From 913f1724fff8fb84825e7c5127e670a3cc0a0e42 Mon Sep 17 00:00:00 2001
From: Thomas A Caswell <tcaswell@gmail.com>
Date: Sat, 3 Mar 2018 16:23:13 -0500
Subject: [PATCH] Backport PR #10613: FIX: resolution of imshow for floats and
 2-D greyscale PIL images

---
 lib/matplotlib/image.py                       |  28 +++++++++++++++++-
 .../test_image/imshow_bignumbers_real.png     | Bin 0 -> 3312 bytes
 lib/matplotlib/tests/test_image.py            |  25 +++++++++++++++-
 3 files changed, 51 insertions(+), 2 deletions(-)
 create mode 100644 lib/matplotlib/tests/baseline_images/test_image/imshow_bignumbers_real.png

diff --git a/lib/matplotlib/image.py b/lib/matplotlib/image.py
index de87c2afaa44..f2a7280af553 100644
--- a/lib/matplotlib/image.py
+++ b/lib/matplotlib/image.py
@@ -396,6 +396,28 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
                 # scaled data
                 A_scaled = np.empty(A.shape, dtype=scaled_dtype)
                 A_scaled[:] = A
+                # clip scaled data around norm if necessary.
+                # This is necessary for big numbers at the edge of
+                # float64's ability to represent changes.  Applying
+                # a norm first would be good, but ruins the interpolation
+                # of over numbers.
+                if self.norm.vmin is not None and self.norm.vmax is not None:
+                    dv = (np.float64(self.norm.vmax) -
+                            np.float64(self.norm.vmin))
+                    vmid = self.norm.vmin + dv / 2
+                    newmin = vmid - dv * 1.e7
+                    if newmin < a_min:
+                        newmin = None
+                    else:
+                        a_min = np.float64(newmin)
+                    newmax = vmid + dv * 1.e7
+                    if newmax > a_max:
+                        newmax = None
+                    else:
+                        a_max = np.float64(newmax)
+                    if newmax is not None or newmin is not None:
+                        A_scaled = np.clip(A_scaled, newmin, newmax)
+
                 A_scaled -= a_min
                 # a_min and a_max might be ndarray subclasses so use
                 # asscalar to ensure they are scalars to avoid errors
@@ -614,7 +636,11 @@ def set_data(self, A):
         """
         # check if data is PIL Image without importing Image
         if hasattr(A, 'getpixel'):
-            self._A = pil_to_array(A)
+            if A.mode == 'L':
+                # greyscale image, but our logic assumes rgba:
+                self._A = pil_to_array(A.convert('RGBA'))
+            else:
+                self._A = pil_to_array(A)
         else:
             self._A = cbook.safe_masked_invalid(A, copy=True)
 
