MongoDB Shell与PyMongo聚合排序结果不一致问题排查

This is almost certainly due to dictionary ordering behavior in older Python versions (pre-3.6) breaking the sort order you defined. Here's the breakdown:

• In Mongo Shell, JSON objects preserve the order of keys. When you specify { "_id.category": 1, "count": -1 } as your sort stage, MongoDB executes the sort exactly in that order: first ascending by _id.category, then descending by count.
• But in Python versions before 3.6, regular dictionaries do not preserve insertion order. When you define your sort stage as a standard Python dict, the order of the keys might get rearranged (e.g., count could end up being processed first by MongoDB). This means the sort would prioritize count over _id.category, making it look like only the count is being sorted on.

Fixes to resolve this: #

1. Upgrade to Python 3.6+: Starting with Python 3.6, regular dictionaries maintain insertion order, so your existing pipeline will work as expected without changes.
2. Use collections.OrderedDict for the sort stage: If you can't upgrade Python, explicitly enforce the key order using OrderedDict:

   from collections import OrderedDict
   
   pipeline = [
       { "$unwind" : "$calls" },
       { "$group": { "_id": {"category": "$calls.category","api": "$calls.api"},"count": { "$sum": 1} } },
       { "$sort": OrderedDict([("_id.category", 1), ("count", -1)]) }
   ]
   output = list(db.DynamicData.aggregate(pipeline, allowDiskUse=True))
   

3. Verify the pipeline being sent: You can print the pipeline before executing it to confirm the sort key order is correct. For older Python versions, the standard dict might show the keys in a different order than you wrote them.

One quick sanity check: Make sure both the Mongo Shell and PyMongo are connecting to the exact same database/collection with identical data—though this is less likely to be the issue here given your description.

内容的提问来源于stack exchange，提问作者Simus