如何在MongoDB中实现唯一的_id值？

Question

I am using Python2.7, Pymongo and MongoDB. I'm trying to get rid of the default _id values in MongoDB. Instead, I want certain fields of columns to go as _id.

For example:

{
    "_id" : ObjectId("568f7df5ccf629de229cf27b"),
    "LIFNR" : "10099",
    "MANDT" : "100",
    "BUKRS" : "2646",
    "NODEL" : "",
    "LOEVM" : ""
}

I would like to concatenate LIFNR+MANDT+BUKRS as 100991002646 and hash it to achieve uniqueness and store it as new _id.

But how far hashing helps for unique ids? And how do I achieve it?

I understood that using default hash function in Python gives different results for different machines (32 bit / 64 bit). If it is true, how would I go about generating _ids?

But I need LIFNR+MANDT+BUKRS to be used however. Thanks in advance.

Answer 1

First you can't update the _id field . Instead you should create a new field and set it value to the concatenated string. To return the concatenated value you need to use the .aggregate() method which provides access to the aggregation pipeline. The only stage in the pipeline is the $project stage where you use the $concat operator which concatenates strings and returns the concatenated string. From there you then iterate the cursor and update each document using "bulk" operations.

bulk = collection.initialize_ordered_bulk_op()
count = 0
cursor = collection.aggregate([
    {"$project": {"value": {"$concat": ["$LIFNR", "$MANDT", "$BUKRS"]}}}
])

for item in cursor:
    bulk.find({'_id': item['_id']}).update_one({'$set': {'id': item['value']}})
    count = count + 1
    if count % 200 == 0:
        bulk.execute()
if count > 0:
    bulk.execute()

---

MongoDB 3.2 deprecates Bulk() and its associated methods so you will need to use the bulk_write() method.

from pymongo import UpdateOne


requests = []
for item in cursor:
    requests.append(UpdateOne({'_id': item['_id']}, {'$set': {'id': item['value']}}))
collection.bulk_write(requests)

---

Your documents will then look like this:

{'BUKRS': '2646',
  'LIFNR': '10099',
  'LOEVM': '',
  'MANDT': '100',
  'NODEL': '',
  '_id': ObjectId('568f7df5ccf629de229cf27b'),
  'id': '100991002646'}

Answer 2

您可以使用自己的哈希函数，那么它就不会依赖于体系结构，更重要的是，您将知道它对变量的作用。