| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 1 | SRC_DIR ?= I5 |
| 2 | |
| 3 | # Discover all *.i5.xml files in SRC_DIR |
| 4 | I5_FILES := $(wildcard $(SRC_DIR)/*.i5.xml) |
| 5 | BASENAMES := $(patsubst %.i5.xml,%,$(notdir $(I5_FILES))) |
| 6 | |
| 7 | BUILD_DIR = build |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 8 | TARGET_DIR ?= ./target |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 9 | MAX_THREADS ?= 8 # $(shell nproc) |
| 10 | MAKE ?= make -j $(shell nproc) |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 11 | # KORAPXMLTOOL_HEAP ?= $(shell echo "$$(($(MAX_THREADS) * 2500))") |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 12 | KORAPXMLTOOL ?= ./bin/korapxmltool |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 13 | KORAPXMLTOOL_MODELS_PATH ?= models |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 14 | |
| 15 | .DELETE_ON_ERROR: |
| 16 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 17 | .PHONY: all clean test index korap check-src |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 18 | |
| Marc Kupietz | 5b3ecc0 | 2025-12-15 14:43:09 +0100 | [diff] [blame] | 19 | .PRECIOUS: $(BUILD_DIR)/%.zip $(BUILD_DIR)/%.tree_tagger.zip $(BUILD_DIR)/%.marmot-malt.zip $(BUILD_DIR)/%.spacy.zip $(BUILD_DIR)/%.corenlp.zip $(BUILD_DIR)/%.cmc.zip $(BUILD_DIR)/%.opennlp.zip $(BUILD_DIR)/%.krill.tar %.i5.xml |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 20 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 21 | all: check-src korap |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 22 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 23 | index: check-src $(TARGET_DIR)/index |
| 24 | |
| 25 | check-src: |
| 26 | @if [ ! -d "$(SRC_DIR)" ]; then \ |
| 27 | echo "Error: SRC_DIR '$(SRC_DIR)' does not exist."; \ |
| 28 | echo "Please create it and place your .i5.xml files there,"; \ |
| 29 | echo "or specify a different directory using SRC_DIR variable."; \ |
| 30 | echo "Example: make SRC_DIR=/path/to/files"; \ |
| 31 | exit 1; \ |
| 32 | fi |
| 33 | @if [ -z "$$(find "$(SRC_DIR)" -maxdepth 1 -name '*.i5.xml' -print -quit)" ]; then \ |
| 34 | echo "Error: No .i5.xml files found in '$(SRC_DIR)'."; \ |
| 35 | echo "Please populate it or set SRC_DIR to a different location."; \ |
| 36 | exit 1; \ |
| 37 | fi |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 38 | |
| 39 | $(BUILD_DIR)/%.zip: $(SRC_DIR)/%.i5.xml |
| 40 | mkdir -p $(BUILD_DIR) |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 41 | tei2korapxml --progress -l warn -s -tk $< > $@ |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 42 | printf "%s\t%s\n" "$(grep -c '<idsText ' $<)" "$(unzip -l $@ | grep data.xml | wc -l)" |
| 43 | |
| 44 | |
| 45 | $(BUILD_DIR)/%.tree_tagger.zip: $(BUILD_DIR)/%.zip bin/korapxmltool |
| Marc Kupietz | 5b3ecc0 | 2025-12-15 14:43:09 +0100 | [diff] [blame] | 46 | $(KORAPXMLTOOL) -j 1 -T treetagger -t zip --force -D $(BUILD_DIR) $< |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 47 | # $(KORAPXMLTOOL) $< | pv | docker run --rm -i korap/conllu2treetagger -l german | conllu2korapxml > $@ |
| 48 | |
| 49 | $(BUILD_DIR)/%.spacy.zip: $(BUILD_DIR)/%.zip bin/korapxmltool |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 50 | $(KORAPXMLTOOL) -P spacy -t zip --force -D $(BUILD_DIR) $< |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 51 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 52 | lib/Krill-Indexer.jar: |
| 53 | mkdir -p lib |
| 54 | curl -sL -o $@ https://github.com/korap/Krill/releases/latest/download/Krill-Indexer.jar |
| Marc Kupietz | 5b3ecc0 | 2025-12-15 14:43:09 +0100 | [diff] [blame] | 55 | |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 56 | bin/korapxmltool: |
| 57 | mkdir -p bin |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 58 | curl -sL -o $@ https://github.com/korap/korapxmltool/releases/latest/download/korapxmltool |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 59 | chmod +x $@ |
| 60 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 61 | $(KORAPXMLTOOL_MODELS_PATH)/de.marmot: |
| 62 | mkdir -p $(KORAPXMLTOOL_MODELS_PATH) |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 63 | curl -sL -o $@ https://cistern.cis.lmu.de/marmot/models/CURRENT/spmrl/de.marmot |
