[Nauty] Warning about LOCALEs in Unix/Linux

[Sterten at aol.com]

I don't know, how things are under Unix/Linux.
Nor do I know anything about
LC_COLLATE
LANG
"en_US.UTF-8"
locales
locale "C".
LC_MESSAGES

For sorting in DOS,Windows I have always been using the free utility
"RPsort" since the Microsoft sort was not very good.
RPsort is fast, has many features but has the small disadvantage
that it needs 3*x free HD-space to sort a file of length x.

Now, for the posted problem , usually I would generate the 
60000 canonical forms -even maybe with n*n Bytes per graph-
then run the output through RPsort and write the sorted
list to a new file. Then I have another program to count the
equal entries in a sorted list.
Well, this is all done from a batch-file zae.bat, which I use often
for such similar tasks :

canonize list1 > list2
call zae list2

However, when the graphs are too long (this rarely happens with
those large HDs these days) then I'd use a "twisted" checksum
program, which calculates a cryptic number of variable length
to identify each line in the file (with high probability) :

canonize list1 > list2
checksum list2 > list3
call zae list3

On HDs it could be a bit slow to access several big files
at the same time, so I sometimes use a ramdrive for this.
Or 2 different HDs for input and output.


Guenter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.anu.edu.au/mailman/private/nauty/attachments/20040203/16102181/attachment.html