Merge pull request #250 from namannimmo10/overload

Overload all existing built-in functions

Author: namannimmo10

integration_tests/test_builtin_abs.py

@@ -1,4 +1,4 @@
-from ltypes import f64
+from ltypes import f32, f64, i32
 
 def test_abs():
     x: f64
@@ -9,5 +9,19 @@ def test_abs():
     assert abs(5.5) == 5.5
     assert abs(-5.5) == 5.5
 
+    x2: f32
+    x2 = -5.5
+    assert abs(x2) == 5.5
+
+    i: i32
+    i = -5
+    assert abs(i) == 5
+    assert abs(-1) == 1
+
+    b: bool
+    b = True
+    assert abs(b) == 1
+    b = False
+    assert abs(b) == 0
 
 test_abs()

integration_tests/test_builtin_pow.py

@@ -17,14 +17,39 @@ def test_pow():
     assert pow(a, b) == 1
     a = 2
     b = -1
-    assert abs(pow(2, -1) - 0.5) < eps
-    # assert abs(pow(a, b) - 0.5) < eps
+    print(pow(a, b))
     a = 6
     b = -4
+    print(pow(a, b))
+
+    a1: f64
+    a2: f64
+    a1 = 4.5
+    a2 = 2.3
+    assert abs(pow(a1, a2) - 31.7971929089206) < eps
+    assert abs(pow(a2, a1) - 42.43998894277659) < eps
+
+    x: i32
+    x = 3
+    y: f64
+    y = 2.3
+    assert abs(pow(x, y) - 12.513502532843182) < eps
+    assert abs(pow(y, x) - 12.166999999999998) < eps
+    assert abs(pow(x, 5.5) - 420.8883462392372) < eps
+
+    assert abs(pow(2, -1) - 0.5) < eps
     assert abs(pow(6, -4) - 0.0007716049382716049) < eps
-    # assert abs(pow(a, b) - 0.0007716049382716049) < eps
     assert abs(pow(-3, -5) + 0.00411522633744856) < eps
     assert abs(pow(6, -4) - 0.0007716049382716049) < eps
+    assert abs(pow(4.5, 2.3) - 31.7971929089206) < eps
+    assert abs(pow(2.3, 0.0) - 1.0) < eps
+    assert abs(pow(2.3, -1.5) - 0.2866871623459944) < eps
+    assert abs(pow(2, 3.4) - 10.556063286183154) < eps
+    assert abs(pow(2, -3.4) - 0.09473228540689989) < eps
+    assert abs(pow(3.4, 9) - 60716.99276646398) < eps
+    assert abs(pow(0.0, 53) - 0.0) < eps
+    assert pow(4, 2) == 16
+    assert abs(pow(-4235.0, 52) - 3.948003805985264e+188) < eps
 
 
 test_pow()

integration_tests/test_builtin_round.py

@@ -1,4 +1,4 @@
-from ltypes import f64
+from ltypes import i32, f32, f64
 
 def test_round():
     f: f64
@@ -22,5 +22,16 @@ def test_round():
     assert round(50.5) == 50
     assert round(56.78) == 57
 
+    i: i32
+    i = -5
+    assert round(i) == -5
+    assert round(4) == 4
+
+    b: bool
+    b = True
+    assert round(b) == 1
+    b = False
+    assert round(b) == 0
+    assert round(False) == 0
 
 test_round()

