using this extension
   mechanism.  Custom section types must include at least one space, (e.g.
   '[My Section]') or they will be treated as a plain property name.
   See the 'peak.config.iniFile.sectionParsers' section in 'peak.ini' for
   more details, along with the 'config.ISettingParser' and 'config.IIniParser'
   interfaces.

 - When creating a 'PropertyName()', it's now possible to force conversion of
   invalid characters to '_', using the 'PropertyName.fromString()'
   constructor.  (Note that the input must be a plain-ASCII string.)  Unless
   you request that wildcards ('?' and '*') be kept, they will also be
   converted to '_' characters.  This can be convenient for converting things
   like filenames or text that might contain spaces, to property names.

 - It's now possible to declare an attribute as offering a wildcard property;
   such lookups now follow the same rules as other wildcard property lookups.
   The 'config.IConfigKey' interface has been changed to cleanly support
   implied keys at both registration and lookup time, so you can implement
   your own key types that work the way interfaces or property names do for
   configuration lookups.

 - The 'EigenRegistry' class has been moved from 'peak.util.EigenData' to
   'peak.config.registries', as it hasn't really been useful outside PEAK for
   a while now.

 - .ini files now support "smart property" objects ('config.ISmartProperty').
   If a property rule defined in an .ini file evaluates at runtime to an object
   that implements 'ISmartProperty', the object will be given a chance to
   compute a value for the property, in place of being used itself.  This helps
   to simplify definition of complex property rules in .ini files, by allowing
   the use of helper classes.  Also, 'naming.LinkRef' and 'naming.Reference'
   (indirectly) support this interface, so you can now use them in .ini files
   to refer to an object via the naming system.  (Previously, 'naming.LinkRef'
   wouldn't do the right thing unless the property was looked up via a
   'config:' URL, and 'naming.Reference' didn't exist.)

 - 'peak.util.imports.whenImported()' can now be used even when the specified
   module has already been loaded.

 - The naming system no longer has 'objectFactories' and 'stateFactories' as
   utilities; they have been replaced with new mechanisms involving adaptation.
   Previously, addresses had a 'retrieve()' method that could be used to
   retrieve the object defined by the address.  Now, to retrieve an object for
   an address, you must either define a context that processes the address, or
   the address must have a 'defaultFactory' attribute, which provides a name
   to be imported to get an 'IObjectFactory' that can construct the referenced
   object.  (This is simpler than it sounds; for URLs that reference
   ManagedConnections, for example, all you need to do is provide the fully
   qualified name of the connection class.)

   Meanwhile, writable naming contexts must have a 'serializationProtocol'
   attribute, specifying what interface an object should be adapted to before
   attempting to store it in that context.

   The naming system no longer processes the 'creationName' keyword argument;
   this is now considered the sole responsibility of 'peak.binding'.  The
   'IComponent.lookupComponent()' method still accepts the keyword argument,
   and attribute bindings still handle the creation name transparently.  It is
   just not available via naming system APIs, and naming contexts no longer
   have to deal with it.

   The naming system base classes no longer use 'attrs' as an input parameter
   or return value.  If you've subclassed anything from 'peak.naming.contexts',
   note that your '_get()' methods should now just return the lookup value,
   rather than a 'state,attrs' tuple.  For most naming contexts, this just
   means you should change 'return foo, None' statements to just 'return foo'.

 - Property definition rules in an .ini file can now refer to 'rulePrefix' and
   'ruleSuffix' variables.  'rulePrefix' is a '.'-terminated string,
   representing the name the rule was defined with.  For example, if the
   rule was defined for '"foo.bar.*"', then 'rulePrefix' will be '"foo.bar."'.
   The 'ruleSuffix' will be the portion of the 'propertyName' that follows
   'rulePrefix'.  So, if looking up property '"foo.bar.baz"', then the
   '"foo.bar.*"' rule will execute with a 'ruleSuffix' of '"baz"'.  This should
   make it easier to work with hierarchical property namespaces.

 - Added simple example scripts and small applications in the 'examples'
   directory.

 - There is a new command-line namespace introspection tool, 'n2', which
   can be accessed by running 'peak n2'.  Type 'peak n2 -h' for help.

