File.byLine ought to return dups?

Steven Schveighoffer schveiguy at yahoo.com
Fri Jun 4 12:41:43 PDT 2010


On Fri, 04 Jun 2010 15:27:03 -0400, Graham Fawcett <fawcett at uwindsor.ca>  
wrote:

> Hi folks,
>
> I expected the following program to print the lines of a file in
> reverse:
>
>     // bad.d
>     import std.stdio;
>     import std.range;
>    void main() {
>       auto f = File("bad.d");
>       foreach(char[] line; retro(array(f.byLine())))
>         writeln(line);
>     }
>
> However, this produces very unusual output: fragments of the same
> lines are printed repeatedly. I suspect it's because the byLine()
> 'generator' is not dup'ing the arrays it reads from the file.
>
> This works as expected:
>
>     // good.d
>     import std.stdio;
>     import std.range;
>    Retro!(char[][]) retroLines(File f) {
>      char[][] lines;
>       foreach(line; f.byLine())
>         lines ~= line.dup;         // note the .dup
>       return retro(lines);
>      }
>    void main() {
>       auto f = File("good.d");
>       foreach(line; retroLines(f))
>         writeln(line);
>     }
>
> If you remove the '.dup', then this behaves badly as well.
>
> So is this a bug in File.byLine, or am I just using it badly? :)

The latter.  File is re-using the buffer for each line, so you are seeing  
the data get overwritten.  This is for performance reasons.  Not everyone  
wants incur heap allocations for every line of a file ;)  As you showed,  
it's possible to get the desired behavior if you need it.  The reverse  
would be impossible.

Now, that being said, a nice addition would be to create a duper range  
that allows you to do one expression:

foreach(char[] line; retro(array(duper(f.byLine()))))

-Steve


More information about the Digitalmars-d-announce mailing list