src/lpython/semantics/python_comptime_eval.h

@@ -97,26 +97,32 @@ struct PythonIntrinsicProcedures {
             ) {
         LFORTRAN_ASSERT(ASRUtils::all_args_evaluated(args));
         if (args.size() != 1) {
-            throw SemanticError("Intrinsic abs function accepts exactly 1 argument", loc);
+            throw SemanticError("abs() takes exactly one argument (" +
+                std::to_string(args.size()) + " given)", loc);
         }
-        ASR::expr_t* trig_arg = args[0];
+        ASR::expr_t* arg = args[0];
         ASR::ttype_t* t = ASRUtils::expr_type(args[0]);
-        if (ASR::is_a<ASR::Real_t>(*t)) {
-            double rv = ASR::down_cast<ASR::ConstantReal_t>(trig_arg)->m_r;
+        ASR::ttype_t *int_type = ASRUtils::TYPE(ASR::make_Integer_t(al, loc, 4, nullptr, 0));
+        ASR::ttype_t *real_type = ASRUtils::TYPE(ASR::make_Real_t(al, loc, 8, nullptr, 0));
+        if (ASRUtils::is_real(*t)) {
+            double rv = ASR::down_cast<ASR::ConstantReal_t>(arg)->m_r;
             double val = std::abs(rv);
-            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantReal_t(al, loc, val, t));
-        } else if (ASR::is_a<ASR::Integer_t>(*t)) {
-            int64_t rv = ASR::down_cast<ASR::ConstantInteger_t>(trig_arg)->m_n;
+            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantReal_t(al, loc, val, real_type));
+        } else if (ASRUtils::is_integer(*t)) {
+            int64_t rv = ASR::down_cast<ASR::ConstantInteger_t>(arg)->m_n;
             int64_t val = std::abs(rv);
-            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantInteger_t(al, loc, val, t));
-        } else if (ASR::is_a<ASR::Complex_t>(*t)) {
-            double re = ASR::down_cast<ASR::ConstantComplex_t>(trig_arg)->m_re;
-            double im = ASR::down_cast<ASR::ConstantComplex_t>(trig_arg)->m_im;
+            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantInteger_t(al, loc, val, int_type));
+        } else if (ASRUtils::is_logical(*t)) {
+            int8_t val = ASR::down_cast<ASR::ConstantLogical_t>(arg)->m_value;
+            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantInteger_t(al, loc, val, int_type));
+        } else if (ASRUtils::is_complex(*t)) {
+            double re = ASR::down_cast<ASR::ConstantComplex_t>(arg)->m_re;
+            double im = ASR::down_cast<ASR::ConstantComplex_t>(arg)->m_im;
             std::complex<double> x(re, im);
             double result = std::abs(x);
-            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantReal_t(al, loc, result, t));
+            return ASR::down_cast<ASR::expr_t>(ASR::make_ConstantReal_t(al, loc, result, real_type));
         } else {
-            throw SemanticError("Argument of the abs function must be Integer, Real or Complex", loc);
+            throw SemanticError("Argument of the abs function must be Integer, Real, Logical or Complex", loc);
         }
     }
 
@@ -249,15 +255,12 @@ struct PythonIntrinsicProcedures {
 
     static ASR::expr_t *eval_pow(Allocator &al, const Location &loc, Vec<ASR::expr_t*> &args) {
         LFORTRAN_ASSERT(ASRUtils::all_args_evaluated(args));
-        ASR::expr_t* arg1 = ASRUtils::expr_value(args[0]);
-        ASR::expr_t* arg2 = ASRUtils::expr_value(args[1]);
+        ASR::expr_t* arg1 = args[0];
+        ASR::expr_t* arg2 = args[1];
         ASR::ttype_t* arg1_type = ASRUtils::expr_type(arg1);
         ASR::ttype_t* arg2_type = ASRUtils::expr_type(arg2);
         ASR::ttype_t *int_type = ASRUtils::TYPE(ASR::make_Integer_t(al, loc, 4, nullptr, 0));
         ASR::ttype_t *real_type = ASRUtils::TYPE(ASR::make_Real_t(al, loc, 8, nullptr, 0));
-        if (!ASRUtils::check_equal_type(arg1_type, arg2_type)) {
-            throw SemanticError("The arguments to pow() must have the same type.", loc);
-        }
         if (ASRUtils::is_integer(*arg1_type) && ASRUtils::is_integer(*arg2_type)) {
             int64_t a = ASR::down_cast<ASR::ConstantInteger_t>(arg1)->m_n;
             int64_t b = ASR::down_cast<ASR::ConstantInteger_t>(arg2)->m_n;
@@ -271,8 +274,35 @@ struct PythonIntrinsicProcedures {
                 return ASR::down_cast<ASR::expr_t>(make_ConstantInteger_t(al, loc,
                     (int64_t)pow(a, b), int_type));
 