 - The PEAK_CONFIG environment variable can now list multiple files, separated
   by the platform's 'os.pathsep' (e.g. ':' on Unix, ';' on Windows).

 - It's no longer necessary to provide a '_defaultState()' implementation
   for an EntityDM: a default implementation is now supplied.

 - Added automatic installation of 'datetime' package for Python < 2.3.

 - CGI support has been moved from 'peak.running.zpublish' into
   'peak.running.commands' (for "raw" CGI/FastCGI) and 'peak.web' (for the
   PEAK high-level publishing framework).  You can use 'peak CGI someName' to
   adapt 'someName' to a 'running.IRerunnableCGI' and run it as a CGI/FastCGI.

 - There is now a 'peak.security' package, available from 'peak.api' as
   'security'.  It provides permission management functions: you can define
   abstract permissions by subclassing 'security.Permission', then create
   permission checking rules by subclassing 'security.RuleSet', and declare
   the permissions needed to access attributes of a class with
   'security.allow()'.  The test suite demonstrates a complex application
   ruleset with dynamic, data-driven permissions.

 - There is now an interface for "Active Descriptors":
   'binding.IActiveDescriptor'.  'peak.binding' now uses this interface to
   identify active descriptors, so you can now create your own.  (Previously,
   'peak.binding' used 'isinstance()' to detect active descriptors.)

 - REMOVED 'naming.ParsedURL'; it was deprecated as of 0.5 alpha 2.

 - The 'provides' keyword argument to various 'peak.binding' APIs has been
   renamed to 'offerAs', and it must be a sequence of configuration keys.
   (Previously, it accepted either a single key or a tuple of keys.)
   The signature of 'binding.Constant()' was changed as well; the first
   positional argument is now the constant value, and 'offerAs' is now a
   keyword argument.  (Previously, 'provides' was the first positional argument
   of 'binding.Constant()'.)  The 'registerProvider()' method of
   'config.IConfigurable()' also now accepts only a single configuration key,
   as does 'EigenRegistry.register()'.

   Also, all 'peak.binding' APIs now only accept positional parameters for
   items unique to that API.  Items common to multiple APIs (such as 'offerAs',
   'doc', 'attrName', etc.) should now be supplied as keyword arguments.

   Bindings also now automatically "suggest" the containing object as a parent
   component for the contained object, whenever a value is assigned to them or
   computed.  If a non-None 'adaptTo' is set on the binding, the value assigned
   or computed will be adapted to the specified protocol before the parent
   component is suggested.  'binding.New()' no longer relies on the
   'IComponentFactory' interface, but instead uses the new adapt/suggest
   mechanisms.

   Previously, parent components were only "suggested" when a binding was set
   via component constructor keyword arguments.  Now, this is done at any time
   bindings are set, but *not* for non-binding keyword arguments.  In other
   words, ordinary attributes of a component do not receive "suggested parent"
   notices, even when set via constructor keyword arguments.  If you want an
   attribute to do this, you must define the attribute with the binding API;
   e.g. via 'requireBinding()' or 'binding.Constant()'.  If you do *not* want
   a binding to suggest a parent component, use 'suggestParent=False' in the
   binding definition.

 Corrected Problems

  - The logging system would raise an error if a log message with no arguments
    contained a '%' sign.

  - There was a typo in peak.naming.arithmetic that caused homogeneous non-URL
    name subtraction to fail.

  - The default reactor supplied in 'peak.running.scheduler' would consume
    CPU continuously if it was waiting for I/O and no tasks were scheduled.

  - The 'peak.util.imports.whenImported' function didn't work.




Fixes and Enhancements since Version 0.5 alpha 1

 Changed, Enhanced, or Newly Deprecated Features

 - Added a 'shellcmd:'  URL scheme that returns a function that calls
   'os.system()' on the body of the URL.  It's intended for use as a command
   factory, as is needed by the 'URLChecker' periodic task.

 - You can now define adapters from arbitrary types to 'binding.IBindingNode',
   and thus be able to use them as part of a component hierarchy - without
   needing to directly add 'getParentComponent()' or 'getComponentName()'
   methods to them.

 - Added experimental 'invoke.c' script for POSIX-ish platforms with funky
   '#!' support, or lack thereof.

