Re: Confusing result from std::string::length()
On Jun 5, 12:48 pm, bork <koznobik...@gmail.com> wrote:
On 4 8N=, 22:46, Le Chaud Lapin <jaibudu...@gmail.com> wrote:
Then I saw that ICU was:
2. using strange modeol for synchronization, testing if state is
mutable, etc. "Ok..maybe I can work around that."
What strange model of synchronization do you mean?
Having trouble find it. It had to do with speccial functions that
programmer must call
3. using type UBOOL as bool, being a typedef for int8_t, which causes
problems with my type system. "Hmm.."
Could you please elaborate a bit on more on the problems it causes?
My type system only recognizes the 14 fundamental C++ scalar types:
char, int, ....long double. This keeps the vast majority of my headers
strictly portable, where strict portability means that the programmer
cannot discern from the .hpp file what is the target platform. This is
generally true also for the majority of .cpp files. By having a
separate UBOOL type that is actually something else, in this case,
int8_t, which is...?...I would have to break this rule and put #ifdef
PLATFORM_YADA..., which, so far, has not been necessary.
4. No use of exceptions. "Wow!, red flag but...still might be OK..."
Then I saw this:
3. UBOOL UnicodeString::IsBogus()
Isn't it the same as "4. No use of exceptions"? "Wow!, red flag
but...still might be OK..."
Yes, I guess, but I made it separate because, generally speaking, not
using exceptions is not necessarily a reason to reject a library. It
is the consequence of not using exceptions that is grounds for
rejection. Also, ICU is a large library. If the reason for not using
exceptions is to accommodate non-conformant compilers, than that would
indicate that this mindset might manifest in other, not-yet-seen
places.
I do not accept the excuse that the library got it start during a
period when exceptions where not available on all supported compilers.
At some point, it is better to just focus on the code and get it
(mostly) right, and if a compiler is still so poor that it cannot
support something as fundamental as exceptions, forsake that compiler.
Why do you use C++ after all? Why don't take something modern with all
bells and whistles (including good Unicode support) like Scala? If
some platform don't support something as fundamental as JVM, at least
16 GB of RAM and 4 cores, forsake that platform. I hardly believe you
really think that it's always possible to "forsake that compiler".
Not always, but exceptions are fundamental, in my book. By contrast,
if I discovered that Visual Studio did not support trigraph sequences,
would be no big deal. Exceptions incfluences the thought process.
There is a certain form that is not reachable without them.
I started anew with my own UNICODE string class.
So, did you throw away UnicodeString or the hole ICU? If "UNICODE"
string is all you need you could try basic_string<UChar> (and probably
specialize char_traits for UChar type).
Since UnicodeString was the essential element, I discarded all of ICU.
Actually, I am using ICU as a crutch for my own class. I looked at the
internal state of UnicodeString to confirm that I was not omitting
anything critical in my own class. Turns out I was not. I am using my
own type of serialization. I am restricting my code to the 14 C++
scalar types (no typedefs against scalars). Lot of little things,
each of which can be added incrementally and opportunistically.
The biggest relief is knowing that, when I am done, I will not have
a ::IsBogus() member function.
-Le Chaud Lapin-
--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]