René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Agg alpha image compositor differs from Matplotlib's software image compositor","articleBody":"On master.\r\n\r\nThe following test is taken from test_images.test_image_composite_alpha:\r\n```\r\nfrom pylab import *\r\n\r\nfig = plt.figure()\r\nax = fig.add_subplot(111)\r\narr = np.zeros((11, 21, 4))\r\narr[:, :, 0] = 1\r\narr[:, :, 3] = np.concatenate((np.arange(0, 1.1, 0.1), np.arange(0, 1, 0.1)[::-1]))\r\narr2 = np.zeros((21, 11, 4))\r\narr2[:, :, 0] = 1\r\narr2[:, :, 1] = 1\r\narr2[:, :, 3] = np.concatenate((np.arange(0, 1.1, 0.1), np.arange(0, 1, 0.1)[::-1]))[:, np.newaxis]\r\nax.imshow(arr, extent=[1, 2, 5, 0], alpha=0.3)\r\nax.imshow(arr, extent=[2, 3, 5, 0], alpha=0.6)\r\nax.imshow(arr, extent=[3, 4, 5, 0])\r\nax.imshow(arr2, extent=[0, 5, 1, 2])\r\nax.imshow(arr2, extent=[0, 5, 2, 3], alpha=0.6)\r\nax.imshow(arr2, extent=[0, 5, 3, 4], alpha=0.3)\r\nax.set_facecolor((0, 0.5, 0, 1))\r\nax.set_xlim([0, 5])\r\nax.set_ylim([5, 0])\r\n\r\nshow()\r\n```\r\nBy default, Agg produces the following result:\r\n![aggcomposed](https://user-images.githubusercontent.com/1322974/27981693-18e23188-6345-11e7-870e-43d515ac5ff9.png)\r\n\r\nMatplotlib can also force software composition of overlapping images, which is, I guess, faster for vector backends (only one image ends up written to the file).  For Agg, this is disabled; the Agg renderer does the compositing itself.  But you can switch Agg back to use software composition with the following patch:\r\n```\r\ndiff --git a/lib/matplotlib/backends/backend_agg.py b/lib/matplotlib/backends/backend_agg.py\r\nindex c6af41b19..580cf0cf8 100644\r\n--- a/lib/matplotlib/backends/backend_agg.py\r\n+++ b/lib/matplotlib/backends/backend_agg.py\r\n@@ -302,7 +302,7 @@ class RendererAgg(RendererBase):\r\n         # It is generally faster to composite each image directly to\r\n         # the Figure, and there's no file size benefit to compositing\r\n         # with the Agg backend\r\n-        return True\r\n+        return False\r\n \r\n     def option_scale_image(self):\r\n         \"\"\"\r\n```\r\nWith the patch on, the result is clearly different.\r\n![mplcomposed](https://user-images.githubusercontent.com/1322974/27981712-5fad6452-6345-11e7-9ebf-f2967ccc058b.png)\r\n\r\nI haven't looked much into which one is correct (it should be easy to manually compute the result), although https://github.com/anntzer/mpl_cairo seems to agree visually with Matplotlib, not Agg.\r\n","author":{"url":"https://github.com/anntzer","@type":"Person","name":"anntzer"},"datePublished":"2017-07-08T01:54:15.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":3},"url":"https://github.com/8847/matplotlib/issues/8847"}