Restrict underscore formats which are converted to float

Disallow numbers with trailing underscore from being converted to float.

The specific change here is to treat 0_e0 as a string.

Fixes issue #838

omegaconf/_utils.py

@@ -155,10 +155,10 @@ def construct_mapping(self, node: yaml.Node, deep: bool = False) -> Any:
         "tag:yaml.org,2002:float",
         re.compile(
             """^(?:
-         [-+]?(?:[0-9][0-9_]*)\\.[0-9_]*(?:[eE][-+]?[0-9]+)?
-        |[-+]?(?:[0-9][0-9_]*)(?:[eE][-+]?[0-9]+)
-        |\\.[0-9_]+(?:[eE][-+][0-9]+)?
-        |[-+]?[0-9][0-9_]*(?::[0-5]?[0-9])+\\.[0-9_]*
+         [-+]?[0-9]+(?:_[0-9]+)*\\.[0-9_]*(?:[eE][-+]?[0-9]+)?
+        |[-+]?[0-9]+(?:_[0-9]+)*(?:[eE][-+]?[0-9]+)
+        |\\.[0-9]+(?:_[0-9]+)*(?:[eE][-+][0-9]+)?
+        |[-+]?[0-9]+(?:_[0-9]+)*(?::[0-5]?[0-9])+\\.[0-9_]*
         |[-+]?\\.(?:inf|Inf|INF)
         |\\.(?:nan|NaN|NAN))$""",
             re.X,

tests/test_create.py

@@ -337,6 +337,42 @@ def test_create_unmodified_loader() -> None:
     assert yaml_cfg["gitrev"] == "100e100"
 
 
+def test_create_float_yaml() -> None:
+    # Note there are some discrepencies with the antrl parser.
+    # The following follow the yaml more closely,
+    # but arguably the antlr interpretation is better (which also
+    # more closely matches python. Specifically:
+    #   c_s not parsed as float. antlr does parse as float
+    #   e_s and f_s not parsed. antlr does parse as float
+    #   h_f and i_f parsed as float. antlr does not parse as float
+    cfg = OmegaConf.create(
+        dedent(
+            """\
+            a_s: 0_e0
+            b_i: 0_0
+            c_s: 1_0e1_0
+            d_f: .5
+            e_s: +.9
+            f_s: -.9
+            g_f: 1_1_2.1
+            h_f: 1__2.1
+            i_f: 1.2_
+            """
+        )
+    )
+    assert cfg == {
+        "a_s": "0_e0",
+        "b_i": 0,
+        "c_s": "1_0e1_0",
+        "d_f": 0.5,
+        "e_s": "+.9",
+        "f_s": "-.9",
+        "g_f": 112.1,
+        "h_f": 12.1,
+        "i_f": 1.2,
+    }
+
+
 def test_create_untyped_list() -> None:
     from omegaconf._utils import get_type_hint
 

tests/test_grammar.py

@@ -89,17 +89,21 @@
     ("float_no_int", ".1", 0.1),
     ("float_no_decimal", "1.", 1.0),
     ("float_minus", "-.2", -0.2),
-    ("float_underscore", "1.1_1", 1.11),
+    ("float_plus", "+.2", 0.2),
+    ("float_underscore_1", "1.1_1", 1.11),
+    ("float_underscore_2", "1_1_2.1", 112.1),
     ("float_bad_1", "1.+2", "1.+2"),
     ("float_bad_2", r"1\.2", r"1\.2"),
     ("float_bad_3", "1.2_", "1.2_"),
+    ("float_bad_4", "1__1.2", "1__1.2"),
     ("float_exp_1", "-1e2", -100.0),
     ("float_exp_2", "+1E-2", 0.01),
     ("float_exp_3", "1_0e1_0", 10e10),
     ("float_exp_4", "1.07e+2", 107.0),
     ("float_exp_5", "1e+03", 1000.0),
     ("float_exp_bad_1", "e-2", "e-2"),
     ("float_exp_bad_2", "01e2", "01e2"),
+    ("float_exp_bad_3", "0_e0", "0_e0"),
     ("float_inf", "inf", math.inf),
     ("float_plus_inf", "+inf", math.inf),
     ("float_minus_inf", "-inf", -math.inf),