From 08ed3669145580ccd5e0d0ee64005d4600e083a5 Mon Sep 17 00:00:00 2001
From: Joerg Tiedemann <jorg.tiedemann@helsinki.fi>
Date: Thu, 4 Nov 2021 09:57:48 +0200
Subject: [PATCH] allas storage commands

---
 Makefile             | 19 +++++++++++++++++++
 lib/allas.mk         | 41 +++++++++++++++++++++++++++++++++++++++++
 scripts/fix_vocab.py |  2 +-
 3 files changed, 61 insertions(+), 1 deletion(-)

diff --git a/Makefile b/Makefile
index 5c4214f9..ebdb50ba 100644
--- a/Makefile
+++ b/Makefile
@@ -378,3 +378,22 @@ train-and-start-bt-jobs: ${WORKDIR}/${MODEL}.${MODELTYPE}.model${NR}.done
 	${MAKE} local-dist
 	${MAKE} -C backtranslate MODELHOME=${MODELDIR} translate-all-wikis-jobs
 
+
+ALL_RELEASED_MODELS = ${wildcard models-tatoeba/*/*.zip}
+ALL_VOCABS_FIXED = ${patsubst %.zip,%.fixed-vocab,${ALL_RELEASED_MODELS}}
+
+fix-released-vocabs: ${ALL_VOCABS_FIXED}
+
+%.fixed-vocab: %.zip
+	@( v=`unzip -l $<  | grep 'vocab.yml$$' | sed 's/^.* //'`; \
+	  if [ "$$v" != "" ]; then \
+	    unzip $< $$v; \
+	    python3 scripts/fix_vocab.py $$v; \
+	    if [ -e $$v.bak ]; then \
+	      echo "update $$v in $<"; \
+	      zip $< $$v $$v.bak; \
+	    else \
+	      echo "vocab $$v is fine in $<"; \
+	    fi; \
+	    rm -f $$v $$v.bak; \
+	  fi )
diff --git a/lib/allas.mk b/lib/allas.mk
index 8b912b21..b28a377a 100644
--- a/lib/allas.mk
+++ b/lib/allas.mk
@@ -20,6 +20,9 @@ WORK_DESTDIR      ?= ${WORKHOME}
 WORK_CONTAINER    ?= OPUS-MT-train_${notdir ${WORKHOME}}-${WHOAMI}
 WORK_CONTAINER_JT ?= OPUS-MT-train_${notdir ${WORKHOME}}-tiedeman
 
+ALLAS_STORAGE_URL = https://object.pouta.csc.fi/
+
+
 ## store workdir on allas
 store:
 	cd ${WORK_SRCDIR} && a-put -b ${WORK_CONTAINER} --nc --follow-links --override ${LANGPAIRSTR}
@@ -43,3 +46,41 @@ fetch-data:
 	mkdir -p ${WORK_DESTDIR}
 	cd ${WORK_DESTDIR} && a-get ${WORK_CONTAINER}/data.tar
 
+
+
+## generic recipe for storing work data and removing it from the file system
+## DANGEROUS --- this really deletes the data!
+## NOTE: makes container also world-readable (see swift post command)
+##       --> this makes it easier to fetch things without login credentials
+##       --> should not store sensitive data here!
+%.stored: %
+	if [ "$(firstword $(subst -, ,$(subst /, ,$@)))" == "work" ]; then \
+	  b=OPUS-MT-train_$(subst /,-,$(dir $@))${WHOAMI}; \
+	  cd $(dir $@); \
+	  a-put -b $$b --nc --follow-links --override $(notdir $<); \
+	  rm -fr $(notdir $<); \
+	  touch $(notdir $@); \
+	  rm -f $(notdir $(@:stored=.fetched)); \
+	  swift post $$b --read-acl ".r:*"
+	fi
+
+
+## TODO: fetch with wget instead of using a-commands
+## fetch work data from allas
+%.fetched:
+	if [ "$(firstword $(subst -, ,$(subst /, ,$@)))" == "work" ]; then \
+	  cd $(dir $@); \
+	  a-get OPUS-MT-train_$(subst /,-,$(dir $@))${WHOAMI}/$(notdir $(@:.fetched=.tar)); \
+	  touch $(notdir $@); \
+	  rm -f $(notdir $(@:fetched=.stored)); \
+	fi
+
+## another way of fetching work data
+## requires settings SRCLANGS and TRGLANGS (or LANGPAIRSTR directly)
+work-%/${LANGPAIRSTR}:
+	mkdir -p $(dir $@)
+	cd $(dir $@) && a-get OPUS-MT-train_$(subst /,-,$(dir $@))${WHOAMI}/${LANGPAIRSTR}.tar
+
+
+UPLOAD_MODELS=$(patsubst %,%.stored,${wildcard work-tatoeba/[dg-rt-z]*})
+upload-workfiles: ${UPLOAD_MODELS}
diff --git a/scripts/fix_vocab.py b/scripts/fix_vocab.py
index c6965880..5e34d0fb 100755
--- a/scripts/fix_vocab.py
+++ b/scripts/fix_vocab.py
@@ -10,7 +10,7 @@ filename = sys.argv[1]
 
 try:
     input = open(filename, 'r')
-    yaml.load(input)
+    yaml.safe_load(input)
 except:
     print('YAML file is broken - try to fix it!')
     print(f'copy {filename} to {filename}.bak')