[Boost-commit] svn:boost r69097 - sandbox/local

Subject: [Boost-commit] svn:boost r69097 - sandbox/local
From: lorcaminiti_at_[hidden]
Date: 2011-02-20 14:38:28

Author: lcaminiti
Date: 2011-02-20 14:38:27 EST (Sun, 20 Feb 2011)
New Revision: 69097
URL: http://svn.boost.org/trac/boost/changeset/69097

Log:
Added TODO file.
Added:
   sandbox/local/TODO.txt   (contents, props changed)

Added: sandbox/local/TODO.txt
==============================================================================
--- (empty file)
+++ sandbox/local/TODO.txt	2011-02-20 14:38:27 EST (Sun, 20 Feb 2011)
@@ -0,0 +1,93 @@
+
+
+* Rework a no-macro version of the factorial example.
+    >>  Ask boost about inline -- is it useful?
+    >>  Tell boost that split-macros can support recursion at the cost of
+        allowing `this` within the function body (so no error if used instead
+        of `this_`).
+
+
+* Rename function_ref to function (because it now has a default constructor so it's not a reference anymore).
+
+
+* Double check all local function syntax elements (from Grammar) -- auto, register, etc.
+
+* Provide bound parameter types within the body (via private typdefs in the functor class) as `BOUND_PARAM_NAME>_type` -- these can be used for concept checking, etc. Actually, provide access to the bind types via a macro BOOST_LOCAL_TYPEOF() so to hide the type name and to strip eventual const& (Vicente's idea).
+
+* In docs that local functions cannot appear on same line (because __LINE__ is used to generate unique symbols).
+
+* Ask Paul why nil-seq are not part of pp-lib.
+
+* Try to use PP_LIST instead of nil-seq.
+
+* Ask (Paul) for interest in addition PP_KEYWORD to pp-lib.
+
+* Rename is_xxx.hpp to xxx.hpp for all PP_KEYWORD and add REMOVE and ADD macros:
+> #include <boost/preprocessor.hpp>
+> #include <boost/preprocessor/detail/is_unary.hpp>
+>
+> #define IS_AUTO_auto (1) /* must expand to unary */
+> #define IS_AUTO(tokens) BOOST_PP_IS_UNARY(BOOST_PP_CAT(IS_AUTO_, tokens))
+>
+> #define REMOVE_AUTO_STRIP_auto /* must expand to nothing */
+> #define REMOVE_AUTO_(tokens) BOOST_PP_CAT(REMOVE_AUTO_STRIP_, tokens)
+> #define REMOVE_AUTO(tokens) \
+>     BOOST_PP_IIF(IS_AUTO(tokens), \
+>         REMOVE_AUTO_ \
+>     , \
+>         tokens BOOST_PP_TUPLE_EAT(1) \
+>     )(tokens)
+>
+> #define ADD_AUTO(tokens) \
+>     BOOST_PP_EXPR_IIF(BOOST_PP_NOT(IS_AUTO(tokens)), auto) tokens
+>
+> IS_AUTO(auto int x) // 1
+> IS_AUTO(int x) // 0
+>
+> REMOVE_AUTO(auto int x) // int x
+> REMOVE_AUTO(int x) // int x
+>
+> ADD_AUTO(auto int x) // auto int x
+> ADD_AUTO(int x) // auto int x
+The names should actually be AUTO_IN_FRONT, AUTO_ADD_FRONT, AUTO_REMOVE_FRONT, AUTO_IN_BACK, AUTO_ADD_BACK, AUTO_REMOVE_BACK (the general stuff could go into keyword/utility/in.hpp, add.hpp, remove.hpp).
+
+* Add to docs that this can never be bound by ref (as in C++).
+
+* Add to docs that MSVC is not C99 because it does not support empty params (so `()` can't be used for empty params and `void` -- or `(void)` -- should be used. However, MSVC supports variadics you can use `..., ...` to separate params instead of parenthesis `(...)...`.
+
+* Add to docs that if `, default ...` syntax seems strange for C99, programmers can always #define and use the following macro (or somenthing similar).
+> #define WITH_DEFAULT(x) , default x
+
+
+
+* Try seriously to make nested locals to work...
+
+* Add to docs that object `this` can never be bound by reference `&this` (as usual in C++ it is not legal to get a reference to the `this` pointer).
+
+* Make sure that non of the binding (outside and inside the functor) code is generated when no param is bound.
+
+* Rename `function_base` to `abstract_function` (because it's abstract) and `function_ref` to `function` (as for Boost.Function because it's no longer a ref since it can be init empty and uses ptr internally).
+
+
+* Send comments on variadic_macro_data:
+
+* Optimize macro expansion code for C++0x, C++03, etc. From John Bytheway:
+> Yes, in C++0x local structs can be passed as template parameters.
+> Obviously, in C++0x there are lambdas too, so you might think your
+> library is useless, but I'd expect most compilers to support passing
+> local structs as template parameters before they support lambdas, so
+> there is some advantage in having this intermediate implementation.
+> Also, your macros would allow writing code that worked in C++0x and
+> C++03, with the virtual function business in only those compilers/modes
+> where it is necessary.
+
+* See if the two local functors `functor_factorial` and `factorial` can be declared `const`.
+
+* Make sure that when no bound params, none of the binding overhead code is generated by the macro expansion.
+
+* Add comparison with Look at Alex-Steven's syntax to docs:
+    http://thread.gmane.org/gmane.comp.lib.boost.devel/168612/focus=168694
+    >> This syntax uses same typeof trick to deduce result type.
+    >> This syntax uses a strange macro-into-macro for bind. I don't.
+
+