[Subversion] / PEAK / CHANGES.txt  

Diff of /PEAK/CHANGES.txt

Parent Directory | Revision Log

version 1122, Sun May 18 00:03:27 2003 UTC version 1350, Tue Aug 19 23:07:30 2003 UTC
Line 1 
Line 1 
   Fixes and Enhancements since Version 0.5 alpha 2
   
    Changed, Enhanced, or Newly Deprecated Features
   
    - 'naming.IName' is now derived from 'binding.IComponentKey', so names and
      addresses must now support the 'findComponent()' method.  All PEAK name and
      address types provide support for this.
   
    - The 'lookup()' method of 'binding.IComponentKey' is now called
      'findComponent()', to better distinguish it from 'lookup()' in
      'naming.IBasicContext', which does something very different.
   
    - 'binding.bindTo()' and 'binding.bindSequence()' now pre-adapt their
      arguments to 'IComponentKey', to speed lookups at runtime, and to ensure
      that errors due to an unusable parameter type occur at class creation time
      instead of waiting until lookup time.
   
    - The following 'binding' forms are now deprecated, and will go away before
      0.5 beta is released:
   
      'bindToProperty(x,y)' -- use 'bindTo(PropertyName(x),default=y)'
   
      'bindToParent()' -- use 'bindTo("..")'
   
      'bindToSelf()'  -- use 'bindTo(".")'
   
      'bindToUtilities()' -- no replacement; let me know if you're using this.
   
    - 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  Fixes and Enhancements since Version 0.5 alpha 1
   
  Changed, Enhanced, or Newly Deprecated Features   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   - Added the UML 1.4 metamodel, and thus the ability to load UML 1.4
    models encoded in XMI 1.1.     models encoded in XMI 1.1.
   
  - Added support in the mof2py code generator for "unprefixing" enumerated   - Added support in the mof2py code generator for "unprefixing" enumerated
    values, so that UML and other metamodels' enumerations work correctly     values, so that UML and other metamodels' enumerations work correctly
    when loading from XMI.     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.Acquire()' now accepts a 'default' value argument, and
    'binding.New()' no longer accepts the 'bindToOwner' flag.     'binding.New()' no longer accepts the 'bindToOwner' flag.
Line 100 
Line 405 
    about this, as they are part of the naming package's extensibility     about this, as they are part of the naming package's extensibility
    framework.)     framework.)
   
  - 'EigenRegistry' and 'PropertyMap' no longer attempt to figure out whether  
    implied (i.e. inherited) interfaces are more or less general with respect  
    to a previous registration.  This was behavior that emulated Zope adapter  
    registries, but what we really wanted was more akin to a Zope "type"  
    registry.  The only parts of the test suite that used the old behavior  
    were the tests specifically written to ensure that behavior!  
   
  - 'binding.bindTo()' now accepts a 'default=' argument, whose value will be   - 'binding.bindTo()' now accepts a 'default=' argument, whose value will be
    used in case of a 'NameNotFound' error.     used in case of a 'NameNotFound' error.
   
Line 168 
Line 466 
   
  Corrected Problems   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 source distributions missing essential setup files
   
  - Fixed a problem with assembly events, where a parent component that didn't   - Fixed a problem with assembly events, where a parent component that didn't
Line 175 
Line 482 
    if they requested the notification after the parent had already received     if they requested the notification after the parent had already received
    it.     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   - Fixed 'naming.lookup()' and related APIs not setting the parent component
    of created objects without an explicitly supplied 'creationParent' keyword     of created objects without an explicitly supplied 'creationParent' keyword
    argument.  This used to "sort of work" when we had implicit configuration     argument.  This used to "sort of work" when we had implicit configuration


Generate output suitable for use with a patch program
Legend:
Removed from v.1122  
changed lines
  Added in v.1350

cvs-admin@eby-sarna.com

Powered by ViewCVS 1.0-dev

ViewCVS and CVS Help