+        } else if (ASRUtils::is_real(*arg1_type) && ASRUtils::is_real(*arg2_type)) {
+            double a = ASR::down_cast<ASR::ConstantReal_t>(arg1)->m_r;
+            double b = ASR::down_cast<ASR::ConstantReal_t>(arg2)->m_r;
+            if (a == 0.0 && b < 0.0) { // Zero Division
+                throw SemanticError("0.0 cannot be raised to a negative power.", loc);
+            }
+            return ASR::down_cast<ASR::expr_t>(make_ConstantReal_t(al, loc,
+                pow(a, b), real_type));
+
+        } else if (ASRUtils::is_integer(*arg1_type) && ASRUtils::is_real(*arg2_type)) {
+            int64_t a = ASR::down_cast<ASR::ConstantInteger_t>(arg1)->m_n;
+            double b = ASR::down_cast<ASR::ConstantReal_t>(arg2)->m_r;
+            if (a == 0 && b < 0.0) { // Zero Division
+                throw SemanticError("0.0 cannot be raised to a negative power.", loc);
+            }
+            return ASR::down_cast<ASR::expr_t>(make_ConstantReal_t(al, loc,
+                pow(a, b), real_type));
+
+        } else if (ASRUtils::is_real(*arg1_type) && ASRUtils::is_integer(*arg2_type)) {
+            double a = ASR::down_cast<ASR::ConstantReal_t>(arg1)->m_r;
+            int64_t b = ASR::down_cast<ASR::ConstantInteger_t>(arg2)->m_n;
+            if (a == 0.0 && b < 0) { // Zero Division
+                throw SemanticError("0.0 cannot be raised to a negative power.", loc);
+            }
+            return ASR::down_cast<ASR::expr_t>(make_ConstantReal_t(al, loc,
+                pow(a, b), real_type));
+
         } else {
-            throw SemanticError("The arguments to pow() must be of type integers for now.", loc);
+            throw SemanticError("The two arguments to pow() must be of type integer or float.", loc);
         }
     }
 
@@ -423,8 +453,14 @@ struct PythonIntrinsicProcedures {
             if (fabs(rv-rounded) == 0.5)
                 rounded = 2.0*round(rv/2.0);
             return ASR::down_cast<ASR::expr_t>(make_ConstantInteger_t(al, loc, rounded, type));
+        } else if (ASRUtils::is_integer(*t)) {
+            int64_t rv = ASR::down_cast<ASR::ConstantInteger_t>(expr)->m_n;
+            return ASR::down_cast<ASR::expr_t>(make_ConstantInteger_t(al, loc, rv, type));
+        } else if (ASRUtils::is_logical(*t)) {
+            int64_t rv = ASR::down_cast<ASR::ConstantLogical_t>(expr)->m_value;
+            return ASR::down_cast<ASR::expr_t>(make_ConstantInteger_t(al, loc, rv, type));
         } else {
-            throw SemanticError("round() argument must be float for now, not '" +
+            throw SemanticError("round() argument must be float, integer, or logical for now, not '" +
                 ASRUtils::type_to_str(t) + "'", loc);
         }
     }

src/runtime/lpython_builtin.py

@@ -28,8 +28,10 @@ def chr(i: i32) -> str:
 #        exit(1)
 
