fix handling input fileGrps where PAGE is mixed with derived images: (#…

…68)

* fix handling input fileGrps where PAGE is mixed with derived images:

- aggregate a list of PAGE pages
- aggregate a list of multi-image pages
- when iterating through files, check if the current page
  has a PAGE file:
  - if so, skip if not a PAGE file
  - otherwise, raise an error if page has multiple images

* require up-to-date pip and ocrd

Makefile

@@ -65,8 +65,8 @@ deps: #deps-ubuntu
 	test -x $(BINDIR)/scribo-cli && \
 	$(BINDIR)/scribo-cli sauvola --help >/dev/null 2>&1 || \
 		$(MAKE) build-olena
-	ocrd ocrd-tool --help >/dev/null 2>&1 || \
-		$(PIP) install ocrd # needed for ocrd CLI (and bashlib)
+	$(PIP) install -U pip
+	$(PIP) install "ocrd>=2.13" # needed for ocrd CLI (and bashlib)
 
 # Install
 install: deps

ocrd-olena-binarize

@@ -349,7 +349,7 @@ function main {
     # shellcheck source=../core/ocrd/bashlib/lib.bash
     source $(ocrd bashlib filename)
     ocrd__wrap "$SHAREDIR/ocrd-tool.json" "ocrd-olena-binarize" "$@"
-    ocrd__minversion 2.11.0
+    ocrd__minversion 2.13.0
 
     scribo_options=(--enable-negate-output)
     case ${params[impl]} in
@@ -393,6 +393,21 @@ function main {
                   -k mimetype        \
                   -k pageId          \
                   --download))
+    page_pages=($(ocrd workspace find \
+                       -G $in_file_grp \
+                       -m $MIMETYPE_PAGE \
+                       -k pageId))
+    multi_pages=($(ocrd workspace find \
+                        -G $in_file_grp \
+                        -m //image/.* \
+                        -k pageId | sort | uniq -d))
+    declare -A is_page is_multi
+    for page in "${page_pages[@]}"; do
+        is_page[$page]=1
+    done
+    for page in "${multi_pages[@]}"; do
+        is_multi[$page]=1
+    done
     local IFS=$' \t\n'
     local n=0 zeros=0000
     for csv in "${files[@]}"; do
@@ -410,8 +425,15 @@ function main {
         if ! test -f "${in_fpath#file://}"; then
            ocrd log error "input file ID=${in_id} (pageId=${in_pageId} MIME=${in_mimetype}) is not on disk"
            continue
-        fi 
-
+        fi
+        # fileGrps may contain PAGE and (derived) images
+        # so if this pageId has a PAGE file, ignore all others
+        # and otherwise if it has multiple images, raise an error
+        if ((${is_page[$in_pageId]:-0})); then
+            test x$in_mimetype != x$MIMETYPE_PAGE && continue
+        elif ((${is_multi[$in_pageId]:-0})); then
+            ocrd_raise "No PAGE-XML for page $in_pageId in fileGrp '$in_file_grp' but multiple images."
+        fi
         local out_id="${in_id//$in_file_grp/$out_file_grp}"
         if [ "x$out_id" = "x$in_id" ]; then
             out_id=${out_file_grp}_${zeros:0:$((4-${#n}))}$n

ocrd-tool.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.2.0",
+  "version": "1.2.1",
   "git_url": "https://github.com/OCR-D/ocrd_olena",
   "tools": {
     "ocrd-olena-binarize": {