Redo mapping entirely; now dumps constant-sized shards for reducing. Much more consis...

[remoteglot-book] / build-book.sh

diff --git a/build-book.sh b/build-book.sh
index 60d5710ed481eda49ee4e372bbf778778688e1dd..67181c68167bdf555bb4157d05fc5c2aa829dbc8 100755 (executable)
--- a/build-book.sh
+++ b/build-book.sh
@@ -3,29 +3,32 @@
 set -e
 
 export SHARDS=40
-export PARALLEL_LOADS=15  # Reduce if you have problems with OOM
 export PARALLEL_MERGES=40
 
-rm -f pgnnames.txt part-*.bin part-*.mtbl part-*.mtbl.part???? open.mtbl.new open.mtbl.part???? open.mtbl.part????.new 2>/dev/null
+rm -f pgnnames.txt tmp.mtbl* open.mtbl.new open.mtbl.part???? open.mtbl.part????.new 2>/dev/null
 
-PGNNUM=0
 for FILE in "$@"; do
-       date | tr -d "\n"
-       echo "  $FILE"
-       ./parallel-parse-pgn.sh "$FILE" "$PGNNUM"
        echo "$FILE" >> pgnnames.txt
-       PGNNUM=$(( PGNNUM + 1 ))
 done
-date 
 
-echo "Loading..."
-parallel -i -j ${PARALLEL_LOADS} bash -c 'FILE="{}"; set -x; ./binloader "$FILE" "${FILE/bin/mtbl}" $SHARDS' -- part-*.bin
-
-rm -f part-*.bin
+for X in $( seq 0 39 ); do
+       (
+               (
+                       PGNNUM=0
+                       for FILE in "$@"; do
+                               START=$( ./find-pgn-split-point.sh "$FILE" $X 40 )
+                               END=$( ./find-pgn-split-point.sh "$FILE" $(( X + 1 )) 40 )
+                               ~/nmu/pgn-extract/pgn-extract --startpos $START --endpos $END --startfilenum $PGNNUM -e -Wsessebin "$FILE" 2>/dev/null
+                               PGNNUM=$(( PGNNUM + 1 ))
+                       done
+               ) | ./binloader - tmp.mtbl.mapper$X $SHARDS 500000
+       ) &
+done
+wait
 
-parallel -i -j ${PARALLEL_MERGES} bash -c 'X={}; set -x; ./binmerger part-*.mtbl.part$( printf %04d $X ) open.mtbl.part$( printf %04d $X ).new' -- $( seq 0 $(( SHARDS - 1 )) )
+parallel -i -j ${PARALLEL_MERGES} bash -c 'X={}; set -x; ./binmerger tmp.mtbl.mapper*.part$( printf %04d $X ).subshard* open.mtbl.part$( printf %04d $X ).new' -- $( seq 0 $(( SHARDS - 1 )) )
 
 for X in $( seq 0 $(( SHARDS - 1 )) ); do
        mv open.mtbl.part$( printf %04d $X ).new open.mtbl.part$( printf %04d $X )
 done
-rm -f part-*.mtbl.part????
+rm -f tmp.mtbl*