Re: [Monotone-devel] some observations about roster node IDs

[Zack Weinberg]

On 6/28/07, Derek Scherger <address@hidden> wrote:
> I'd be interested in the crazy one-liners... I'm sure they'd be
> interesting in their own right.

Well, I'll be helpful, and not write it all on one line for you:

mtn automate select b:net.venge.monotone |
 mtn automate toposort address@hidden |
 (i=1; echo roster type nid
  while read id
  do mtn get_roster $id | sed -ne '
    /file/{
       s/^ *//
       s/ ".*$//
       h
   }
   /dir/{
       s/^ *//
       s/ ".*$//
       h
   }
   /ident/{
       s/^ *ident "//
       s/"$//
       H; g; y/\n/ /; p
   }' | sort -k 1,1 -k 2,2n | sed "s/^/$i /"
  i=$(($i + 1))
  done) > nvm-nids.txt

Replace the first two lines with just

 mtn automate select '' |

if you want the full data set, not just net.venge.monotone.  The
result has three space-separated, labeled columns: an arbitrary number
for the roster, the word 'dir' or 'file' indicating what kind of node,
and the node ID.  This (not accidentally) is the format expected by
R's read.table(, header=T).  You'll want gigs of RAM in order to do
anything useful with it without waiting hours for swap.

> I'd also be interested in what you are doing in terms of profiling. I
> suspect that my use of raw gprof and oprofile output is rather primitive
> but never seem to find time to improve on it. Any chance you would like
> to write up a quick summary of what you're doing and stick it in
> notes/README.profiling or somewhere. Then anyone else who has other
> ideas has somewhere to add them. ;)

I don't have any clever techniques either; oprofile still won't work
for me, so I've just been running
valgrind --tool=callgrind on the scary fast amd64 box and waiting a
couple hours.

zw