Fix #143 Disallow quotes in roles and symbols

Quotes are now only allowed to delimit strings or as escaped
characters within strings.

This commit also changes the PEG grammar to disallow line-breaking
spaces within strings so it more accurately represents the line-based
parsing behavior.

docs/notation.rst

@@ -90,8 +90,10 @@ grammar to allow for surface alignments.
    Symbol    <- NameChar+
    Role      <- ':' NameChar*
    Alignment <- '~' ([a-zA-Z] '.'?)? Digit+ (',' Digit+)*
-   String    <- '"' (!'"' ('\\' . / .))* '"'
-   NameChar  <- ![ \n\t\r\f\v()/:~] .
+   String    <- '"' (!'"' (StrEscape / StrChar))* '"'
+   StrEscape <- '\\' StrChar
+   StrChar   <- ![\n\r\f\v] .
+   NameChar  <- ![ \n\t\r\f\v"()/:~] .
    Digit     <- [0-9]
 
 This grammar has some seemingly unnecessary ambiguity in that both the

penman/_lexer.py

@@ -22,8 +22,8 @@
     'ALIGNMENT': r'~(?:[a-z]\.?)?[0-9]+(?:,[0-9]+)*',
     # ROLE cannot be made up of COLON + SYMBOL because it then becomes
     # difficult to detect anonymous roles: (a : b) vs (a :b c)
-    'ROLE': r':[^ \t\r\n\v\f()\/:~]*',
-    'SYMBOL': r'[^ \t\r\n\v\f()\/:~]+',
+    'ROLE': r':[^ \t\r\n\v\f"()\/:~]*',
+    'SYMBOL': r'[^ \t\r\n\v\f"()\/:~]+',
     'LPAREN': r'\(',
     'RPAREN': r'\)',
     'SLASH': r'\/',  # concept (node label) role

tests/test_codec.py

@@ -229,6 +229,16 @@ def test_decode_recursion_limit(self):
         assert len(g.triples) == (n         # n :instance triples
                                   + n - 1)  # n - 1 :ARG0 triples
 
+    def test_decode_issue_143(self):
+        # https://github.com/goodmami/penman/issues/143
+        with pytest.raises(penman.DecodeError):
+            decode('(a :op ")')
+        with pytest.raises(penman.DecodeError):
+            decode('(a :op1 " :op2 "foo")')
+        with pytest.raises(penman.DecodeError):
+            decode('(a :" foo)')
+
+
     def test_encode(self, x1):
         # empty graph
         g = penman.Graph([])

tests/test_lexer.py

@@ -74,3 +74,18 @@ def test_nonbreaking_space_issue_99():
     assert [tok.type for tok in lexer.lex('1\r2')] == ['SYMBOL', 'SYMBOL']
     assert [tok.type for tok in lexer.lex('1\u00a02')] == ['SYMBOL']
     assert [tok.type for tok in lexer.lex('あ　い')] == ['SYMBOL']
+
+
+def test_unterminated_string_issue_143():
+    # https://github.com/goodmami/penman/issues/143
+    # unmatched quotes result in unexpected tokens
+    assert [tok.type for tok in lexer.lex('(a :op ")')] == [
+        'LPAREN', 'SYMBOL', 'ROLE', 'UNEXPECTED', 'RPAREN'
+    ]
+    assert [tok.type for tok in lexer.lex('(a :op1 " :op2 "foo")')] == [
+        'LPAREN', 'SYMBOL', 'ROLE', 'STRING', 'SYMBOL', 'UNEXPECTED', 'RPAREN'
+    ]
+    # also disallow quotes in role names
+    assert [tok.type for tok in lexer.lex('(a :" b)')] == [
+        'LPAREN', 'SYMBOL', 'ROLE', 'UNEXPECTED', 'SYMBOL', 'RPAREN'
+    ]