"David Abrahams" <dave@boost-consulting.com> wrote
"Arkadiy Vertleyb" <vertleyb@hotmail.com> writes:
parse the actual typeof expression. Unfortunately it completely blows up the parser (probably an internal limit is exceeded) and doesn't leave me with much error reporting. Is there a configuration option to scale down the size of the MPL-vector used for the encoding ?
BOOST_TYPEOF_LIMIT_SIZE
By default it's 50 -- try to set it to 15 -- should be enough to compile main.cpp (but not test_compliant.cpp).
I don't understand all the issues, but I get the impression that BOOST_TYPEOF_LIMIT_SIZE isn't a good analogue for BOOST_MPL_LIMIT_XXX_SIZE because most of the time you're not supposed to change the latter. Should we be changing one of these names, or maybe do something to clearly advertise any differences?
Well, it's not exactly an analogue. Typeof uses only numbered (or what is the correct name for it?) versions of the mpl::vector. Setting BOOST_TYPEOF_LIMIT_SIZE to, e.g., 15 means that vector15 is used to store the encoded type, and numbered versions up to vector15 are used to build it, regardless what the value of BOOST_MPL_LIMIT_VECTOR_SIZE is. If typeof limit happens to exceed the mpl limit, additional numbered MPL vector definitions are built with PP lib, as suggested by Aleksey some time ago. To add a little to the confusion, by default I don't use mpl::vector. Instead I use my own very simple MPL-compatible vector definition, which allows me to avoid the gcc-mpl::vector performance problem discussed some time ago. To really use MPL, BOOST_TYPEOF_USE_MPL_VECTOR has to be defined. This may be reconsidered after the MPL problem is fixed. Most of the time you wouldn't have to change BOOST_TYPEOF_LIMIT_SIZE either. I would use BOOST_MPL_LIMIT_VECTOR_SIZE as a convenient default, but I think it's currently 10 (?) and too low for typeof... Do you have a better name in mind or is it enough to just put the above description somewhere in the document? Regards, Arkadiy