diff --git a/lib/matplotlib/tests/baseline_images/test_image/imshow_bignumbers_real.png b/lib/matplotlib/tests/baseline_images/test_image/imshow_bignumbers_real.png
new file mode 100644
index 0000000000000000000000000000000000000000..ce3404488ebb88daece7f11dc5ba9519bccb6766
GIT binary patch
literal 3312
zcmeH~ZA_C_6vt1yrYpt@lO?F-r3t<yuC}Fu6b3C4bfTnnn+9u?7ok?B@DwOepl?7H
zFd`(fz&5bVEE&sYK&%O<Kp{Aa>$Ed$Rz@k*QajT^TRMeO+U~(G`?3$)M>bA!b8}Db
z$<00a-TyiF@^Q`w#7#Rk0RS=mNZLmLSOoxPmVie+aZe0$=!7dxO+P_EO+v_TK-WRS
zBfL_8;M}!=`Ig;PfI3fz4xJKxjGPx$<d*QEFjpkJgorK`oFkX>OG*onA~J)@pwdh7
zVyHr-RK$p*Q}g-d^!EfY^jQ7@Ml2(jLcSmp2@`3wzhbFKiGb!m8LI}^%t%i=ctYGU
zKXvfCGx{2hDf(rczV6PCIpf3q`N9@r{?B=|uj&UCkxN@X+4=|h_?2<`v2i@(D89y6
zbEEiC;mNyG^T*z=+K$^hzGG+RUdigRG|Dg{z2MYiZbVs+awAbgue_Zr2nLwd>;ep&
z!;vAS5pdA(jshSC8v^9RfQ1Ky6i{r#5FjLaW7EG&!B5k6O!LEg-FavAHm^<Jqsh#c
z#<5m@|JpV8u;sEA6!Kq2M@N@mv?F7hD0z~SRcQP)XqecfG54$0oyl)Of6c_pSe9q5
zWpe*F;zg&^Nw0p8=NcsqyFE!+1L7FJg)NU$?u*JegO|JAfzts$tv2e3L*|~E44oym
z1uBzSK8w6LvnT87bWwcu<9CZ7{!}ZUqFsEd>)aPbNRcrX7FJ7fveU{ysXLU!mRjUP
z7VfTf!-3&Zr1SHew#27pgG1TWw#yd*j8{o-Quu7XTrW`1=#)6P^#vGah0I2`M<V5P
zP1w@`+C8{|k=v9!UhCw3H0BCvok3oqdOK;i?Os#a*&aummj8SQC@Rc&VV&8<EWbbf
zYSOW!+Jk{hyCI$#zQrtIWp|<^Rd0pi^DGQ<i=ss-c)b}EOV9D^=b0~@4;E)UEsL<-
z%bKCJX&=M^IL3&xc#XN(UVh)EnsErYSa_BJMM7rym`5_i>Y6Bcy%Q9VBv@k(ueFz=
zwQombeju(Nq5E6FQJTd`O<u3JPv%Dg{`%fkd*^OSZJ*D@4J1}$2TWn=K=Yac^r(GB
zjV>Fi9*aTZx=|DTvDN)5;X3CH8!u2?_`I>(6qNvAAS`~JiT(x+{|~vl+nzid!_$}D
zv{gyfXQ-ii6joBaf)w7Y>7API*2&OH83Z}|G5sSZ;Wn$=v(mxonrs0$;6|DMj{vT-
zy0hheb%1?w>wv}UP`4*zjLwxQwoVC%*e93!?ys6mxr^5-l}Vq;mq>#WPl;g$t$i3U
z42NY)Vpv|g3RQs+1!>iZcmDc)rg?(Pj>wjIENGbwZA+G+HN?K;El#{X=C<kty2?8p
z=6R-es<^e!6pmOWtmZIUR%4IaUnN;K)vK%-rcftdd%Ka(Gp=}w`rP%d!%DV%fr1Pt
z7@JfB5rRt6V^_J7HSf=CSa5^``RH0FnN;mrv}h-zBUJ*O@~Vy<kQk;tU;2kSEQL(X
x+F51_GMZrI_$a<TA!yUtso3aB|8dXl$NpBEk@}(jf)@RqK>8t0T6bzr?Vm7^lmP$$

literal 0
HcmV?d00001

diff --git a/lib/matplotlib/tests/test_image.py b/lib/matplotlib/tests/test_image.py
index 16be520e325f..26e3b4a7ea26 100644
--- a/lib/matplotlib/tests/test_image.py
+++ b/lib/matplotlib/tests/test_image.py
@@ -511,6 +511,18 @@ def test_nonuniformimage_setnorm():
     im.set_norm(plt.Normalize())
 
 
+@needs_pillow
+def test_jpeg_2d():
+    # smoke test that mode-L pillow images work.
+    imd = np.ones((10, 10), dtype='uint8')
+    for i in range(10):
+        imd[i, :] = np.linspace(0.0, 1.0, 10) * 255
+    im = Image.new('L', (10, 10))
+    im.putdata(imd.flatten())
+    fig, ax = plt.subplots()
+    ax.imshow(im)
+
+
 @needs_pillow
 def test_jpeg_alpha():
     plt.figure(figsize=(1, 1), dpi=300)
@@ -855,7 +867,18 @@ def test_imshow_bignumbers():
     img = np.array([[1, 2, 1e12],[3, 1, 4]], dtype=np.uint64)
     pc = ax.imshow(img)
     pc.set_clim(0, 5)
-    plt.show()
+
+
+@image_comparison(baseline_images=['imshow_bignumbers_real'],
+                  remove_text=True, style='mpl20',
+                  extensions=['png'])
+def test_imshow_bignumbers_real():
+    # putting a big number in an array of integers shouldn't
+    # ruin the dynamic range of the resolved bits.
+    fig, ax = plt.subplots()
+    img = np.array([[2., 1., 1.e22],[4., 1., 3.]])
+    pc = ax.imshow(img)
+    pc.set_clim(0, 5)
 
 
 @pytest.mark.parametrize(