1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
|
This carries an "aint nuffin clever in 'ere" warning: So far All I have done
is get yasvec compiling again, remove a few bugs and extend reallocate
for some cases where msize and expand are availible.
enable_if removed, its function replaced with type_selector in the case
of namespace level functions and the use of dummy Int2Type arguments
in member functions. I don't like dummy arguments but they seemed better than
the other options (namely external, explicitly templated functions 'hidden' in
an implementation namespace).
I should have kept a log of all of the bugs I pulled out and other modifications,
Indeed, I almost did but somehow it didn't quite happen - sorry.
uninitialised_copy I have defined for InputItr = const T* and InputItr = T*
seperately, this is less than desireable but previously with only the
const-data version defined any non const pointers were being eaten up by
the default templated function. This is fairly minor but I mentions
because you may disagree with me.
To do:
Cut vector back vector to a buffer then decide whether it is sensible to
re-implement vector using this buffer.
Implement valarray and related classes using yasli_nstd::buffer and other
yasli components. The structure of these classes will follow that found in
Mingw32 - expressions are to be evaluated during assignment.
Add yasli_nstd::reference_valarray, a class which behaves like
valarray and can be used in conjuction with it but which consists of an
array of pointers.
|
