[mypyc] Raise ValueError if int too big for native int type (#20385)

Previously we used OverflowError, which seems inconsistent with
CPython, since a conversion/range check is arguably not an arithmetic
operation, and OverflowError is a subclass of ArithmeticError.

This is an extract for the docs for OverflowError: "Raised when the
result of an arithmetic operation is too large to be represented."

Author: JukkaL

mypyc/lib-rt/int_ops.c

@@ -400,7 +400,7 @@ int64_t CPyLong_AsInt64_(PyObject *o) {
         if (PyErr_Occurred()) {
             return CPY_LL_INT_ERROR;
         } else if (overflow) {
-            PyErr_SetString(PyExc_OverflowError, "int too large to convert to i64");
+            PyErr_SetString(PyExc_ValueError, "int too large to convert to i64");
             return CPY_LL_INT_ERROR;
         }
     }
@@ -453,7 +453,7 @@ int32_t CPyLong_AsInt32_(PyObject *o) {
         if (PyErr_Occurred()) {
             return CPY_LL_INT_ERROR;
         } else if (overflow) {
-            PyErr_SetString(PyExc_OverflowError, "int too large to convert to i32");
+            PyErr_SetString(PyExc_ValueError, "int too large to convert to i32");
             return CPY_LL_INT_ERROR;
         }
     }
@@ -495,7 +495,7 @@ int32_t CPyInt32_Remainder(int32_t x, int32_t y) {
 }
 
 void CPyInt32_Overflow() {
-    PyErr_SetString(PyExc_OverflowError, "int too large to convert to i32");
+    PyErr_SetString(PyExc_ValueError, "int too large to convert to i32");
 }
 
 // i16 unboxing slow path
@@ -510,7 +510,7 @@ int16_t CPyLong_AsInt16_(PyObject *o) {
         if (PyErr_Occurred()) {
             return CPY_LL_INT_ERROR;
         } else if (overflow) {
-            PyErr_SetString(PyExc_OverflowError, "int too large to convert to i16");
+            PyErr_SetString(PyExc_ValueError, "int too large to convert to i16");
             return CPY_LL_INT_ERROR;
         }
     }
@@ -552,7 +552,7 @@ int16_t CPyInt16_Remainder(int16_t x, int16_t y) {
 }
 
 void CPyInt16_Overflow() {
-    PyErr_SetString(PyExc_OverflowError, "int too large to convert to i16");
+    PyErr_SetString(PyExc_ValueError, "int too large to convert to i16");
 }
 
 // u8 unboxing slow path
@@ -567,15 +567,15 @@ uint8_t CPyLong_AsUInt8_(PyObject *o) {
         if (PyErr_Occurred()) {
             return CPY_LL_UINT_ERROR;
         } else if (overflow) {
-            PyErr_SetString(PyExc_OverflowError, "int too large or small to convert to u8");
+            PyErr_SetString(PyExc_ValueError, "int too large or small to convert to u8");
             return CPY_LL_UINT_ERROR;
         }
     }
     return result;
 }
 
 void CPyUInt8_Overflow() {
-    PyErr_SetString(PyExc_OverflowError, "int too large or small to convert to u8");
+    PyErr_SetString(PyExc_ValueError, "int too large or small to convert to u8");
 }
 
 double CPyTagged_TrueDivide(CPyTagged x, CPyTagged y) {

mypyc/test-data/run-i16.test

@@ -15,10 +15,10 @@ def test_box_and_unbox() -> None:
         o2: Any = x
         assert o == o2
         assert x == i
-    with assertRaises(OverflowError, "int too large to convert to i16"):
+    with assertRaises(ValueError, "int too large to convert to i16"):
         o = 2**15
         x2: i16 = o
-    with assertRaises(OverflowError, "int too large to convert to i16"):
+    with assertRaises(ValueError, "int too large to convert to i16"):
         o = -2**15 - 1
         x3: i16 = o
 
