diff --git a/Mathlib.lean b/Mathlib.lean
index 2eb0fe3bc67d02..6cc203f72f99e9 100644
--- a/Mathlib.lean
+++ b/Mathlib.lean
@@ -385,6 +385,7 @@ import Mathlib.Data.Rat.Sqrt
 import Mathlib.Data.Real.Basic
 import Mathlib.Data.Real.CauSeq
 import Mathlib.Data.Real.CauSeqCompletion
+import Mathlib.Data.Real.Sign
 import Mathlib.Data.Rel
 import Mathlib.Data.Semiquot
 import Mathlib.Data.Seq.Computation
diff --git a/Mathlib/Data/Real/Sign.lean b/Mathlib/Data/Real/Sign.lean
new file mode 100644
index 00000000000000..b0cb5e00c5d7e3
--- /dev/null
+++ b/Mathlib/Data/Real/Sign.lean
@@ -0,0 +1,129 @@
+/-
+Copyright (c) 2021 Kexing Ying. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Kexing Ying, Eric Wieser
+
+! This file was ported from Lean 3 source module data.real.sign
+! leanprover-community/mathlib commit 9003f28797c0664a49e4179487267c494477d853
+! Please do not edit these lines, except to modify the commit id
+! if you have ported upstream changes.
+-/
+import Mathlib.Data.Real.Basic
+
+/-!
+# Real sign function
+
+This file introduces and contains some results about `Real.sign` which maps negative
+real numbers to -1, positive real numbers to 1, and 0 to 0.
+
+## Main definitions
+
+ * `Real.sign r` is $\begin{cases} -1 & \text{if } r < 0, \\
+                               ~~\, 0 & \text{if } r = 0, \\
+                               ~~\, 1 & \text{if } r > 0. \end{cases}$
+
+## Tags
+
+sign function
+-/
+
+
+namespace Real
+
+/-- The sign function that maps negative real numbers to -1, positive numbers to 1, and 0
+otherwise. -/
+noncomputable def sign (r : ℝ) : ℝ :=
+  if r < 0 then -1 else if 0 < r then 1 else 0
+#align real.sign Real.sign
+
+theorem sign_of_neg {r : ℝ} (hr : r < 0) : sign r = -1 := by rw [sign, if_pos hr]
+#align real.sign_of_neg Real.sign_of_neg
+
+theorem sign_of_pos {r : ℝ} (hr : 0 < r) : sign r = 1 := by rw [sign, if_pos hr, if_neg hr.not_lt]
+#align real.sign_of_pos Real.sign_of_pos
+
+@[simp]
+theorem sign_zero : sign 0 = 0 := by rw [sign, if_neg (lt_irrefl _), if_neg (lt_irrefl _)]
+#align real.sign_zero Real.sign_zero
+
+@[simp]
+theorem sign_one : sign 1 = 1 :=
+  sign_of_pos <| by norm_num
+#align real.sign_one Real.sign_one
+
+theorem sign_apply_eq (r : ℝ) : sign r = -1 ∨ sign r = 0 ∨ sign r = 1 := by
+  obtain hn | rfl | hp := lt_trichotomy r (0 : ℝ)
+  · exact Or.inl <| sign_of_neg hn
+  · exact Or.inr <| Or.inl <| sign_zero
+  · exact Or.inr <| Or.inr <| sign_of_pos hp
+#align real.sign_apply_eq Real.sign_apply_eq
+
+/-- This lemma is useful for working with `ℝˣ` -/
+theorem sign_apply_eq_of_ne_zero (r : ℝ) (h : r ≠ 0) : sign r = -1 ∨ sign r = 1 := by
+  obtain hn | rfl | hp := lt_trichotomy r (0 : ℝ)
+  · exact Or.inl <| sign_of_neg hn
+  · exact (h rfl).elim
+  · exact Or.inr <| sign_of_pos hp
+#align real.sign_apply_eq_of_ne_zero Real.sign_apply_eq_of_ne_zero
+
+@[simp]
+theorem sign_eq_zero_iff {r : ℝ} : sign r = 0 ↔ r = 0 := by
+  refine' ⟨fun h => _, fun h => h.symm ▸ sign_zero⟩
+  obtain hn | rfl | hp := lt_trichotomy r (0 : ℝ)
+  · rw [sign_of_neg hn, neg_eq_zero] at h
+    exact (one_ne_zero h).elim
+  · rfl
+  · rw [sign_of_pos hp] at h
+    exact (one_ne_zero h).elim
+#align real.sign_eq_zero_iff Real.sign_eq_zero_iff
+
+theorem sign_int_cast (z : ℤ) : sign (z : ℝ) = ↑(Int.sign z) := by
+  obtain hn | rfl | hp := lt_trichotomy z (0 : ℤ)
+  · rw [sign_of_neg (Int.cast_lt_zero.mpr hn), Int.sign_eq_neg_one_of_neg hn, Int.cast_neg,
+      Int.cast_one]
+  · rw [Int.cast_zero, sign_zero, Int.sign_zero, Int.cast_zero]
+  · rw [sign_of_pos (Int.cast_pos.mpr hp), Int.sign_eq_one_of_pos hp, Int.cast_one]
+#align real.sign_int_cast Real.sign_int_cast
+
+theorem sign_neg {r : ℝ} : sign (-r) = -sign r := by
+  obtain hn | rfl | hp := lt_trichotomy r (0 : ℝ)
+  · rw [sign_of_neg hn, sign_of_pos (neg_pos.mpr hn), neg_neg]
+  · rw [sign_zero, neg_zero, sign_zero]
+  · rw [sign_of_pos hp, sign_of_neg (neg_lt_zero.mpr hp)]
+#align real.sign_neg Real.sign_neg
+
+theorem sign_mul_nonneg (r : ℝ) : 0 ≤ sign r * r := by
+  obtain hn | rfl | hp := lt_trichotomy r (0 : ℝ)
+  · rw [sign_of_neg hn]
+    exact mul_nonneg_of_nonpos_of_nonpos (by norm_num) hn.le
+  · rw [mul_zero]
+  · rw [sign_of_pos hp, one_mul]
+    exact hp.le
+#align real.sign_mul_nonneg Real.sign_mul_nonneg
+
+theorem sign_mul_pos_of_ne_zero (r : ℝ) (hr : r ≠ 0) : 0 < sign r * r := by
+  refine' lt_of_le_of_ne (sign_mul_nonneg r) fun h => hr _
+  have hs0 := (zero_eq_mul.mp h).resolve_right hr
+  exact sign_eq_zero_iff.mp hs0
+#align real.sign_mul_pos_of_ne_zero Real.sign_mul_pos_of_ne_zero
+
+@[simp]
+theorem inv_sign (r : ℝ) : (sign r)⁻¹ = sign r := by
+  obtain hn | hz | hp := sign_apply_eq r
+  · rw [hn]
+    norm_num
+  · rw [hz]
+    exact inv_zero
+  · rw [hp]
+    exact inv_one
+#align real.inv_sign Real.inv_sign
+
+@[simp]
+theorem sign_inv (r : ℝ) : sign r⁻¹ = sign r := by
+  obtain hn | rfl | hp := lt_trichotomy r (0 : ℝ)
+  · rw [sign_of_neg hn, sign_of_neg (inv_lt_zero.mpr hn)]
+  · rw [sign_zero, inv_zero, sign_zero]
+  · rw [sign_of_pos hp, sign_of_pos (inv_pos.mpr hp)]
+#align real.sign_inv Real.sign_inv
+
+end Real