refs #205 Add regression test for skipws/noskipws issue

tests/functional/regressions/test_issue205_skipws_propagation.py

@@ -0,0 +1,41 @@
+from __future__ import unicode_literals
+
+import pytest
+from textx import metamodel_from_str, TextXSyntaxError
+
+
+def test_issue205_skipws_propagation():
+    """
+    Test a problem found in Arpeggio not maintaining `skipws/noskipws` applied
+    to ordered choice root rule.
+    Reported at SO: https://stackoverflow.com/questions/57944531/how-do-you-correctly-mix-textx-skipws-non-skipws  # noqa
+    """
+
+    mm = metamodel_from_str(r'''
+            Sentence[skipws]: words*=Word;
+            Word[noskipws]: ID '.' | ID | '.';
+        ''')
+
+    # If ```noskipws``` from the ```Word``` rule is obeyed then this won't
+    # parse as there is no rule to consume spaces in between the words. In the
+    # unfixed Arpeggio version ```skipws``` from ```Sentence``` would nullify
+    # ```noskipws``` from Word so the parse will not fail but it would not be
+    # what should be expected.
+    with pytest.raises(TextXSyntaxError):
+        mm.model_from_str('''foo bar .''')
+
+    # We should consume spaces explicitly. Here, we show that a rule with the
+    # ```noskipws``` rule modifier is able to consume spaces explicitly. Match
+    # suppression (-) is employed to exclude these spaces from the target value
+    # (this means the ```Word``` does not include the spaces in its value).
+    # Note: ```skipws``` of Sentence does not consume the spaces between the
+    # words, because the rule modifiers
+    # (http://textx.github.io/textX/stable/grammar/#rule-modifiers) apply
+    # immediately. Thus, we need to consume the extra spaces before a word.
+    mm = metamodel_from_str(r'''
+            Sentence[skipws]: words*=Word;
+            Word[noskipws]: /\s*/- (ID '.' | ID | '.');
+        ''')
+
+    m = mm.model_from_str('''foo bar .''')
+    assert len(m.words) == 3