util.MarkDiff align=both

Author: martinpopel

udapi/block/util/markdiff.py

@@ -9,7 +9,8 @@ class MarkDiff(Block):
     """Mark differences between parallel trees."""
 
     def __init__(self, gold_zone, attributes='form,lemma,upos,xpos,deprel,feats,misc',
-                 mark=1, mark_attr="Mark", add=False, print_stats=0, ignore_parent=False, **kwargs):
+                 mark=1, mark_attr='Mark', add=False, print_stats=0, ignore_parent=False,
+                 align=False, align_attr='Align', **kwargs):
         """Create the Mark block object.
         Params:
         gold_zone: Which of the zones should be treated as gold?
@@ -18,10 +19,19 @@ def __init__(self, gold_zone, attributes='form,lemma,upos,xpos,deprel,feats,misc
             The tree topology, i.e. node parent is always considered.
         mark: What value should be used in `node.misc['Mark']` of the differing nodes?
         mark_attr: use this MISC attribute name instead of "Mark".
+            Use mark_attr=0 to prevent marking diffs in MISC.
         add: If False, node.misc attributes Mark, ToDo and Bug will be deleted before running this block,
             so that the marked_only option (e.g. via `udapy -TM`) prints only nodes marked by this block.
         print_stats: How many lines of statistics should be printed? -1 means all.
         ignore_parent: ignore differences in dependency parents
+        align: store word alignment, possible values are False (no alignment stored, the default)
+            "from-pred", i.e. pred_node.misc["Align"] = aligned_gold_node.ord,
+            "from-gold", i.e. gold_node.misc["Align"] = aligned_pred_node.ord and
+            "both", i.e. both from-pred and from-gold.
+            If only forms should be considered for inducing the word alignment,
+            you should use "util.MarkDiff attributes='form' ignore_parent=1 align=1".
+            Only one-to-one alignment is supported.
+        align_attr: use this MISC attribute name instead of "Align".
         """
         super().__init__(**kwargs)
         self.gold_zone = gold_zone
@@ -31,7 +41,11 @@ def __init__(self, gold_zone, attributes='form,lemma,upos,xpos,deprel,feats,misc
         self.add = add
         self.print_stats = print_stats
         self.ignore_parent = ignore_parent
+        self.align = align
+        self.align_attr = align_attr
         self.stats = collections.Counter()
+        if not mark_attr and not align and not print_stats:
+            raise ValueError('mark_attr=0 does not make sense without align or print_stats')
 
     def process_tree(self, tree):
         gold_tree = tree.bundle.get_tree(self.gold_zone)
@@ -45,7 +59,7 @@ def process_tree(self, tree):
 
         pred_nodes, gold_nodes = tree.descendants, gold_tree.descendants
         # Make sure both pred and gold trees are marked, even if one has just deleted nodes.
-        if len(pred_nodes) != len(gold_nodes):
+        if len(pred_nodes) != len(gold_nodes) and self.mark_attr:
             tree.add_comment(f'{self.mark_attr} = {self.mark}')
             gold_tree.add_comment(f'{self.mark_attr} = {self.mark}')
         pred_tokens = ['_'.join(n.get_attrs(self.attrs)) for n in pred_nodes]
@@ -59,18 +73,24 @@ def process_tree(self, tree):
             if edit in {'equal', 'replace'}:
                 for i in range(pred_lo, pred_hi):
                     alignment[i] = i - pred_lo + gold_lo
+                    if self.align in ("both", "from-pred"):
+                        pred_nodes[i].misc[self.align_attr] = i - pred_lo + gold_lo + 1
+                    if self.align in ("both", "from-gold"):
+                        gold_nodes[i - pred_lo + gold_lo].misc[self.align_attr] = i + 1
 
         for diff in diffs:
             edit, pred_lo, pred_hi, gold_lo, gold_hi = diff
             if edit == 'equal':
                 for p_node, g_node in zip(pred_nodes[pred_lo:pred_hi], gold_nodes[gold_lo:gold_hi]):
                     if not self.ignore_parent and alignment.get(p_node.parent.ord - 1) != g_node.parent.ord - 1:
-                        p_node.misc[self.mark_attr] = self.mark
-                        g_node.misc[self.mark_attr] = self.mark
                         self.stats['ONLY-PARENT-CHANGED'] += 1
+                        if self.mark_attr:
+                            p_node.misc[self.mark_attr] = self.mark
+                            g_node.misc[self.mark_attr] = self.mark
             else:
-                for node in pred_nodes[pred_lo:pred_hi] + gold_nodes[gold_lo:gold_hi]:
-                    node.misc[self.mark_attr] = self.mark
+                if self.mark_attr:
+                    for node in pred_nodes[pred_lo:pred_hi] + gold_nodes[gold_lo:gold_hi]:
+                        node.misc[self.mark_attr] = self.mark
                 if self.print_stats:
                     if edit == 'replace':
                         # first n nodes are treated as aligned, the rest is treated as ADDED/DELETED