diff options
-rw-r--r-- | .gitmodules | 3 | ||||
-rw-r--r-- | Makefile | 70 | ||||
m--------- | wikiextractor | 0 |
3 files changed, 51 insertions, 22 deletions
diff --git a/.gitmodules b/.gitmodules index eb4e7ca..8f00840 100644 --- a/.gitmodules +++ b/.gitmodules @@ -1,6 +1,3 @@ [submodule "nutrimatic"] path = nutrimatic url = https://github.com/egnor/nutrimatic -[submodule "wikiextractor"] - path = wikiextractor - url = https://github.com/attardi/wikiextractor @@ -5,6 +5,8 @@ all: enwiktionary-latest.version all: enwiktionary-latest.index all: enwiki-latest.version all: enwiki-latest.index +all: phraser-words500K-latest.index +all: phraser-phrases5M-latest.index .PHONY: newest newest: @@ -20,8 +22,13 @@ sync: newest .PHONY: trim trim: - ls enwiki-*-snap.index | sort | head -n -2 | xargs -rt rm - ls enwiktionary-*-snap.index | sort | head -n -2 | xargs -rt rm + ls enwiki-*-snap.index | sort | head -n -2 | xargs -t rm + ls enwiktionary-*-snap.index | sort | head -n -2 | xargs -t rm + ls enwiki-*-pages-articles.xml.bz2 | sort | head -n -2 | xargs -t rm + ls enwiktionary-*-pages-articles.xml.bz2 | sort | head -n -2 | xargs -t rm + ls phraser-words500K-*-snap.index | sort | head -n -2 | xargs -t rm + ls phraser-phrases5M-*-snap.index | sort | head -n -2 | xargs -t rm + ls phraser-*.txt | sort | head -n -2 | xargs -t rm %/.git: git submodule update --init $* @@ -29,30 +36,55 @@ trim: nutrimatic/bin: nutrimatic/.git cd nutrimatic && ./build.py -wikiextractor/WikiExtractor.py: wikiextractor/.git - -%-latest.version: - wget -O - https://dumps.wikimedia.org/$*/latest/$*-latest-pages-articles.xml.bz2-rss.xml |\ - xpath '//rss/channel/link/text()' |\ +en%-latest.version: + wget -O - https://dumps.wikimedia.org/en$*/latest/en$*-latest-pages-articles.xml.bz2-rss.xml |\ + xpath -e '//rss/channel/link/text()' |\ sed 's/^.*\///' > $@ -enwiki-%-snap.merge: nutrimatic/bin wikiextractor/WikiExtractor.py - # ~7h30m - wget -O - https://dumps.wikimedia.org/enwiki/$*/enwiki-$*-pages-articles.xml.bz2 |\ - bzip2 -d |\ - python wikiextractor/WikiExtractor.py -q -o- - |\ - ./merge.sh $@ +enwiki-%-pages-articles.xml.bz2: + wget -O $@ https://dumps.wikimedia.org/enwiki/$*/enwiki-$*-pages-articles.xml.bz2 -enwiktionary-%-snap.merge: nutrimatic/bin wikiextractor/WikiExtractor.py +enwiki-%-snap.merge: nutrimatic/bin enwiki-%-pages-articles.xml.bz2 # ~1h - wget -O - https://dumps.wikimedia.org/enwiktionary/$*/enwiktionary-$*-pages-articles.xml.bz2 |\ - bzip2 -d |\ - python wikiextractor/WikiExtractor.py -q -o- - |\ + wikiextractor -q -o- enwiki-$*-pages-articles.xml.bz2 |\ + ./merge.sh $@ + +enwiktionary-%-pages-articles.xml.bz2: + wget -O $@ https://dumps.wikimedia.org/enwiktionary/$*/enwiktionary-$*-pages-articles.xml.bz2 + +enwiktionary-%-snap.merge: nutrimatic/bin enwiktionary-%-pages-articles.xml.bz2 + wikiextractor -q -o- enwiktionary-$*-pages-articles.xml.bz2 |\ ./merge.sh $@ -%-snap.index: %-snap.merge nutrimatic/bin - nutrimatic/bin/merge-indexes 5 $*-snap.merge $@ +en%-snap.index: en%-snap.merge nutrimatic/bin + nutrimatic/bin/merge-indexes 5 en$*-snap.merge $@ %-latest.index: %-latest.version make $*-$(shell cat $*-latest.version)-snap.index ln -sf $*-$(shell cat $*-latest.version)-snap.index $@ + +phraser-words500K-latest.version: + date -d \ + "$$(curl -I https://lahosken.san-francisco.ca.us/frivolity/prog/phraser/words_500K.txt |\ + awk -F': ' '/^last-modified: / { print $$2 }')" \ + +"%Y%m%d%H%M%S" > $@ + +phraser-phrases5M-latest.version: + date -d \ + "$$(curl -I https://lahosken.san-francisco.ca.us/frivolity/prog/phraser/phrases_5M.txt |\ + awk -F': ' '/^last-modified: / { print $$2 }')" \ + +"%Y%m%d%H%M%S" > $@ + +phraser-words500K-%.txt: + wget -O $@ https://lahosken.san-francisco.ca.us/frivolity/prog/phraser/words_500K.txt + +phraser-phrases5M-%.txt: + wget -O $@ https://lahosken.san-francisco.ca.us/frivolity/prog/phraser/phrases_5M.txt + +phraser-words500K-%-snap.index: phraser-words500K-%.txt nutrimatic/bin + # ~ 3m + awk -F'\t' '{ for (i=0; i < $$1; i++) print $$2 }' phraser-words500K-$*.txt | ./merge.sh $@ + +phraser-phrases5M-%-snap.index: phraser-phrases5M-%.txt nutrimatic/bin + # ~ 10m + awk -F'\t' '{ for (i=0; i < $$1; i++) print $$2 }' phraser-phrases5M-$*.txt | ./merge.sh $@ diff --git a/wikiextractor b/wikiextractor deleted file mode 160000 -Subproject 3162bb6c3c9ebd2d15be507aa11d6fa818a454a |