   'invoke' is designed to be used like this::

     #!/usr/local/bin/invoke peak somearg otherarg...

   This should work on most sane platforms with a long-enough commandline.
   (See "this page":http://homepages.cwi.nl/~aeb/std/hashexclam-1.html for
   details on the insanely incompatible ways different Unixes interpret '#!'
   lines.)

   The script is not currently built or installed by setup.py.  On the
   platforms it's targeted at, you should be able to build it with::

     gcc -o invoke invoke.c

   (Yes, it really is that simple of a script.)

 - Added a ZConfig schema for 'running.commands.EventDriven' applications,
   a ZConfig component definition for adaptive tasks, and a running shortcut
   called 'EventDriven'.

   It should now be possible to do this::

     #!/usr/bin/env peak EventDriven

   at the top of a ZConfig file formatted according to the new schema, and
   have it run.  There are two periodic tasks that can be configured and
   run from such a file: 'CleanupFiles' and 'URLChecker'.  'CleanupFiles' will
   delete files matching a pattern that are older than a certain age, while
   'URLChecker' will check to see if the target of a naming system URL is
   up/available/working, and if not, runs a command to restart it.  As an
   amusing demo, try specifying a 'file:' URL with a 'shellcmd:touch theFile'
   to recreate the file, then add a 'CleanupFiles' that deletes the file the
   checker looks for.  This can be hours (well, minutes) of exciting fun as you
   watch the dueling daemons undoing each others' work.

 - Added 'zconfig.schema' URL scheme that loads an enhanced ZConfig schema
   object that can act as a command line interpreter using the 'peak' script.

   To use it, run 'peak zconfig.schema:urlToSchema urlOfConfig'.  Or, add
   a line like this::

     #!/usr/bin/env peak zconfig.schema:pkgfile:some.package/schema.xml

   to the top of a configuration file, and make the configuration file
   executable.  Note that the schema specified must convert to an object
   that's usable with the commands bootstrap framework.  Also note that
   if you have a local PEAK_CONFIG file, you can add a 'peak.running.shortcuts'
   entry to shorten the URL reference in your #! line.  E.g.::

     #!/usr/bin/env peak mySchema

   will suffice if you have defined 'peak.running.shortcuts.mySchema' as
   'naming.LinkRef("zconfig.schema:pkgfile:some.package/schema.xml")'.

   There is also a 'peak ZConfig urlOfSchema urlOfConfig' variant, that was
   added to support putting '#!/usr/bin/env peak ZConfig' at the top of
   schema files, but unfortunately that's not valid XML.

 - Standardized file-based URL syntaxes (e.g logfiles and lockfiles) to
   follow RFC 1738/2396, and Python 'urllib'.  This shouldn't affect much
   besides the canonical forms of the URLs.  Added 'pkgfile:some.pkg/filepath'
   URL syntax for ease of referring to files near modules.  (A convenience
   intended mainly for referencing ZConfig schemas.)

 - Added the UML 1.4 metamodel, and thus the ability to load UML 1.4
   models encoded in XMI 1.1.

 - Added support in the mof2py code generator for "unprefixing" enumerated
   values, so that UML and other metamodels' enumerations work correctly
   when loading from XMI.  Also, mof2py no longer emits 'config.setupModule()'
   calls in generated code, as in practice they are not needed.

 - Running 'peak test' from the command line is roughly equivalent to running
   'unittest.py', except that the test suite defaults to the PEAK test suite.
   You can, however run any test suite from the command line with a dotted
   module/attribute path, e.g 'peak test foo.bar.test_suite'.

 - 'binding.Acquire()' now accepts a 'default' value argument, and
   'binding.New()' no longer accepts the 'bindToOwner' flag.

 - There is a new 'binding.IComponentKey' interface that is used to implement
   'IComponent.lookupComponent()'.  Now you can implement this interface,
   or create an adapter for it, in order to make an object usable as an
   argument to 'binding.lookupComponent()' - and therefore usable as a key
   for 'binding.bindTo()' or 'binding.bindToSequence()'.  Not that it's
   necessarily very useful to do so; you're probably better off simply
   creating a naming scheme.  But it might be useful for lookups done
   in the context of classes, since naming schemes aren't usable there.
   (It was actually added in order to factor out all the type testing that
   'lookupComponent' used to do, so it doesn't matter if it's useful for
   much else.)

