René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"[ENH]: Should we hide _preprocess_data from the stack trace?","articleBody":"### Problem\n\nWhen a method with `@_preprocess_data` raises an exception the decorator shows up in the call stack:\n\n```\nValueError                                Traceback (most recent call last)\nCell In[7], line 6\n      4 fig, ax = plt.subplots()\n      5 data = [10, 20, float('nan'), 40]\n----\u003e 6 ax.pie(data, labels=[\"A\", \"B\", \"C\", \"D\"])\n      7 plt.show()\n\nFile ~\\AppData\\Local\\miniforge3\\envs\\jupyter\\Lib\\site-packages\\matplotlib\\__init__.py:1473, in _preprocess_data.\u003clocals\u003e.inner(ax, data, *args, **kwargs)\n   1470 @functools.wraps(func)\n   1471 def inner(ax, *args, data=None, **kwargs):\n   1472     if data is None:\n-\u003e 1473         return func(\n   1474             ax,\n   1475             *map(sanitize_sequence, args),\n   1476             **{k: sanitize_sequence(v) for k, v in kwargs.items()})\n   1478     bound = new_sig.bind(ax, *args, **kwargs)\n   1479     auto_label = (bound.arguments.get(label_namer)\n   1480                   or bound.kwargs.get(label_namer))\n\nFile ~\\AppData\\Local\\miniforge3\\envs\\jupyter\\Lib\\site-packages\\matplotlib\\axes\\_axes.py:3334, in Axes.pie(self, x, explode, labels, colors, autopct, pctdistance, shadow, labeldistance, startangle, radius, counterclock, wedgeprops, textprops, center, frame, rotatelabels, normalize, hatch)\n   3331 x += expl * math.cos(thetam)\n   3332 y += expl * math.sin(thetam)\n-\u003e 3334 w = mpatches.Wedge((x, y), radius, 360. * min(theta1, theta2),\n   3335                    360. * max(theta1, theta2),\n   3336                    facecolor=get_next_color(),\n   3337                    hatch=next(hatch_cycle),\n   3338                    clip_on=False,\n   3339                    label=label)\n   3340 w.set(**wedgeprops)\n   3341 slices.append(w)\n```\n\nThis is rather distracting and makes the stacktrace harder to read.\n\n### Proposed solution\n\nWe could filter the decorator out of the stacktrace like this\n\n```\ndef hidden_decorator(func):\n    @functools.wraps(func)\n    def wrapper(*args, **kwargs):\n        try:\n            return func(*args, **kwargs)\n        except Exception as e:\n            # Modify the traceback to remove the wrapper from the stack trace\n            tb = traceback.extract_tb(sys.exc_info()[2])\n            # Filter out the wrapper's frame\n            filtered_tb = [frame for frame in tb if frame.name != wrapper.__name__]\n            # Raise the exception with the modified traceback\n            raise e.with_traceback(traceback.extract_stack(filtered_tb))\n    return wrapper\n```\n\nIs this a reasonable idea? On the one hand, it's removing irrelevant information (the decorator is an implemenation detail that should not bother the user or anybody debugging internal errors. On the other hand, a stack trace is a technical detail and it may be better to tell the full story there.","author":{"url":"https://github.com/timhoffm","@type":"Person","name":"timhoffm"},"datePublished":"2025-04-03T12:00:58.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":6},"url":"https://github.com/29863/matplotlib/issues/29863"}