bpo-16806: Fix lineno and col_offset for multi-line string tokens (GH-10021)

Author: asottile

Lib/test/test_ast.py

@@ -683,6 +683,25 @@ def test_get_docstring_none(self):
         node = ast.parse('async def foo():\n  x = "not docstring"')
         self.assertIsNone(ast.get_docstring(node.body[0]))
 
+    def test_multi_line_docstring_col_offset_and_lineno_issue16806(self):
+        node = ast.parse(
+            '"""line one\nline two"""\n\n'
+            'def foo():\n  """line one\n  line two"""\n\n'
+            '  def bar():\n    """line one\n    line two"""\n'
+            '  """line one\n  line two"""\n'
+            '"""line one\nline two"""\n\n'
+        )
+        self.assertEqual(node.body[0].col_offset, 0)
+        self.assertEqual(node.body[0].lineno, 1)
+        self.assertEqual(node.body[1].body[0].col_offset, 2)
+        self.assertEqual(node.body[1].body[0].lineno, 5)
+        self.assertEqual(node.body[1].body[1].body[0].col_offset, 4)
+        self.assertEqual(node.body[1].body[1].body[0].lineno, 9)
+        self.assertEqual(node.body[1].body[2].col_offset, 2)
+        self.assertEqual(node.body[1].body[2].lineno, 11)
+        self.assertEqual(node.body[2].col_offset, 0)
+        self.assertEqual(node.body[2].lineno, 13)
+
     def test_literal_eval(self):
         self.assertEqual(ast.literal_eval('[1, 2, 3]'), [1, 2, 3])
         self.assertEqual(ast.literal_eval('{"foo": 42}'), {"foo": 42})

Lib/test/test_fstring.py

@@ -270,10 +270,7 @@ def test_ast_line_numbers_duplicate_expression(self):
         self.assertEqual(binop.right.col_offset, 7)  # FIXME: this is wrong
 
     def test_ast_line_numbers_multiline_fstring(self):
-        # FIXME: This test demonstrates invalid behavior due to JoinedStr's
-        # immediate child nodes containing the wrong lineno.  The enclosed
-        # expressions have valid line information and column offsets.
-        # See bpo-16806 and bpo-30465 for details.
+        # See bpo-30465 for details.
         expr = """
 a = 10
 f'''
@@ -298,19 +295,16 @@ def test_ast_line_numbers_multiline_fstring(self):
         self.assertEqual(type(t.body[1].value.values[1]), ast.FormattedValue)
         self.assertEqual(type(t.body[1].value.values[2]), ast.Constant)
         self.assertEqual(type(t.body[1].value.values[2].value), str)
-        # NOTE: the following invalid behavior is described in bpo-16806.
-        # - line number should be the *first* line (3), not the *last* (8)
-        # - column offset should not be -1
-        self.assertEqual(t.body[1].lineno, 8)
-        self.assertEqual(t.body[1].value.lineno, 8)
-        self.assertEqual(t.body[1].value.values[0].lineno, 8)
-        self.assertEqual(t.body[1].value.values[1].lineno, 8)
-        self.assertEqual(t.body[1].value.values[2].lineno, 8)
-        self.assertEqual(t.body[1].col_offset, -1)
-        self.assertEqual(t.body[1].value.col_offset, -1)
-        self.assertEqual(t.body[1].value.values[0].col_offset, -1)
-        self.assertEqual(t.body[1].value.values[1].col_offset, -1)
-        self.assertEqual(t.body[1].value.values[2].col_offset, -1)
+        self.assertEqual(t.body[1].lineno, 3)
+        self.assertEqual(t.body[1].value.lineno, 3)
+        self.assertEqual(t.body[1].value.values[0].lineno, 3)
+        self.assertEqual(t.body[1].value.values[1].lineno, 3)
+        self.assertEqual(t.body[1].value.values[2].lineno, 3)
+        self.assertEqual(t.body[1].col_offset, 0)
+        self.assertEqual(t.body[1].value.col_offset, 0)
+        self.assertEqual(t.body[1].value.values[0].col_offset, 0)
+        self.assertEqual(t.body[1].value.values[1].col_offset, 0)
+        self.assertEqual(t.body[1].value.values[2].col_offset, 0)
         # NOTE: the following lineno information and col_offset is correct for
         # expressions within FormattedValues.
         binop = t.body[1].value.values[1].value
@@ -321,8 +315,8 @@ def test_ast_line_numbers_multiline_fstring(self):
         self.assertEqual(binop.lineno, 4)
         self.assertEqual(binop.left.lineno, 4)
         self.assertEqual(binop.right.lineno, 6)
-        self.assertEqual(binop.col_offset, 3)
-        self.assertEqual(binop.left.col_offset, 3)
+        self.assertEqual(binop.col_offset, 4)
+        self.assertEqual(binop.left.col_offset, 4)
         self.assertEqual(binop.right.col_offset, 7)
 
     def test_docstring(self):

