masses/runGA

#!/bin/sh

# TODO: add FPRATE instead of HAM_PREFERENCE
# set SCORESET

# must use a / in the arg to a 'source' command to avoid searching the PATH
. ./config

LEARN_RATE="${LEARN_RATE:-2.0}"

NAME="set$SCORESET"
# TODO: add $FPRATE instead of HAM_PREFERENCE
LOGDIR="gen-$NAME-$HAM_PREFERENCE-$THRESHOLD-$EPOCHS-ga"

# ensure sandbox T_ rules aren't used in the GA and don't appear in output
KILL_SANDBOX_RULES=y

###########################################################################

[ -d gen-cache ] || mkdir gen-cache     # a cache, woo

if [ "$NOTE" != "" ]; then
	LOGDIR="$LOGDIR-$NOTE"
fi

if [ "x$1" = "x" ]; then

# -------------------------------------------------------------------------
# Initial rescoring

if [ ! -f "ORIG/ham-$NAME.log" -o ! -f "ORIG/spam-$NAME.log" ]; then
	echo "Couldn't find logs for $NAME" >&2
	exit 1
fi

(	# log this

set -x	# trace commands to the log

# Create a directory to organize the logs with this group of settings
mkdir -p $LOGDIR $LOGDIR/NSBASE $LOGDIR/SPBASE

if ! [ -d $LOGDIR ] ; then
  echo "Failed to mkdir $LOGDIR, dying" 1>&2
  exit 1
fi

# This should be in here instead.  Prevents testing.
# svn revert ../rules/50_scores.cf

rm -rf tmprules
cp -r ../rules tmprules

cp tmprules/50_scores.cf orig_scores.cf

# fix all scores to non-zero (avoid a possible bug, not quite sure)
./enable-all-evolved-rules < tmprules/50_scores.cf \
	> tmprules/50_scores.cf.new || exit 1
mv tmprules/50_scores.cf.new tmprules/50_scores.cf

[ $KILL_SANDBOX_RULES = y ] && rm tmprules/70_sandbox.cf

echo "[Doing a scoreset $SCORESET score-generation run]"

# Clean out old runs
echo "[Cleaning up]"
rm -rf spam-test.log ham-test.log spam.log ham.log \
	NSBASE SPBASE tmp freqs perceptron.scores \
	garescorer.scores
make clean

# Generate 90/10 split logs
# keep the *-split*.logs in cwd so it's cacheable
echo "[Generating 90/10 split ham]"
perl tenpass/split-log-into-buckets-cached \
    9:gen-cache/ham-split9.log 1:gen-cache/ham-split1.log ORIG/ham-$NAME.log
ln -f gen-cache/ham-split9.log $LOGDIR/NSBASE/ham.log
ln -f gen-cache/ham-split1.log $LOGDIR/NSBASE/ham-test.log

echo "[Generating 90/10 split spam]"
perl tenpass/split-log-into-buckets-cached \
    9:gen-cache/spam-split9.log 1:gen-cache/spam-split1.log ORIG/spam-$NAME.log
ln -f gen-cache/spam-split9.log $LOGDIR/SPBASE/spam.log
ln -f gen-cache/spam-split1.log $LOGDIR/SPBASE/spam-test.log

echo "[Setting up for gen run]"
# Ok, setup for a run
ln -s $LOGDIR/SPBASE/spam.log .
ln -s $LOGDIR/NSBASE/ham.log .
ln -s $LOGDIR/SPBASE/spam-test.log .
ln -s $LOGDIR/NSBASE/ham-test.log .

# try to find number of processors
ostype=`uname`
if [ $ostype = "FreeBSD" ]; then
  numcpus=`/sbin/sysctl -n kern.smp.cpus`
elif [ $ostype = "SunOS" ]; then
  numcpus=`/usr/sbin/psrinfo | wc -l`
else
  numcpus=`cpucount 2>/dev/null || egrep -c '^processor\b' /proc/cpuinfo 2>/dev/null || echo 1`
fi
if [ ${numcpus:=0} -le 0 ]; then numcpus=1; fi

echo "[Generating GA]"
# Generate GA with full logs
make -j $numcpus SCORESET=$SCORESET garescorer > $LOGDIR/make.output 2>&1
cp freqs $LOGDIR/freqs

echo "[config]"
cat config
echo "[gen run start]"
pwd
date

# TODO: use -f $FPRATE instead of -b $HAM_PREFERENCE
time ./garescorer -b $HAM_PREFERENCE -e $EPOCHS -t $THRESHOLD || exit $?
date

# POST-GA COMMANDS:

mv garescorer.scores $LOGDIR/scores
echo "[gen run end]"

cp orig_scores.cf tmprules/50_scores.cf
perl ./rewrite-cf-with-new-scores --scoreset $SCORESET \
        --old-scores tmprules/50_scores.cf \
	--new-scores $LOGDIR/scores \
        --cffile tmprules \
        > tmprules/50_newscores.cf

mv tmprules/50_newscores.cf tmprules/50_scores.cf
cp tmprules/50_scores.cf $LOGDIR/50_scores.cf

perl ./fp-fn-statistics --ham ham-test.log --spam spam-test.log \
    --scoreset $SCORESET --cffile=tmprules \
    --fnlog $LOGDIR/false_negatives --fplog $LOGDIR/false_positives \
    > $LOGDIR/test

# END OF POST-GA COMMANDS

) | tee $LOGDIR/log

else

# -------------------------------------------------------------------------
# Statistics generation, once everyone likes the scores


  # use the logs we saved
  fulllogh=$LOGDIR/NSBASE/ham.log
  fulllogs=$LOGDIR/SPBASE/spam.log
  testlogh=$LOGDIR/NSBASE/ham-test.log
  testlogs=$LOGDIR/SPBASE/spam-test.log

  if [ ! -f "$testlogh" -o ! -f "$testlogs" ]; then
    echo "Couldn't find logs for $NAME: $testlogh $testlogs" >&2
    exit 1
  fi

  rm -f ham-test.log spam-test.log
  ln -s $testlogh ham-test.log
  ln -s $testlogs spam-test.log

  rm -f ham.log spam.log
  ln -s $fulllogh ham.log
  ln -s $fulllogs spam.log

  [ $KILL_SANDBOX_RULES = y ] && rm ../rules/70_sandbox.cf

  # This needs to have ../rules/50_scores.cf in place first ...
  echo "[gen test results for set $SCORESET]"
  perl ./fp-fn-statistics --ham $testlogh --spam $testlogs \
      --scoreset $SCORESET --cffile=../rules | tee $LOGDIR/test

  echo "[STATISTICS file generation for set $SCORESET]"
  bash ./mk-baseline-results $SCORESET | tee $LOGDIR/statistics

  cp $LOGDIR/statistics ../rules/STATISTICS-set${SCORESET}.txt
  ls -l ../rules/STATISTICS-set${SCORESET}.txt

fi

exit 0