check if IDs (node.ord) form a valid sequence in *.conll and *.conllup

martinpopel · martinpopel · commit 38f8c0b9f1a3 · 2025-03-27T15:43:50.000+01:00
There are e.g. files with gaps (1,2,3,5). Failing to detect this
results in wrong parsing and even `fix_cycles=1` may result in infinite loop.

When a node is created using `root.create_child()`,
it already has `node.ord` set up correctly as the last node in the tree.

For *.conllu we could do the same, but we don't do that
because Read.Conllu is speed critical
and invalid files can be detected using `validate.py`.
diff --git a/udapi/block/read/conll.py b/udapi/block/read/conll.py
@@ -91,7 +91,8 @@ def parse_node_line(self, line, root, nodes, parents, mwts):
                     else:
                         raise exception
             elif attribute_name == 'ord':
-                setattr(node, 'ord', int(value))
+                if int(value) != node._ord:
+                    raise ValueError(f"Node {node} ord mismatch: {value}, but expecting {node._ord} at:\n{line}")
             elif attribute_name == 'deps':
                 setattr(node, 'raw_deps', value)
             elif attribute_name != '_' and value != '_':
diff --git a/udapi/block/read/conllup.py b/udapi/block/read/conllup.py
@@ -79,7 +79,8 @@ def parse_node_line(self, line, root, nodes, parents, mwts):
                     else:
                         raise exception
             elif attribute_name == 'ord':
-                setattr(node, 'ord', int(value))
+                if int(value) != node._ord:
+                    raise ValueError(f"Node {node} ord mismatch: {value}, but expecting {node._ord} at:\n{line}")
             elif attribute_name == 'deps':
                 setattr(node, 'raw_deps', value)
             elif value == '_' and attribute_name != 'form':