 - PEAK has been refactored to avoid the use of 'isImplementedBy()' and
   similar introspection, in favor of 'adapt()'.  As a result, some
   'peak.naming' interfaces have changed.  This should not affect you
   if you are only subclassing PEAK-provided naming components and not
   implementing these interfaces "from scratch".  However, the various
   'isAddress', 'isAddressClass', 'isResolver', and 'isName' APIs have
   also been removed, as they were based on 'isImplementedBy()'.

 - REMOVED ability to use '__implements__' and '__class_implements__' to
   declare support for interfaces.  Use 'protocols.advise()' or a related
   API to do this now.  The 'protocols' package is available automatically
   from 'peak.api'.

   Similarly, the ability  to use 'isImplementedBy()' with interfaces declared
   by PEAK is REMOVED.  You can still use 'isImplementedBy()' with Zope
   interfaces, of course, but we recommend you switch to 'adapt()', which
   should work with both PEAK and Zope interfaces.

 - Replaced all use of 'zope.interface' with 'protocols' package because
   the 'protocols' package:

   * is considerably smaller and simpler than 'zope.interface'

   * produces Interface objects that can be inspected with the Python
     'pydoc' and 'help()' tools

   * supports and implements the PEP 246 'adapt()' protocol

   * transparently supports transitive adaptation - i.e. if adapter AB
     adapts from A to B, and adapter BC adapts from B to C, then an adapt(x,C)
     where 'x' is an 'A', will be implemented as BC(AB(x)).

   * Supports "open protocols" that allow you to "superclass" a protocol
     to create a subset protocol; objects that support the first protocol
     will automatically support the subset protocol.  For example, if one
     person defines a "dictionary" protocol, someone else can create a
     "read-only dictionary" protocol, and all objects supporting the
     "dictionary protocol" will be considered to implement the "read-only
     dictionary" protocol.

   * can interoperate with other interface packages, including Zope's, but
     does not require them

   * works with module inheritance (for everything but moduleProvides(), and
     we should get to that by 0.5a2)

   * lets you use Interfaces as abstract base classes (i.e., you can
     inherit from an interface and turn it into an implementation, and
     you can define default attribute values or method implementations in
     your interfaces

   * Lets you mix interface declarations from any number of frameworks and
     any number of interface types, in a single 'implements()' or
     'classProvides()'

   * uses adaptation as the fundamental approach to dealing with interfaces,
     and avoids the use of 'isImplementedBy()'.  In the *rare* case that you
     need to introspect rather than adapt, you can always call adapt() and
     check the result.  (But introspection usually means that you're using
     interfaces as a form of metadata; it's better to create an explicit
     interface that provides the metadata you seek, and adapt to that
     interface, than to use interfaces as data.)

   Most of these features are unavailable in 'zope.interface', and some have
   been declared by the Zope Pope to be unacceptable or undesirable features
   for Zope interfaces.  (Others may be available in some form in future
   versions of Zope X3.)  So, we no longer require or distribute
   'zope.interface'.

 - The signatures of the 'getObjectInstance()', 'getStateToBind()', and
   'getURLContext()' methods in the 'peak.naming' package have changed, to
   place the context or parent component as the first, non-optional argument.
   (If you don't know what these methods are for, you don't need to do anything
   about this, as they are part of the naming package's extensibility
   framework.)

 - 'binding.bindTo()' now accepts a 'default=' argument, whose value will be
   used in case of a 'NameNotFound' error.

 - DEPRECATED 'naming.ParsedURL'.  It will disappear in 0.5 alpha 3 or beta.
   It is replaced by the new 'naming.URL.Base'.  The 'naming.URL' package
   provides a new URL parsing framework based on 'peak.model'.  Upgrading from
   'ParsedURL' to 'URL.Base' is trivial for ParsedURL subclasses that used
   only the 'scheme' and 'body' fields, and in fact may not require any
   changes except for the choice of base class.  Also, the 'retrieve()' method
   of URLs is deprecated; please begin defining the 'getObjectInstance()'
   method instead.  This is to cut down a bit on the number of ways that the
   naming package spells the idea of retrieving something!