@@ -209,13 +209,13 @@ def test_mixed_comparisons() -> None:
 
     int_too_big = int() + (1 << 15)
     int_too_small = int() - (1 << 15) - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i16_3 < int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_big < i16_3
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i16_3 > int_too_small
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small < i16_3
 
 def test_mixed_arithmetic_and_bitwise_ops() -> None:
@@ -235,9 +235,9 @@ def test_mixed_arithmetic_and_bitwise_ops() -> None:
 
     int_too_big = int() + (1 << 15)
     int_too_small = int() - (1 << 15) - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i16_3 & int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small & i16_3
 
 def test_coerce_to_and_from_int() -> None:
@@ -277,11 +277,11 @@ def test_explicit_conversion_overflow() -> None:
     assert int(y) == min_i16
 
     too_big = int() + 2**15
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = i16(too_big)
 
     too_small = int() - 2**15 - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = i16(too_small)
 
 def test_i16_from_large_small_literal() -> None:
@@ -320,9 +320,9 @@ def test_explicit_conversion_from_float() -> None:
     assert from_float(2**15 - 1) == 2**15 - 1
     assert from_float(-2**15) == -2**15
     # The error message could be better, but this is acceptable
-    with assertRaises(OverflowError, "int too large to convert to i16"):
+    with assertRaises(ValueError, "int too large to convert to i16"):
         assert from_float(float(2**15))
-    with assertRaises(OverflowError, "int too large to convert to i16"):
+    with assertRaises(ValueError, "int too large to convert to i16"):
         # One ulp below the lowest valid i64 value
         from_float(float(-2**15 - 1))
 

mypyc/test-data/run-i32.test

@@ -15,10 +15,10 @@ def test_box_and_unbox() -> None:
         o2: Any = x
         assert o == o2
         assert x == i
-    with assertRaises(OverflowError, "int too large to convert to i32"):
+    with assertRaises(ValueError, "int too large to convert to i32"):
         o = 2**31
         x2: i32 = o
-    with assertRaises(OverflowError, "int too large to convert to i32"):
+    with assertRaises(ValueError, "int too large to convert to i32"):
         o = -2**32 - 1
         x3: i32 = o
 
@@ -209,13 +209,13 @@ def test_mixed_comparisons() -> None:
 
     int_too_big = int() + (1 << 31)
     int_too_small = int() - (1 << 31) - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i32_3 < int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_big < i32_3
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i32_3 > int_too_small
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small < i32_3
 
 def test_mixed_arithmetic_and_bitwise_ops() -> None:
@@ -235,9 +235,9 @@ def test_mixed_arithmetic_and_bitwise_ops() -> None:
 
     int_too_big = int() + (1 << 31)
     int_too_small = int() - (1 << 31) - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i32_3 & int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small & i32_3
 
 def test_coerce_to_and_from_int() -> None:
@@ -281,11 +281,11 @@ def test_explicit_conversion_overflow() -> None:
     assert int(y) == min_i32
 
     too_big = int() + 2**31
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = i32(too_big)
 
     too_small = int() - 2**31 - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = i32(too_small)
 
 def test_i32_from_large_small_literal() -> None:
@@ -318,9 +318,9 @@ def test_explicit_conversion_from_float() -> None:
     assert from_float(2**31 - 1) == 2**31 - 1
     assert from_float(-2**31) == -2**31
     # The error message could be better, but this is acceptable
-    with assertRaises(OverflowError, "int too large to convert to i32"):
+    with assertRaises(ValueError, "int too large to convert to i32"):
         assert from_float(float(2**31))
-    with assertRaises(OverflowError, "int too large to convert to i32"):
+    with assertRaises(ValueError, "int too large to convert to i32"):
         # One ulp below the lowest valid i64 value
         from_float(float(-2**31 - 2048))
 

mypyc/test-data/run-i64.test

@@ -299,11 +299,11 @@ def test_explicit_conversion_overflow() -> None:
     assert int(y) == min_i64
 
     too_big = int() + 2**63
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = i64(too_big)
 
     too_small = int() - 2**63 - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = i64(too_small)
 
