Merge pull request #328 from kba/loctype

set LOCTYPE and OTHERLOCTYPE as necessary, fix #310

ocrd_models/ocrd_models/ocrd_file.py

@@ -17,7 +17,7 @@ class OcrdFile():
     #  def create(mimetype, ID, url, local_filename):
     #      el_fileGrp.SubElement('file')
 
-    def __init__(self, el, mimetype=None, instance=None, local_filename=None, mets=None, url=None):
+    def __init__(self, el, mimetype=None, loctype='OTHER', instance=None, local_filename=None, mets=None, url=None):
         """
         Args:
             el (LxmlElement):
@@ -33,6 +33,7 @@ def __init__(self, el, mimetype=None, instance=None, local_filename=None, mets=N
         self.local_filename = local_filename
         self._instance = instance
         self.mets = mets
+        self.loctype = loctype
 
         if url:
             self.url = url
@@ -117,6 +118,48 @@ def pageId(self, pageId):
         self.mets.set_physical_page_for_file(pageId, self)
 
 
+    @property
+    def loctype(self):
+        """
+        Get the ``LOCTYPE``.
+        """
+        el_FLocat = self._el.find('mets:FLocat', NS)
+        return '' if el_FLocat is None else el_FLocat.get('LOCTYPE')
+
+    @loctype.setter
+    def loctype(self, loctype):
+        """
+        Set the ``LOCTYPE``.
+        """
+        if loctype is None:
+            return
+        loctype = loctype.upper()
+        el_FLocat = self._el.find('mets:FLocat', NS)
+        if el_FLocat is None:
+            el_FLocat = ET.SubElement(self._el, TAG_METS_FLOCAT)
+        el_FLocat.set('LOCTYPE', loctype)
+        if loctype == 'OTHER':
+            self.otherloctype = 'FILE'
+        else:
+            self.otherloctype = None
+
+    @property
+    def otherloctype(self):
+        el_FLocat = self._el.find('mets:FLocat', NS)
+        return '' if el_FLocat is None else el_FLocat.get('OTHERLOCTYPE')
+
+    @otherloctype.setter
+    def otherloctype(self, otherloctype):
+        el_FLocat = self._el.find('mets:FLocat', NS)
+        if el_FLocat is None:
+            el_FLocat = ET.SubElement(self._el, TAG_METS_FLOCAT)
+        if not otherloctype:
+            if 'OTHERLOCTYPE' in el_FLocat.attrib:
+                del el_FLocat.attrib['OTHERLOCTYPE']
+        else:
+            el_FLocat.set('LOCTYPE', 'OTHER')
+            el_FLocat.set('OTHERLOCTYPE', otherloctype)
+
     @property
     def mimetype(self):
         """

ocrd_validators/ocrd_validators/workspace_validator.py

@@ -192,7 +192,7 @@ def _validate_mets_files(self):
                 self.report.add_notice("File '%s' has GROUPID attribute - document might need an update" % f.ID)
             if not f.pageId:
                 self.report.add_error("File '%s' does not manifest any physical page." % f.ID)
-            if 'url' not in self.skip and ':/' in f.url:
+            if 'url' not in self.skip and f.url and ':/' in f.url:
                 if re.match(r'^file:/[^/]', f.url):
                     self.report.add_warning("File '%s' has an invalid (Java-specific) file URL '%s'" % (f.ID, f.url))
                 scheme = f.url[0:f.url.index(':')]

tests/cli/test_workspace.py

@@ -185,13 +185,17 @@ def test_remove_file_group(self):
     def test_copy_vs_clone(self):
         src_dir = assets.path_to('kant_aufklaerung_1784/data')
         with TemporaryDirectory() as tempdir:
+            # cloned without download
             shallowcloneddir = join(tempdir, 'cloned-shallow')
+            # cloned with download
             fullcloneddir = join(tempdir, 'cloned-all')
+            # copied
             copieddir = join(tempdir, 'copied')
 
             Path(fullcloneddir).mkdir()
             Path(shallowcloneddir).mkdir()
 
+
             result = self.runner.invoke(workspace_cli, ['clone', join(src_dir, 'mets.xml'), shallowcloneddir])
             self.assertEqual(result.exit_code, 0)
 
@@ -200,9 +204,20 @@ def test_copy_vs_clone(self):
 
             with copy_of_directory(src_dir, copieddir):
                 shallow_vs_copied = dircmp(shallowcloneddir, copieddir)
-                full_vs_copied = dircmp(fullcloneddir, copieddir)
                 self.assertEqual(set(shallow_vs_copied.right_only), set(['OCR-D-GT-ALTO', 'OCR-D-GT-PAGE', 'OCR-D-IMG']))
-                self.assertEqual(full_vs_copied.diff_files, [])
+
+                full_vs_copied = dircmp(fullcloneddir, copieddir)
+                #  print(full_vs_copied)
+                #  from ocrd_utils import pushd_popd
+                #  with pushd_popd(tempdir):
+                    #  import os
+                    #  os.system("diff %s/mets.xml %s/mets.xml" % (fullcloneddir, copieddir))
+                # XXX mets.xml will not have the exact same content because
+                # URLs that are actually files will be marked up as such with
+                # @LOCTYPE/@OTHERLOCTYPE
+                #  self.assertEqual(full_vs_copied.diff_files, [])
+                self.assertEqual(full_vs_copied.left_only, [])
+                self.assertEqual(full_vs_copied.right_only, [])
 
 if __name__ == '__main__':
     main()

tests/model/test_ocrd_file.py

@@ -1,3 +1,4 @@
+from lxml import etree as ET
 from tests.base import TestCase, main
 from ocrd_models import OcrdFile
 
@@ -10,6 +11,18 @@ def test_no_pageid_without_mets(self):
         with self.assertRaisesRegex(Exception, ".*has no member 'mets' pointing.*"):
             f.pageId = 'foo'
 
+    def test_loctype(self):
+        f = OcrdFile(None)
+        self.assertEqual(f.loctype, 'OTHER')
+        self.assertEqual(f.otherloctype, 'FILE')
+        f.otherloctype = 'foo'
+        self.assertEqual(f.otherloctype, 'foo')
+        f.loctype = 'URN'
+        self.assertEqual(f.loctype, 'URN')
+        self.assertEqual(f.otherloctype, None)
+        f.otherloctype = 'foo'
+        self.assertEqual(f.loctype, 'OTHER')
+
     def test_set_url(self):
         f = OcrdFile(None)
         f.url = None

tests/test_workspace.py

@@ -47,7 +47,7 @@ def test_workspace_add_file_basename_no_content(self):
             ws1 = self.resolver.workspace_from_nothing(directory=tempdir)
             ws1.add_file('GRP', ID='ID1', mimetype='image/tiff')
             f = ws1.mets.find_files()[0]
-            self.assertEqual(f.url, '')
+            self.assertEqual(f.url, None)
 
     def test_workspace_add_file_binary_content(self):
         with TemporaryDirectory() as tempdir: