Allow custom_formats for known format

e.g. to focus on dates in a particular range.

Author: Zac-HD

CHANGELOG.md

@@ -1,5 +1,8 @@
 # Changelog
 
+#### 0.20.0 - 2021-06-02
+- Allow `custom_formats` for known string formats (#83)
+
 #### 0.19.2 - 2021-05-17
 - Fix support `date` and `time` formats (#79)
 

src/hypothesis_jsonschema/__init__.py

@@ -3,7 +3,7 @@
 The only public API is `from_schema`; check the docstring for details.
 """
 
-__version__ = "0.19.2"
+__version__ = "0.20.0"
 __all__ = ["from_schema"]
 
 from ._from_schema import from_schema

src/hypothesis_jsonschema/_from_schema.py

@@ -11,7 +11,7 @@
 
 import jsonschema
 from hypothesis import assume, provisional as prov, strategies as st
-from hypothesis.errors import InvalidArgument
+from hypothesis.errors import HypothesisWarning, InvalidArgument
 from hypothesis.internal.conjecture import utils as cu
 
 from ._canonicalise import (
@@ -76,8 +76,11 @@ def from_schema(
 ) -> st.SearchStrategy[JSONType]:
     """Take a JSON schema and return a strategy for allowed JSON objects.
 
-    Schema reuse with "definitions" and "$ref" is not yet supported, but
-    everything else in drafts 04, 05, and 07 is fully tested and working.
+    To generate specific string formats, pass a ``custom_formats`` dict
+    mapping the format name to a strategy for allowed strings.
+
+    Supports JSONSchema drafts 04, 06, and 07, with the exception of
+    recursive references.
     """
     try:
         return __from_schema(schema, custom_formats=custom_formats)
@@ -127,13 +130,18 @@ def __from_schema(
         for name, strat in custom_formats.items():
             if not isinstance(name, str):
                 raise InvalidArgument(f"format name {name!r} must be a string")
-            if name in STRING_FORMATS:
-                raise InvalidArgument(f"Cannot redefine standard format {name!r}")
             if not isinstance(strat, st.SearchStrategy):
                 raise InvalidArgument(
                     f"custom_formats[{name!r}]={strat!r} must be a Hypothesis "
                     "strategy which generates strings matching this format."
                 )
+            if name in STRING_FORMATS:
+                warnings.warn(
+                    f"Overriding standard format {name!r} - was "
+                    f"{STRING_FORMATS[name]!r}, now {strat!r}",
+                    HypothesisWarning,
+                    stacklevel=2,
+                )
         format_checker = jsonschema.FormatChecker()
         custom_formats = {
             name: _get_format_filter(name, format_checker, strategy)
@@ -406,7 +414,7 @@ def string_schema(
     min_size = schema.get("minLength", 0)
     max_size = schema.get("maxLength")
     strategy = st.text(min_size=min_size, max_size=max_size)
-    known_formats = {**(custom_formats or {}), **STRING_FORMATS}
+    known_formats = {**STRING_FORMATS, **(custom_formats or {})}
     if schema.get("format") in known_formats:
         # Unknown "format" specifiers should be ignored for validation.
         # See https://json-schema.org/latest/json-schema-validation.html#format

tests/test_from_schema.py

@@ -18,7 +18,7 @@
     settings,
     strategies as st,
 )
-from hypothesis.errors import FailedHealthCheck, InvalidArgument
+from hypothesis.errors import FailedHealthCheck, HypothesisWarning, InvalidArgument
 from hypothesis.internal.reflection import proxies
 
 from hypothesis_jsonschema._canonicalise import (
@@ -430,7 +430,6 @@ def validate_card_format(string):
     [
         {"foo": "not a strategy"},
         {5: st.just("name is not a string")},
-        {"full-date": st.just("2000-01-01")},  # can't override a standard format
         {"card-test": st.just("not a valid card")},
         {"card-test": st.none()},  # Not a string
     ],
@@ -463,6 +462,18 @@ def test_allowed_unknown_custom_format(string):
     assert "not registered" not in jsonschema.FormatChecker().checkers
 
 
+@given(data=st.data())
+def test_overriding_standard_format(data):
+    expected = "2000-01-01"
+    schema = {"type": "string", "format": "full-date"}
+    custom_formats = {"full-date": st.just(expected)}
+    with pytest.warns(
+        HypothesisWarning, match="Overriding standard format 'full-date'"
+    ):
+        value = data.draw(from_schema(schema, custom_formats=custom_formats))
+    assert value == expected
+
+
 with warnings.catch_warnings():
     warnings.simplefilter("ignore", UserWarning)