René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"[Bug]: matplotlib.path.Path.to_polygons fails with TriContourSet paths","articleBody":"### Bug summary\n\nThe [Path](https://matplotlib.org/stable/api/path_api.html#matplotlib.path.Path) objects produced by [tricontourf](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.tricontourf.html) to draw its contours lead to spurious polygons when running the respective `to_polygons` method.\r\n\n\n### Code for reproduction\n\n```python\nimport numpy as np\r\nimport matplotlib as mpl\r\nimport matplotlib.pyplot as plt\r\nimport matplotlib.tri\r\n\r\n# Example for TriContour based from:\r\n# https://matplotlib.org/2.0.2/examples/pylab_examples/tricontour_smooth_delaunay.html\r\n\r\n\r\ndef experiment_res(x, y):\r\n    \"\"\" An analytic function representing experiment results \"\"\"\r\n    x = 2.*x\r\n    r1 = np.sqrt((0.5 - x)**2 + (0.5 - y)**2)\r\n    theta1 = np.arctan2(0.5 - x, 0.5 - y)\r\n    r2 = np.sqrt((-x - 0.2)**2 + (-y - 0.2)**2)\r\n    theta2 = np.arctan2(-x - 0.2, -y - 0.2)\r\n    z = (4*(np.exp((r1/10)**2) - 1)*30. * np.cos(3*theta1) +\r\n         (np.exp((r2/10)**2) - 1)*30. * np.cos(5*theta2) +\r\n         2*(x**2 + y**2))\r\n    return (np.max(z) - z)/(np.max(z) - np.min(z))\r\n\r\nn = 200\r\nrandom_gen = np.random.mtrand.RandomState(seed=127260)\r\nx = random_gen.uniform(-1., 1., size=n)\r\ny = random_gen.uniform(-1., 1., size=n)\r\nv = experiment_res(x, y)\r\n\r\n## triangulate\r\ntri = mpl.tri.Triangulation(x, y)\r\nmask = mpl.tri.TriAnalyzer(tri).get_flat_tri_mask(min_circle_ratio=.01)\r\ntri.set_mask(mask)\r\n\r\n## refine triangular mesh\r\nrefiner = mpl.tri.UniformTriRefiner(tri)\r\ntriFiner, vFiner = refiner.refine_field(v, subdiv=2)  # uses CubicTriInterpolator\r\n\r\n## plot the mesh\r\nplt.close('all')\r\nplt.triplot(tri, color='k', lw=.8)\r\nplt.triplot(triFiner, color='k', alpha=.3, lw=.5)\r\nplt.show()\r\n\r\n## plot and create TriCounterSet object\r\nlevels = [.7, .9, 1]\r\ncmap = mpl.cm.plasma\r\nplt.close('all')\r\ncl = plt.tricontour(triFiner, vFiner, levels=levels, colors='k', linewidths=2, linestyles='--')\r\ncf = plt.tricontourf(triFiner, vFiner, levels=levels, cmap=cmap, extend='both', alpha=.8)\r\n_ = [collection.set_edgecolor(\"face\") for collection in cf.collections]  # https://stackoverflow.com/a/73805556/921580\r\nplt.triplot(tri, color='k', lw=1.5, alpha=.1)  # refined mesh\r\nplt.triplot(triFiner, color='k', lw=.5, alpha=.1)  # original mesh\r\ncb = plt.colorbar(cf)\r\nplt.show()\r\n\r\n## retrieve contours from TriContourSet (paths) and convert to polygons\r\nfor i in range(len(cf.collections)):\r\n    plt.gca().set_xlim(-1.1, 1.1)\r\n    plt.gca().set_ylim(-1.1, 1.1)\r\n    [plt.gca().add_patch(\r\n            mpl.patches.PathPatch(\r\n                cf.collections[i].get_paths()[j],\r\n                facecolor=cf.collections[i].get_facecolor()\r\n            )\r\n        ) for j in range(len(cf.collections[i].get_paths()))]\r\n    [plt.plot(*list(zip(*p)), c='red', ls='--') \r\n            for path in cf.collections[i].get_paths()\r\n            for p in path.to_polygons()\r\n        ]\r\n    if i == (len(cf.collections) - 2):\r\n        plt.savefig('bug_actual_outcome.png')\r\n    plt.show()\r\n\r\n## compare path vertices with output from to_polygons method\r\npath = cf.collections[-2].get_paths()[0]\r\nprint(\"\\npath vertices and codes:\\n\", path)\r\nprint(\"\\npath.to_polygons() = \\n\", path.to_polygons())\n```\n\n\n### Actual outcome\n\n![bug_actual_outcome](https://user-images.githubusercontent.com/1119116/215825498-d59e4fbc-a439-448d-bfff-e9c27b07b8f8.png)\n\n### Expected outcome\n\n![bug_expected_outcome](https://user-images.githubusercontent.com/1119116/215825869-a279a2e5-d19f-47ca-b6af-1741b503b7a5.png)\r\n\n\n### Additional information\n\nThe issue seems to be related to the [Path](https://matplotlib.org/stable/api/path_api.html#matplotlib.path.Path) object and its `to_polygons` method. More specifically the problem may happen when a Path contains multiple polylines but is defined without the `CLOSEPOLY` codes ending each polyline. When producing contour plots the code can cope with this data without problems, but the `to_polygons` method cannot.\r\n\r\nI have tested in both 3.4.1 and 3.0.2 versions and both have the same problem.\r\n\r\nI have a fix where I defined a function `path2polygons`:\r\n\r\n```python\r\ndef path2polygons(path):\r\n    assert isinstance(path, mpl.path.Path), f\"Type of input is {type(path)} instead of Path.\"\r\n    codesOld = path.codes[path.codes != mpl.path.Path.CLOSEPOLY]\r\n    vertsOld = path.vertices[path.codes != mpl.path.Path.CLOSEPOLY]\r\n    jj = np.append(np.flatnonzero(codesOld == path.MOVETO), [len(codesOld)])\r\n    verts, codes = [], []\r\n    for ji, je in zip(jj[:-1], jj[1:]):\r\n        codes.extend(   [mpl.path.Path.MOVETO]\r\n                      + [mpl.path.Path.LINETO]*(je - ji - 1)\r\n                      + [mpl.path.Path.CLOSEPOLY]\r\n                    )\r\n        verts.extend( list(vertsOld[ji:je]) + [vertsOld[ji]] )\r\n    return mpl.path.Path(verts, codes).to_polygons()\r\n```\r\nbut is not tested with other Path codes such as `CURVE3` and `CURVE4`.\r\n\n\n### Operating system\n\nDebian 10\n\n### Matplotlib Version\n\n3.4.1\n\n### Matplotlib Backend\n\nTkAgg\n\n### Python version\n\n3.7.3\n\n### Jupyter version\n\n_No response_\n\n### Installation\n\npip","author":{"url":"https://github.com/cvr","@type":"Person","name":"cvr"},"datePublished":"2023-01-31T17:00:35.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":3},"url":"https://github.com/25114/matplotlib/issues/25114"}