 
-# This is an implementation for f64.
-# TODO: implement abs() as a generic procedure, and implement for all types
+#: abs() as a generic procedure.
+#: supported types for argument:
+#: i32, f32, f64, bool, c32, c64
+@overload
 def abs(x: f64) -> f64:
     """
     Return the absolute value of `x`.
@@ -39,6 +41,35 @@ def abs(x: f64) -> f64:
     else:
         return -x
 
+@overload
+def abs(x: f32) -> f32:
+    if x >= 0.0:
+        return x
+    else:
+        return -x
+
+@overload
+def abs(x: i32) -> i64:
+    if x >= 0:
+        return x
+    else:
+        return -x
+
+@overload
+def abs(b: bool) -> i32:
+    if b:
+        return 1
+    else:
+        return 0
+
+@overload
+def abs(c: c32) -> f32:
+    pass
+
+@overload
+def abs(c: c64) -> f64:
+    pass
+
 
 def str(x: i32) -> str:
     """
@@ -115,12 +146,30 @@ def len(s: str) -> i32:
     """
     pass
 
+#: pow() as a generic procedure.
+#: supported types for arguments:
+#: (i32, i32), (f64, f64), (i32, f64), (f64, i32)
+@overload
+def pow(x: i32, y: i32) -> i32:
+    """
+    Returns x**y.
+    """
+    return x**y
 
-def pow(x: i32, y: i32) -> f64:
+@overload
+def pow(x: f64, y: f64) -> f64:
     """
     Returns x**y.
     """
-    return 1.0*x**y
+    return x**y
+
+@overload
+def pow(x: i32, y: f64) -> f64:
+    return x**y
+
+@overload
+def pow(x: f64, y: i32) -> f64:
+    return x**y
 
 
 def int(f: f64) -> i32:
@@ -206,7 +255,10 @@ def oct(n: i32) -> str:
         res += _values[remainder]
     return prep + res[::-1]
 
-
+#: round() as a generic procedure.
+#: supported types for argument:
+#: i32, f64, bool
+@overload
 def round(value: f64) -> i32:
     """
     Rounds a floating point number to the nearest integer.
@@ -216,6 +268,14 @@ def round(value: f64) -> i32:
     else:
         return int(value) + 1
 
+@overload
+def round(value: i32) -> i64:
+    return value
+
+@overload
+def round(b: bool) -> i32:
+    return abs(b)
+
 def complex(x: f64, y: f64) -> c64:
     pass
 

tests/constants1.py

@@ -19,15 +19,15 @@ def test_ord_chr():
 
 
 def test_abs():
-    # a: i32
-    # a = abs(5)
-    # a = abs(-500)
-    # a = abs(False)
-    # a = abs(True)
+    a: i32
+    a = abs(5)
+    a = abs(-500)
+    a = abs(False)
+    a = abs(True)
     b: f32
     b = abs(3.45)
     b = abs(-5346.34)
-    # b = abs(complex(3.45, 5.6))
+    b = abs(complex(3.45, 5.6))
 
 
 def test_len():

tests/expr7.py

@@ -1,16 +1,15 @@
 def test_pow():
-    a: f64
+    a: i32
     a = pow(2, 2)
 
-
-def test_pow_1(a: i32, b: i32) -> f64:
-    res: f64
+def test_pow_1(a: i32, b: i32) -> i32:
+    res: i32
     res = pow(a, b)
     return res
 
 def main0():
     test_pow()
-    c: f64
+    c: i32
     c = test_pow_1(1, 2)
 
 main0()

tests/reference/asr-complex1-f26c460.json

@@ -6,7 +6,7 @@
     "outfile": null,
     "outfile_hash": null,
     "stdout": "asr-complex1-f26c460.stdout",
-    "stdout_hash": "ad6d981d374283dd344f860604121d3777ad15fd84cf52982d400ef4",
+    "stdout_hash": "219fe9f1acbedd5d18e1ed0d17c82f9a616e4b2e2b7ec8618fe43b1b",
     "stderr": null,
     "stderr_hash": null,
     "returncode": 0