Lib/test/test_opcodes.py

@@ -27,7 +27,7 @@ def test_setup_annotations_line(self):
             with open(ann_module.__file__) as f:
                 txt = f.read()
             co = compile(txt, ann_module.__file__, 'exec')
-            self.assertEqual(co.co_firstlineno, 6)
+            self.assertEqual(co.co_firstlineno, 3)
         except OSError:
             pass
 

Lib/test/test_string_literals.py

@@ -117,7 +117,7 @@ def test_eval_str_invalid_escape(self):
             eval("'''\n\\z'''")
         self.assertEqual(len(w), 1)
         self.assertEqual(w[0].filename, '<string>')
-        self.assertEqual(w[0].lineno, 2)
+        self.assertEqual(w[0].lineno, 1)
 
         with warnings.catch_warnings(record=True) as w:
             warnings.simplefilter('error', category=SyntaxWarning)
@@ -126,7 +126,7 @@ def test_eval_str_invalid_escape(self):
             exc = cm.exception
         self.assertEqual(w, [])
         self.assertEqual(exc.filename, '<string>')
-        self.assertEqual(exc.lineno, 2)
+        self.assertEqual(exc.lineno, 1)
 
     def test_eval_str_raw(self):
         self.assertEqual(eval(""" r'x' """), 'x')
@@ -166,7 +166,7 @@ def test_eval_bytes_invalid_escape(self):
             eval("b'''\n\\z'''")
         self.assertEqual(len(w), 1)
         self.assertEqual(w[0].filename, '<string>')
-        self.assertEqual(w[0].lineno, 2)
+        self.assertEqual(w[0].lineno, 1)
 
         with warnings.catch_warnings(record=True) as w:
             warnings.simplefilter('error', category=SyntaxWarning)
@@ -175,7 +175,7 @@ def test_eval_bytes_invalid_escape(self):
             exc = cm.exception
         self.assertEqual(w, [])
         self.assertEqual(exc.filename, '<string>')
-        self.assertEqual(exc.lineno, 2)
+        self.assertEqual(exc.lineno, 1)
 
     def test_eval_bytes_raw(self):
         self.assertEqual(eval(""" br'x' """), b'x')

Misc/ACKS

@@ -1845,3 +1845,4 @@ Gennadiy Zlobin
 Doug Zongker
 Peter Åstrand
 Zheao Li
+Carsten Klein

Misc/NEWS.d/next/Core and Builtins/2018-10-20-18-05-58.bpo-16806.zr3A9N.rst

@@ -0,0 +1 @@
+Fix ``lineno`` and ``col_offset`` for multi-line string tokens.

Parser/parsetok.c

@@ -205,6 +205,8 @@ parsetok(struct tok_state *tok, grammar *g, int start, perrdetail *err_ret,
         size_t len;
         char *str;
         col_offset = -1;
+        int lineno;
+        const char *line_start;
 
         type = PyTokenizer_Get(tok, &a, &b);
         if (type == ERRORTOKEN) {
@@ -253,8 +255,15 @@ parsetok(struct tok_state *tok, grammar *g, int start, perrdetail *err_ret,
             }
         }
 #endif
-        if (a != NULL && a >= tok->line_start) {
-            col_offset = Py_SAFE_DOWNCAST(a - tok->line_start,
+
+        /* Nodes of type STRING, especially multi line strings
+           must be handled differently in order to get both
+           the starting line number and the column offset right.
+           (cf. issue 16806) */
+        lineno = type == STRING ? tok->first_lineno : tok->lineno;
+        line_start = type == STRING ? tok->multi_line_start : tok->line_start;
+        if (a != NULL && a >= line_start) {
+            col_offset = Py_SAFE_DOWNCAST(a - line_start,
                                           intptr_t, int);
         }
         else {
@@ -263,7 +272,7 @@ parsetok(struct tok_state *tok, grammar *g, int start, perrdetail *err_ret,
 
         if ((err_ret->error =
              PyParser_AddToken(ps, (int)type, str,
-                               tok->lineno, col_offset,
+                               lineno, col_offset,
                                &(err_ret->expected))) != E_OK) {
             if (err_ret->error != E_DONE) {
                 PyObject_FREE(str);

Parser/tokenizer.c

@@ -1519,6 +1519,13 @@ tok_get(struct tok_state *tok, char **p_start, char **p_end)
         int quote_size = 1;             /* 1 or 3 */
         int end_quote_size = 0;
 
