Lars Finsen wrote:
> Den 8. feb. 2007 kl. 15.21 skrev Paul Herring:
>> It is very rare that you should be using EOF functions directly in a C
>> or C++ program - you should instead be checking for an error in the
>> reading of data from a file. Then, and only then, should you be using
>> the EOF functions to determine if indeed it is the end of the file (or
>> someone's removed the floppy disc from the drive etc.)
>
> Why is that? I've always used it when reading from a text file.
> Checking for other errors usually isn't necessary.
>
>> Show your code!
>
> It's very simple. Essentially it goes like this:
>
> ifstream(filename) ui;
> while(!ui.eof())
> {
> if(baner==NULL) baner = b = new bane;
> else { b->neste = new bane; b=b->neste; }
> ui >> b->fork >> b->type >>....>> b->ist;
> }
> ui.close();
>
> It reads all the actual lines correctly, but then goes on to read
> endless null lines. The file does contain linefeeds, but as far as I
> have understood, these are skipped when the skipws flag is set, which
> (as far as I have understood) it is by default.
>
> I am trying to convert from a half-automated system using MS Excel by
> converting the Excel table to text and reading the text file into the
> C++ program. Not quite decided on how to store the data in the C++
> system yet. It consists of several tables, corresponding to several
> classes in C++.
>
> LEF
Lars, have you looked at using COM to completely automate whatever
process you are doing? Excel has a pretty extensive COM automation
system exposed...although it uses IDispatch, which isn't C/C++ friendly
(IDispatch is VB-friendly, though). Then you wouldn't have to deal with
export/import issues.
I would recommend BString (from Safe C++ Design Principles - see
c-prog's Files section) but, IIRC, you said something about Borland C++
in your first post. BString depends on Block which, in turn, uses
'placement new' and 'placement delete'. The latter are two advanced and
dangerous C++ concepts, which Borland's compiler didn't fully support
the last time I checked. If you try it and it _does_ work, you could do
this:
size_t z;
BString Data, CurrLine;
Block<BString> SplitData;
List<bane> BaneList;
bane b;
if (!Data.LoadFromFile(filename))
{
// Barf an error message.
return 1;
}
z = 0;
CurrLine = Data.LineInput();
while (z < Data.ftell())
{
// Process line.
CurrLine.Explode(SplitData, '\t');
b.fork = SplitData[0];
b.type = SplitData[1];
...
b.ist = SplitData[???];
BaneList += b;
z = Data.ftell();
CurrLine = Data.LineInput();
}
(NOTE: LoadFromFile() dumps the entire file into memory. A good enough
solution for files smaller than 1MB...perhaps up to 2MB. Then again,
the code isn't exactly designed for performance as it is to just simply
work as a quick-n-dirty first pass solution. Don't know why you need a
linked list, but you shouldn't be rolling your own.)
--
Thomas Hruska
CubicleSoft President
Ph: 517-803-4197
*NEW* VerifyMyPC 2.0
Change tracking and management tool.
Reduce tech. support times from 2 hours to 5 minutes.
Free for personal use, $10 otherwise.
http://www.CubicleSoft.com/VerifyMyPC/
