views:

207

answers:

3

Hi,

I'm trying to create a file which has the following structure:
- Each line has 32 bytes - Each line looks like this format string: "%10i %3.7f %3.7f\n"

My Problem is the following: When i have a negative floating point numbers the line gets longer by one or even two characters because the - sign does not count to the "%3.7f".

Is there any way to do this more nicely than this?

if( node->lng > 0 && node->lat > 0 ) { 
    sprintf( osm_node_repr, "%10i %3.7f %3.7f\n", node->id, node->lng, node->lat );
} else if (node->lng > 0 && node->lat < 0) {
    sprintf( osm_node_repr, "%10i %3.7f %3.6f\n", node->id, node->lng, node->lat );
} else if (node->lng < 0 && node->lat > 0) {
    sprintf( osm_node_repr, "%10i %3.6f %3.7f\n", node->id, node->lng, node->lat );
} else if ( node->lng < 0 && node->lat < 0 ) { 
    sprintf( osm_node_repr, "%10i %3.6f %3.6f\n", node->id, node->lng, node->lat );
}

Thanks for your Answers,
Andreas

+1  A: 

I wouldn't use sprintf() for this. It looks like something that belongs in Node::operator>>(ostream&).

MSalters
This. The question is tagged C++, and I'm not sure why the OP is writing C code here.
DeadMG
I'm using this code in a C++ program so I tagged the question with C++ but obviously this is wrong :-). Thanks
Kungi
+1  A: 

You can force the sign to be printed, too with the + prefix: printf( "+3.6f", 1.0 ); will result in a fixed size printout.

(courtesy to the handiest printf documentation I ever saw).

xtofl
Thank you that is exactly what i was looking for.
Kungi
+1  A: 

Why not just write a binary file, where the sign bit isn't a concern? Added benefit is that (in general) you'll be writing 12 bytes vs 32 for each item (a green solution :-). After all, you lose precision when doing a sprintf and an atof on the other side.

If this isn't viable,

int len_lng = lng < 0 ? 6: 7;
int len_lat = lat < 0 ? 6: 7;
char fmt[128];
sprintf(fmt, "%%10i %%3.%df %%3.%df\n", len_lng, len_lat);
sprintf(osm_node_repr, fmt, node->id, node->lng, node->lat);
xcramps
I agree strongly with losslessly writing 'valuable data'. +1!
xtofl
Writing a binary file is definitely worth consideration. I didn't even think of that. Thanks for the idea.
Kungi
Writing floats in a binary file is fine as long as the file is always read by the same program. Also, you can use '*' in your format to allow passing of the width or precision as an argument.
tomlogic