gcf-python 2.5.2__tar.gz → 2.6.0__tar.gz

{gcf_python-2.5.2 → gcf_python-2.6.0}/CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## v2.6.0 (2026-08-14)
+
+- **Numeric domain (spec v3.5.3, SPEC 2.3.2).** Specifies the canonical numeric domain as signed `int64` for integers and IEEE-754 double for non-integers. Earlier versions left integers beyond the double-exact range (2^53) to the host numeric type; this version parses integer literals to an exact `int64` on decode and on the JSON-to-value bridge, returns an out-of-range error for a value outside `int64` on both decode and encode, and models larger values (unsigned-64 identifiers, exact decimals) as strings. Canonical number formatting aligns to the domain: a double at or above 2^53 renders in exponent notation. Verified against new `numbers/017-024` and `errors-v2/041-042` conformance fixtures and the cross-SDK differential fuzz. An out-of-range value raises `ValueError`.
+
+## v2.5.3 (2026-08-10)
+
+- **Losslessness fix (spec v3.5.2, SPEC 2.3/2.4).** The number grammar and numeric-like classification now pin digits to ASCII `[0-9]`. Python's `re` compiles `\d` in Unicode mode (matching `\p{Nd}`), so a value like `1.٥` (ASCII `1`, `.`, U+0665) was treated as number-shaped: quoted on encode, and a bare `1.٥` decoded through `float()` to the number `1.5`. That diverged from the ASCII SDKs and silently turned a string into a number when decoding wire produced by another SDK. `\d` is replaced with `[0-9]`; such tokens stay strings and encode bare. Verified against new `scalar/029-031` and `decode/007` conformance fixtures and the cross-SDK differential fuzz.
+
 ## v2.5.2 (2026-08-09)
 
 - **Spec v3.5.1 conformance (SPEC 5, score-rounding errata).** SPEC 5 now pins the graph `score` two-decimal wire form to round-half-to-even on the exact IEEE-754 double, resolving a midpoint divergence in the JavaScript and Kotlin SDKs. This SDK's `f"{score:.2f}"` formatter already rounds half-to-even, so there is no behavior change; re-verified against the new `graph-encode/004_score_midpoint_rounding` conformance fixture.

{gcf_python-2.5.2 → gcf_python-2.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.4
+Metadata-Version: 2.5
 Name: gcf-python
-Version: 2.5.2
+Version: 2.6.0
 Summary: The AI-native wire format for structured data. 50-92% fewer tokens than JSON, with multi-turn delta encoding for agent loops. 100% comprehension on every frontier model. Zero dependencies.
 Project-URL: Homepage, https://github.com/blackwell-systems/gcf-python
 Project-URL: Documentation, https://gcformat.com/
@@ -304,7 +304,7 @@ GCF wins 15/16 datasets on the expanded [token efficiency benchmark](https://git
 
 **Zero runtime dependencies. Permanently.** All six implementations depend only on their language's standard library. No transitive dependencies. No supply chain risk. This is a permanent commitment: GCF will never take on external runtime dependencies. MIT licensed. All implementations support both generic profile (`encodeGeneric`) and graph profile (`encode`). CLI included in all 6 languages.
 
-**Specification:** [SPEC v3.5.0 Stable](https://github.com/blackwell-systems/gcf/blob/main/SPEC.md) with 264 conformance fixtures, 43,000,000,000+ lossless round-trips verified across 5 formats and 6 languages. All implementations at v2.5.1+ (Go v1.6.1, Swift v2.6.0). Cross-language 6x6 matrix verified.
+**Specification:** [SPEC v3.5.2 Stable](https://github.com/blackwell-systems/gcf/blob/main/SPEC.md) with 269 conformance fixtures, 43,000,000,000+ lossless round-trips verified across 5 formats and 6 languages. Current versions: Go v1.6.2, TypeScript v2.5.2, Python v2.5.3, Rust v2.5.3, Swift v2.6.2, Kotlin v2.5.2, .NET v0.1.2. Cross-language conformance verified across all seven SDKs.
 
 ## Adopted by
 

{gcf_python-2.5.2 → gcf_python-2.6.0}/README.md

@@ -279,7 +279,7 @@ GCF wins 15/16 datasets on the expanded [token efficiency benchmark](https://git
 
 **Zero runtime dependencies. Permanently.** All six implementations depend only on their language's standard library. No transitive dependencies. No supply chain risk. This is a permanent commitment: GCF will never take on external runtime dependencies. MIT licensed. All implementations support both generic profile (`encodeGeneric`) and graph profile (`encode`). CLI included in all 6 languages.
 
-**Specification:** [SPEC v3.5.0 Stable](https://github.com/blackwell-systems/gcf/blob/main/SPEC.md) with 264 conformance fixtures, 43,000,000,000+ lossless round-trips verified across 5 formats and 6 languages. All implementations at v2.5.1+ (Go v1.6.1, Swift v2.6.0). Cross-language 6x6 matrix verified.
+**Specification:** [SPEC v3.5.2 Stable](https://github.com/blackwell-systems/gcf/blob/main/SPEC.md) with 269 conformance fixtures, 43,000,000,000+ lossless round-trips verified across 5 formats and 6 languages. Current versions: Go v1.6.2, TypeScript v2.5.2, Python v2.5.3, Rust v2.5.3, Swift v2.6.2, Kotlin v2.5.2, .NET v0.1.2. Cross-language conformance verified across all seven SDKs.
 