   For more complex URL classes, the '__init__' methods go away, 'parse'
   methods change slightly, and explicit field definitions (using
   'model.structField' or similar) are required.  See PEAK's 'URL.Base'
   subclasses for examples.  There is also a sophisticated parsing and
   formatting framework (see the 'peak.naming.URL' and 'peak.util.fmtparse'
   modules) that can be used in place of the old regex-based approach.

 - Added 'peak.util.fmtparse', a parsing and formatting framework, and
   integrated it with 'peak.model' so that any element type can have a
   syntax for parsing from, or formatting to, a string.

 - Added 'binding.whenAssembled(...)' as syntax sugar for
   'binding.Once(...,activateUponAssembly=True)'.

 - Removed 'LOG_XYZ' convenience functions from 'peak.api', and refactored
   'peak.running.logs' to use a PEP 282-like interface, 'running.ILogger'.
   Under the new scheme, messages must be sent to a specific entry point
   (e.g. 'self.logger.warning("foo")').  Components can bind an attribute
   directly to a logger object, or via configuration properties or utilities.
   PEAK components that do logging all define a 'logger' attribute, bound
   to a configuration property in the 'peak.logs' property namespace.  By
   a default in 'peak.ini', 'peak.logs.*' is configured to output messages
   of 'WARNING' priority or higher to 'sys.stderr'.

   For compatibility with the PEP 282 logging package, a 'logging.logger:'
   URL scheme has been added; looking up the URL '"logging.logger:foo.bar"'
   is equivalent to 'logging.getLogger("foo.bar")', unless the 'logging'
   package is not available, in which case the configuration property
   'peak.logs.foo.bar' will be looked up in the target context of the
   lookup.  Optionally, you can configure the 'logging.logger' URL scheme so
   that it only uses PEAK loggers, and never uses the PEP 282 loggers.

 - Added 'binding.metamethod()' wrapper for metaclass methods that might
   not be accessible from their instances if the instances (classes) also
   defined the method for *their* instances.  You must now use this wrapper
   on any such metaclass-defined methods, as PEAK no longer works around
   this via the 'x.__class__.foo(x,...)' trick that was used previously.
   In particular, if you have metaclass definitions of 'getParentComponent',
   '_getConfigData', 'getComponentName', or 'notifyUponAssembly', you need
   to wrap them with 'binding.metamethod' now.

 - Made 'NOT_GIVEN' and 'NOT_FOUND' recognizable by humans (they 'repr'
   and 'str' to their names) and by Python (they can be pickled, and
   when restored they come back as the same object).


 Corrected Problems

 - Fixed a problem in ZConfig 'schema.dtd'; I used 'PCDATA' where I should've
   used 'CDATA'.

 - Fixed a problem with 'binding.supertype()' not working correctly if the MRO
   it was searching contained a "classic" class.  Now 'supertype()' skips any
   classic classes it finds.  (It probably should be rewritten entirely.)

 - Fixed misc. problems with 'fromZConfig()' component constructor

 - Fixed source distributions missing essential setup files

 - Fixed a problem with assembly events, where a parent component that didn't
   need assembly notification, wouldn't ever notify its children of assembly
   if they requested the notification after the parent had already received
   it.

 - Fixed a bug in automatic metaclass generation that caused extra unneeded
   metaclasses to be generated.

 - Fixed 'naming.lookup()' and related APIs not setting the parent component
   of created objects without an explicitly supplied 'creationParent' keyword
   argument.  This used to "sort of work" when we had implicit configuration
   parents, but was broken when we went "all explicit" for 0.5 alpha 1.

 - Fixed a problem where initializing single-valued immutable fields of
  'peak.model' types did not perform type/value normalization.

 - Fixed a problem where bindTo would use the attribute name as the
   default value for a lookup, if the requested name/property/utility
   was not found.

 - Fixed 'mof2py' generator script not working

 - Fixed model.Element not getting parent component set when passed as a
   constructor argument.

 - Fixed property/utility lookups not working correctly on model.*
   objects.

 - Fixed IndentedStream generating all-whitespace lines


cvs-admin@eby-sarna.com

Powered by ViewCVS 1.0-dev

ViewCVS and CVS Help