+        /* Nodes of type STRING, especially multi line strings
+           must be handled differently in order to get both
+           the starting line number and the column offset right.
+           (cf. issue 16806) */
+        tok->first_lineno = tok->lineno;
+        tok->multi_line_start = tok->line_start;
+
         /* Find the quote size and start of string */
         c = tok_nextc(tok);
         if (c == quote) {

Parser/tokenizer.h

@@ -38,6 +38,8 @@ struct tok_state {
     int pendin;         /* Pending indents (if > 0) or dedents (if < 0) */
     const char *prompt, *nextprompt;          /* For interactive prompting */
     int lineno;         /* Current line number */
+    int first_lineno;   /* First line of a single line or multi line string
+                           expression (cf. issue 16806) */
     int level;          /* () [] {} Parentheses nesting level */
             /* Used to allow free continuations inside them */
 #ifndef PGEN
@@ -58,6 +60,9 @@ struct tok_state {
     char *encoding;         /* Source encoding. */
     int cont_line;          /* whether we are in a continuation line. */
     const char* line_start;     /* pointer to start of current line */
+    const char* multi_line_start; /* pointer to start of first line of
+                                     a single line or multi line string
+                                     expression (cf. issue 16806) */
 #ifndef PGEN
     PyObject *decoding_readline; /* open(...).readline */
     PyObject *decoding_buffer;

Python/ast.c

@@ -4284,9 +4284,13 @@ fstring_fix_node_location(const node *parent, node *n, char *expr_str)
                 start--;
             }
             cols += (int)(substr - start);
-            /* Fix lineno in mulitline strings. */
-            while ((substr = strchr(substr + 1, '\n')))
-                lines--;
+            /* adjust the start based on the number of newlines encountered
+               before the f-string expression */
+            for (char* p = parent->n_str; p < substr; p++) {
+                if (*p == '\n') {
+                    lines++;
+                }
+            }
         }
     }
     fstring_shift_node_locations(n, lines, cols);

Python/importlib.h

@@ -484,7 +484,7 @@ const unsigned char _Py_M__zipimport[] = {
     64,0,0,0,114,65,0,0,0,114,78,0,0,0,114,82,
     0,0,0,114,83,0,0,0,114,9,0,0,0,114,9,0,
     0,0,114,9,0,0,0,114,10,0,0,0,114,4,0,0,
-    0,45,0,0,0,115,24,0,0,0,8,13,4,5,8,46,
+    0,45,0,0,0,115,24,0,0,0,8,1,4,17,8,46,
     10,32,10,12,8,10,8,21,8,11,8,26,8,13,8,38,
     8,18,122,12,95,95,105,110,105,116,95,95,46,112,121,99,
     84,114,60,0,0,0,70,41,3,122,4,46,112,121,99,84,
@@ -1044,7 +1044,7 @@ const unsigned char _Py_M__zipimport[] = {
     34,0,0,0,114,182,0,0,0,114,183,0,0,0,114,184,
     0,0,0,114,189,0,0,0,114,9,0,0,0,114,9,0,
     0,0,114,9,0,0,0,114,10,0,0,0,114,80,0,0,
-    0,212,2,0,0,115,14,0,0,0,8,5,4,1,4,2,
+    0,212,2,0,0,115,14,0,0,0,8,1,4,5,4,2,
     8,4,8,9,8,6,8,11,114,80,0,0,0,41,45,114,
     84,0,0,0,90,26,95,102,114,111,122,101,110,95,105,109,
     112,111,114,116,108,105,98,95,101,120,116,101,114,110,97,108,
@@ -1065,8 +1065,8 @@ const unsigned char _Py_M__zipimport[] = {
     0,0,114,170,0,0,0,114,152,0,0,0,114,150,0,0,
     0,114,44,0,0,0,114,80,0,0,0,114,9,0,0,0,
     114,9,0,0,0,114,9,0,0,0,114,10,0,0,0,218,
-    8,60,109,111,100,117,108,101,62,13,0,0,0,115,88,0,
-    0,0,4,4,8,1,16,1,8,1,8,1,8,1,8,1,
+    8,60,109,111,100,117,108,101,62,1,0,0,0,115,88,0,
+    0,0,4,16,8,1,16,1,8,1,8,1,8,1,8,1,
     8,1,8,2,8,3,6,1,14,3,16,4,4,2,8,2,
     4,1,4,1,4,2,14,127,0,127,0,1,12,1,12,1,
     2,1,2,252,4,9,8,4,8,9,8,31,8,126,2,254,