| 64 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 65 | $(KORAPXMLTOOL_MODELS_PATH)/german.mco: |
| 66 | mkdir -p $(KORAPXMLTOOL_MODELS_PATH) |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 67 | curl -sL -o $@ https://corpora.ids-mannheim.de/tools/$@ |
| 68 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 69 | $(KORAPXMLTOOL_MODELS_PATH)/dereko_domains_s.classifier: |
| 70 | mkdir -p $(KORAPXMLTOOL_MODELS_PATH) |
| 71 | curl -sL -o $@ https://corpora.ids-mannheim.de/tools/models/$@ |
| 72 | |
| 73 | $(KORAPXMLTOOL_MODELS_PATH)/german-fast.tagger: |
| 74 | mkdir -p $(KORAPXMLTOOL_MODELS_PATH) |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 75 | curl -sL -o $@ https://corpora.ids-mannheim.de/tools/$@ |
| 76 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 77 | $(KORAPXMLTOOL_MODELS_PATH)/germanSR.ser.gz: |
| 78 | mkdir -p $(KORAPXMLTOOL_MODELS_PATH) |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 79 | curl -sL -o $@ https://corpora.ids-mannheim.de/tools/$@ |
| 80 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 81 | $(KORAPXMLTOOL_MODELS_PATH)/de-pos-maxent.bin: |
| 82 | mkdir -p $(KORAPXMLTOOL_MODELS_PATH) |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 83 | curl -sL -o $@ https://corpora.ids-mannheim.de/tools/$@ |
| 84 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 85 | $(BUILD_DIR)/%.marmot-malt.zip: $(BUILD_DIR)/%.zip $(KORAPXMLTOOL_MODELS_PATH)/de.marmot $(KORAPXMLTOOL_MODELS_PATH)/german.mco bin/korapxmltool |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 86 | $(KORAPXMLTOOL) -T marmot:models/de.marmot -P malt:models/german.mco -t zip --force -D $(BUILD_DIR) $< |
| Marc Kupietz | 5b3ecc0 | 2025-12-15 14:43:09 +0100 | [diff] [blame] | 87 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 88 | $(BUILD_DIR)/%.corenlp.zip: $(BUILD_DIR)/%.zip $(KORAPXMLTOOL_MODELS_PATH)/german-fast.tagger $(KORAPXMLTOOL_MODELS_PATH)/germanSR.ser.gz bin/korapxmltool |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 89 | $(KORAPXMLTOOL) -T corenlp -P corenlp -t zip --force -D $(BUILD_DIR) $< |
| 90 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 91 | $(BUILD_DIR)/%.opennlp.zip: $(BUILD_DIR)/%.zip $(KORAPXMLTOOL_MODELS_PATH)/de-pos-maxent.bin bin/korapxmltool |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 92 | $(KORAPXMLTOOL) -T opennlp -t zip --force -D $(BUILD_DIR) $< |
| 93 | |
| Marc Kupietz | 800f9bb | 2025-12-11 17:30:24 +0100 | [diff] [blame] | 94 | $(BUILD_DIR)/%.cmc.zip: $(BUILD_DIR)/%.zip bin/korapxmltool |
| Marc Kupietz | 5b3ecc0 | 2025-12-15 14:43:09 +0100 | [diff] [blame] | 95 | $(KORAPXMLTOOL) -j 1 -A "docker run --rm -i korap/conllu-cmc -s" -l error -F cmc -t zip --force -D $(BUILD_DIR) $< |
| Marc Kupietz | 800f9bb | 2025-12-11 17:30:24 +0100 | [diff] [blame] | 96 | |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 97 | # udpipe target removed as requested |
| 98 | # %.ud.zip: %.zip |
| 99 | # $(KORAPXMLTOOL) $< | pv | ./scripts/udpipe2 | conllu2korapxml > $@ |
| 100 | |
| Marc Kupietz | 800f9bb | 2025-12-11 17:30:24 +0100 | [diff] [blame] | 101 | $(BUILD_DIR)/%.krill.tar: $(BUILD_DIR)/%.zip $(BUILD_DIR)/%.marmot-malt.zip $(BUILD_DIR)/%.tree_tagger.zip $(BUILD_DIR)/%.spacy.zip $(BUILD_DIR)/%.corenlp.zip $(BUILD_DIR)/%.opennlp.zip $(BUILD_DIR)/%.cmc.zip |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 102 | K2K_PUBLISHER_STRING=1 K2K_TRANSLATOR_TEXT=1 $(KORAPXMLTOOL) --non-word-tokens -f -t krill -D $(BUILD_DIR) $(basename $<)*.zip |
| 103 | |
| Marc Kupietz | 15b313d | 2025-12-11 12:20:56 +0100 | [diff] [blame] | 104 | $(TARGET_DIR)/index: $(foreach base,$(BASENAMES),$(BUILD_DIR)/$(base).krill.tar) |
| 105 | make lib/Krill-Indexer.jar |
| 106 | touch lib/krill.cfg |
| 107 | mkdir -p $(TARGET_DIR) |
| 108 | java -jar lib/Krill-Indexer.jar -c lib/krill.cfg --progress -i $(subst " ",;,$^) -o $@ |
| 109 | |
| 110 | korap: check-src $(TARGET_DIR)/index |
| 111 | curl https://raw.githubusercontent.com/KorAP/KorAP-Docker/master/compose.yaml | INDEX='$(TARGET_DIR)/index' docker compose -p korap -f - --profile=lite --profile=example up |
| Marc Kupietz | 89a4407 | 2025-12-11 07:14:00 +0100 | [diff] [blame] | 112 | |
| 113 | $(TARGET_DIR)/index.tar.xz: $(TARGET_DIR)/index |
| 114 | tar -I 'xz -T0' -C $(dir $<) -cf $@ $(notdir $<) |
| 115 | |
| 116 | clean: |
| 117 | rm -rf $(BUILD_DIR) $(TARGET_DIR) |