~mwhudson/pypy/imported-builtinrefactor : changes

~mwhudson/pypy/imported-builtinrefactor » Changes from revision 846

From Revision 846 to 827

Rev	Summary	Authors	Date
846	the final merge of the builtinrefactor branch into the trunk... the final merge of the builtinrefactor branch into the trunk. See http://codespeak.net/pipermail/pypy-dev/2003q3/001012.html for in-depth discussion and description of what is new.	hpk	20 years ago
845	A version of fixeol that finds .txt files as well, and can ... A version of fixeol that finds .txt files as well, and can correct existing files on your working copy that have inconsistent line endings (This is a replay in the builtinrefactor branch of the trunk revision 1337.)	arigo	20 years ago
844	Still some light refactoring of the multimethod mecanisms. Still some light refactoring of the multimethod mecanisms. Now we give enough static clues about the types to allow for a complete list of what-to-try-in-which-order to be built for each input argument types. This cleanly separates method selection from method call, as wished by mwh and Samuele. It should also speed up pypy a bit but not dramatically enough to notice I'm afraid. (This is a replay in the builtinrefactor branch of the trunk revision 1131.)	arigo	20 years ago
843	one last hack to have string-% behave in stdobjspace one last hack to have string-% behave in stdobjspace	arigo	20 years ago
842	minor check minor check	arigo	20 years ago
841	- fixed two places to run with python-2.2 - fixed two places to run with python-2.2 M pypy/interpreter/extmodule.py - fixed the introspection loop to work with python-2.2 - minimized it a bit M pypy/interpreter/baseobjspace.py - early initialize w_builtins to a dummy value in order to allow frame creation (which needs space.w_builtins) during sys/builtin init. M pypy/module/builtin.py - allow generators for the 'compile' builtin by default to allow tests working on python-2.2	hpk	20 years ago
840	Big "final" check-in in this branch. Big "final" check-in in this branch. Individual comments about changes will be made during the merge.	arigo	20 years ago
839	test failures reduced test failures reduced	arigo	20 years ago
838	interpreter tests working again interpreter tests working again	arigo	20 years ago
837	getting closer, some tests start passing again getting closer, some tests start passing again	arigo	20 years ago
836	Still nothing that works, far from it, but progress neverthe... Still nothing that works, far from it, but progress nevertheless (I think).	arigo	20 years ago
835	More intermediate check-ins, after discussion with hpk... More intermediate check-ins, after discussion with hpk...	arigo	20 years ago
834	This is a middle-work check-in. Nothing works. This is a middle-work check-in. Nothing works. A interpreter/eval.py New abstract base classes Code and Frame A interpreter/pycode.py M interpreter/pyframe.py Concrete subclasses of Code and Frame D interpreter/opcode.py AM interpreter/pyopcode.py A interpreter/pyfastscope.py A interpreter/pynestedscope.py Opcodes moved into three seperate files defining a subclass of PyFrame each M interpreter/gateway.py M interpreter/baseobjspace.py Various changes here, work in progress A interpreter/function.py The one and unique Function class, with argument parsing M interpreter/executioncontext.py AM interpreter/error.py OperationError moved to its own module	arigo	20 years ago
833	- eliminated unused 'pycode.py' (forgot it last time) - eliminated unused 'pycode.py' (forgot it last time) - got rid of two-phase initialization by dissallowing app-level class definitions in '__builtin__' initialization. We don't want to invoke the interpreter for making the __builtin__ module as the interpreter needs it to function. - cleaned up uses of app2interp in some places. app-level methods now don't take a 'self' argument because it doesn't make sense to have them at app-space. app2interp and InterpretedFunction don't care about bound/unbound methods. These are/should be an implementation detail of the objspaces. - the branch still only runs on trivial objspace.	hpk	20 years ago
832	- gigantic intermediary checkpoint-checkin - gigantic intermediary checkpoint-checkin this checkin is mainly for internal 'checkpointing' reasons please don't look at it too closely until you want to participate in the refactorings. it's definitely neccessary to cleanup the refactored code although it's beginning to work ... - trivial object space works with a more unified approach to functions, code and wrapping app-level stuff into interpreter-level code. StdObjSpace fails because multimethods deeply depend on interpreter/pycode.py depends deep - moved some tests from appspace/test to interpreter/test where they belonged (test_exec stdspace and annspace fail miserably (don't even try them) - got rid of 'extmodule' and builtin_app stuff - revamped PyFrame initialization (now a 'ScopedCode' invokes either 'setdictscope' or 'setfastscope' to set the local scope of the frame. the frame doesn't invoke any analysis itself anymore. - introduced a 'ScopedCode' class which takes a space, a cpython code object, global and closure scope. Currently you call 'eval_frame(w_locals)' to trigger interpretation of the interp-level defined function. - introduced an 'InterpretedFunction' which takes a space, a function object, global and closure scope. Currently you call 'eval_frame(w_args, w_kwargs)' to trigger interpretation of the interp-level defined function. The arguments are parsed with another app-level function 'parse_args_complex' which is a rewrite (50% code) of 'decode_code_arguments' - introduced an 'AppVisibleModule' class that allows to easily define builtin modules (which can access interp-level stuff). See the doc-string for more info. - all the previous new classes live live in 'interpreter/gateway.py' which seems to be a good name to define stuff that 'connects' interp-level and app-level execution. - disabled 'mod' and 'divmod' complexobject-tests because they fail in python-2.3 (which also issues deprecation warnings) - improved the tool.test to skip non-importable modules (and print a warning) so much for now.	hpk	20 years ago
831	- moved build_class from opcode_app to opcode.py - moved build_class from opcode_app to opcode.py - initialized co_lnotab, co_firstline, co_filename in AppCode to get better tracebacks - little tweak to use w_builtins instead of w_builtin during basobjcspace-initialization	hpk	20 years ago
830	- fixed wrong cloning of exceptions - fixed wrong cloning of exceptions (try raise ValueError on the cmdline in the previous revision) - excluded 'xrange' from the list of "imported" types from __builtin__ probably 'standard_types' should be defined more explicitely...	hpk	20 years ago
829	various little (cosmetical) cleanups to make the code more r... various little (cosmetical) cleanups to make the code more readable	hpk	20 years ago
828	- got rid of pyframe_app, and pyframe_app by introducing a n... - got rid of pyframe_app, and pyframe_app by introducing a new mechanism for exposing app-level defined methods at interp-level. This is not really perfect but i begins to work so i check it in. the mechanism is invoked like so (it's not perfect, i know) class someclass: def app_method(self, fixed, number, of, args): # do app-level stuff, don't use 'self' for the time beeing # examples: decode_code_arguments, normalize_exception method = app2interp(app_method) def othermethod(self): # 'self.space' must contain an objectspace instance self.method(fixed, number, of, args) - there is a new (experimental, of course!) approach to executing code objects. Actually the "app2interp" method when called makes up an "AppBuiltinCode" instance which is a simple form of a code object. Instead of the whole "build_arguments" to-dict-locals- and-then-get-it-out-of-locals business it constructs arguments in a straightforward manner. Thus it's currently not possible to pass anything like the exact number of arguments (no defaults, etc.) It probably makes sense to extend argument-parsing by using Michael's extmodule's subclassing technique for handling various argument-passing cases. Not that AppBuiltinCode creates a new kind of frame called "AppFrame" which subclass PyFrame but has an easy initialization (see above) - moved tests to applevel and added one for execfile	hpk	20 years ago
827	fixed a typo fixed a typo	hpk	20 years ago

Older »