Re: KI_FILE count precondition

On Thu, 18 Jan 2007 17:30:19 +0100, Eric Bezault <ericb-D6Qt/9opevxWk0Htik3J/[email protected]>  
wrote:

> Bernd Schoeller wrote:
>> Can anybody motivate for me the reason of the 'is_closed' precondition  
>> in
>> KI_FILE?
>
> If the file is not closed, then we might be writing
> to it. The new characters may still be in the buffer
> or already on disk. What does `count' means in that
> case?

Well - on an open file, the count should be exactly what is visible  
through the open file handle. I understand that there are potential  
problems with concurrent access to the same file, but I do not see how  
these might get resolved by requiring only to measure the size of a closed  
file.

> If you want to know the size of a file on disk,
> use KL_SHARED_FILE_SYSTEM.file_system.file_count.

The code in question reads the content of file into a string, a common  
task that should be reasonably easy to implement.

Currently, it looks as follows:

file.reset(filename)
filesize := file.count
file.open_read
if file.is_open_read then
   file.read_string(filesize)
   s := file.last_string
   file.close
end

This looks a little complex to me and still has problems. I am a little  
afraid to just call

   file.read_string(Platform.Maximum_integer)

as, looking at the implementation in KL_TEXT_INPUT_FILE, this might create  
a buffer of that size.

It would be best if there was a 'read_string(-1)' that would just read in  
everything available.

Bernd