 def test_i64_from_large_small_literal() -> None:
@@ -323,9 +323,9 @@ def test_explicit_conversion_from_float() -> None:
     assert from_float(2**63 - 1024) == 2**63 - 1024
     assert from_float(-2**63) == -2**63
     # The error message could be better, but this is acceptable
-    with assertRaises(OverflowError, "int too large to convert to i64"):
+    with assertRaises(ValueError, "int too large to convert to i64"):
         assert from_float(float(2**63))
-    with assertRaises(OverflowError, "int too large to convert to i64"):
+    with assertRaises(ValueError, "int too large to convert to i64"):
         # One ulp below the lowest valid i64 value
         from_float(float(-2**63 - 2048))
 
@@ -430,13 +430,13 @@ def test_mixed_comparisons() -> None:
 
     int_too_big = int() + (1 << 63)
     int_too_small = int() - (1 << 63) - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i64_3 < int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_big < i64_3
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i64_3 > int_too_small
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small < i64_3
 
 def test_mixed_comparisons_32bit() -> None:
@@ -476,9 +476,9 @@ def test_mixed_arithmetic_and_bitwise_ops() -> None:
 
     int_too_big = int() + (1 << 63)
     int_too_small = int() - (1 << 63) - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert i64_3 & int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small & i64_3
 
 def test_for_loop() -> None:
@@ -591,10 +591,10 @@ def test_unbox_int_fails() -> None:
     with assertRaises(TypeError, msg):
         x: i64 = o
     o2: Any = 1 << 63
-    with assertRaises(OverflowError, "int too large to convert to i64"):
+    with assertRaises(ValueError, "int too large to convert to i64"):
         y: i64 = o2
     o3: Any = -(1 << 63 + 1)
-    with assertRaises(OverflowError, "int too large to convert to i64"):
+    with assertRaises(ValueError, "int too large to convert to i64"):
         z: i64 = o3
 
 class Uninit:

mypyc/test-data/run-u8.test

@@ -14,10 +14,10 @@ def test_box_and_unbox() -> None:
         o2: Any = x
         assert o == o2
         assert x == i
-    with assertRaises(OverflowError, "int too large or small to convert to u8"):
+    with assertRaises(ValueError, "int too large or small to convert to u8"):
         o = 256
         x2: u8 = o
-    with assertRaises(OverflowError, "int too large or small to convert to u8"):
+    with assertRaises(ValueError, "int too large or small to convert to u8"):
         o = -1
         x3: u8 = o
 
@@ -166,13 +166,13 @@ def test_mixed_comparisons() -> None:
 
     int_too_big = int() + 256
     int_too_small = int() -1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert u8_3 < int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_big < u8_3
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert u8_3 > int_too_small
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small < u8_3
 
 def test_mixed_arithmetic_and_bitwise_ops() -> None:
@@ -192,9 +192,9 @@ def test_mixed_arithmetic_and_bitwise_ops() -> None:
 
     int_too_big = int() + 256
     int_too_small = int() - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert u8_3 & int_too_big
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         assert int_too_small & u8_3
 
 def test_coerce_to_and_from_int() -> None:
@@ -233,11 +233,11 @@ def test_explicit_conversion_overflow() -> None:
     assert int(y) == min_u8
 
     too_big = int() + 256
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = u8(too_big)
 
     too_small = int() - 1
-    with assertRaises(OverflowError):
+    with assertRaises(ValueError):
         x = u8(too_small)
 
 def test_u8_from_large_small_literal() -> None:
@@ -277,9 +277,9 @@ def test_explicit_conversion_from_float() -> None:
     assert from_float(0) == 0
     assert from_float(-0.999) == 0
     # The error message could be better, but this is acceptable
-    with assertRaises(OverflowError, "int too large or small to convert to u8"):
+    with assertRaises(ValueError, "int too large or small to convert to u8"):
         assert from_float(float(256))
-    with assertRaises(OverflowError, "int too large or small to convert to u8"):
+    with assertRaises(ValueError, "int too large or small to convert to u8"):
         # One ulp below the lowest valid i64 value
         from_float(float(-1.0))