 ## Adopted by
 

{gcf_python-2.5.2 → gcf_python-2.6.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "gcf-python"
-version = "2.5.2"
+version = "2.6.0"
 description = "The AI-native wire format for structured data. 50-92% fewer tokens than JSON, with multi-turn delta encoding for agent loops. 100% comprehension on every frontier model. Zero dependencies."
 readme = "README.md"
 license = {text = "MIT"}

{gcf_python-2.5.2 → gcf_python-2.6.0}/src/gcf/__init__.py

@@ -102,4 +102,4 @@ __all__ = [
     "DEFAULT_REANCHOR_N",
 ]
 
-__version__ = "2.3.0"
+__version__ = "2.6.0"

{gcf_python-2.5.2 → gcf_python-2.6.0}/src/gcf/scalar.py

@@ -9,8 +9,10 @@ from typing import Any
 # \Z (end of string), not $, so a trailing newline does not count as the end:
 # in Python $ also matches just before a final \n, which would misclassify a
 # string like "5\n" as a number or "W\n" as a bare key and break round-trip.
-_JSON_NUMBER_RE = re.compile(r"^-?(?:0|[1-9]\d*)(?:\.\d+)?(?:[eE][+-]?\d+)?\Z")
-_NUMERIC_LIKE_RE = re.compile(r"^[+-]\.?\d|^\.\d|^0\d")
+# Digits are ASCII 0-9 (SPEC 2.3): \d is avoided because in Python's re it also
+# matches Unicode decimal digits (\p{Nd}), which would accept e.g. "1.<U+0665>".
+_JSON_NUMBER_RE = re.compile(r"^-?(?:0|[1-9][0-9]*)(?:\.[0-9]+)?(?:[eE][+-]?[0-9]+)?\Z")
+_NUMERIC_LIKE_RE = re.compile(r"^[+-]\.?[0-9]|^\.[0-9]|^0[0-9]")
 _INLINE_ARRAY_RE = re.compile(r"\[[^\]]*\]\s*:")
 _BARE_KEY_RE = re.compile(r"^[a-zA-Z_][a-zA-Z0-9_]*\Z")
 
@@ -91,6 +93,15 @@ def format_scalar(v: Any, delimiter: str = "") -> str:
     if isinstance(v, bool):
         return "true" if v else "false"
     if isinstance(v, int) and not isinstance(v, bool):
+        # The encoder enforces the int64 domain (SPEC 2.3.2): a Python int is
+        # arbitrary-precision, so a host integer outside int64 is rejected here
+        # rather than emitted as a bare token the decoder would reject.
+        if v < -(2**63) or v > 2**63 - 1:
+            raise ValueError(
+                f"out_of_range: integer {v} is outside the canonical int64 "
+                "domain [-9223372036854775808, 9223372036854775807]; "
+                "model larger values as strings (SPEC 2.3.2)"
+            )
         return str(v)
     if isinstance(v, float):
         return format_number(v)
@@ -109,7 +120,12 @@ def format_number(f: float) -> str:
         # Negative zero canonicalizes to 0 (SPEC 2.3.1): -0.0 equals 0.0 by value.
         return "0"
     a = abs(f)
-    if 1e-6 <= a < 1e21:
+    # Plain decimal only below 2^53. Every double at or above 2^53 is integer-valued,
+    # so a plain rendering would emit a bare-integer token: indistinguishable from an
+    # int64 on the wire and beyond the binary64 safe-integer range (2^53-1), so a
+    # JavaScript decoder rejects it under its default policy. Exponent shape keeps bare
+    # tokens int64 and decimal/exponent tokens doubles (SPEC 2.3.1). Ints format above.
+    if 1e-6 <= a < 2**53:
         # Use repr for shortest round-trippable form.
         s = repr(f)
         # If repr chose scientific notation, convert to plain decimal.
@@ -163,12 +179,20 @@ def parse_scalar(s: str, tabular_context: bool = False) -> Any:
     if s == "false":
         return False
     if _JSON_NUMBER_RE.match(s):
+        # Token shape follows domain (SPEC 2.3.2): a bare-integer literal (no
+        # fraction, no exponent) is an int64-domain integer parsed exactly, not
+        # routed through float(); a decimal or exponent literal is a double.
+        if "." not in s and "e" not in s and "E" not in s:
+            n = int(s)
+            if n < -(2**63) or n > 2**63 - 1:
+                raise ValueError(
+                    f"out_of_range: integer {s} is outside the canonical int64 "
+                    "domain [-9223372036854775808, 9223372036854775807]; "
+                    "model larger values as strings (SPEC 2.3.2)"
+                )
+            return n
         try:
-            f = float(s)
-            if "." not in s and "e" not in s and "E" not in s:
-                if abs(f) <= 2**53:
-                    return int(f)
-            return f
+            return float(s)
         except ValueError:
             pass
     return s

{gcf_python-2.5.2 → gcf_python-2.6.0}/tests/test_conformance_v2.py

@@ -183,6 +183,23 @@ def test_conformance(rel_path, data):
         with pytest.raises((ValueError, Exception)):
             decode_generic(data["input"])
 
+    elif op == "roundtrip-wire":
+        # Input and expected are wire strings: decode then re-encode and require the
+        # result to equal the input wire. The value never becomes a host number, so a
+        # value that a JSON parser would float (an integer beyond 2^53) is pinned here.
+        decoded = decode_generic(data["input"])
+        reencoded = encode_generic(decoded)
+        assert reencoded == data["expected"], (
+            f"wire idempotence mismatch:\n  got: {reencoded!r}\n  exp: {data['expected']!r}"
+        )
+
+    elif op == "encode-error":
+        # Input is a JSON value (encode-side) out of the numeric domain; encoding it
+        # must raise. In Python the JSON parser preserves big integers exactly, so the
+        # encoder is the domain-enforcement site.
+        with pytest.raises((ValueError, Exception)):
+            encode_generic(data["input"])
+
     elif op == "generic-pack-root":
         from gcf.generic_delta import GenericSet, generic_pack_root